WEBVTT 00:00:03.537 --> 00:00:10.497 This is the Convergent Science Network podcast. Leading researchers in the domain 00:00:10.497 --> 00:00:16.777 of neuroscience, brain theory and technology are interviewed by Paul Verschure and Tony Prescott. 00:00:25.917 --> 00:00:29.437 This is Paul Verschure with the Convergent Science Network podcast. 00:00:30.017 --> 00:00:36.457 And I'm here with Tim Pearce, one of the speakers in our summer school, the 2013 version of it. 00:00:36.737 --> 00:00:43.557 And Tim is a great specialist in olfactory systems, both their biological versions 00:00:43.557 --> 00:00:44.917 and artificial versions. 00:00:46.337 --> 00:00:48.997 So, Tim, what's so special about natural olfaction? 00:00:51.657 --> 00:00:54.917 Great. Well, I was very pleased to be here. It's very exciting talks. 00:00:55.677 --> 00:01:01.497 So, natural olfaction. Yeah, well, one of the most amazing features, I guess. 00:01:01.617 --> 00:01:08.377 Well, there are many, but you have to consider pretty amazing the ability to 00:01:08.377 --> 00:01:11.497 detect literally thousands of molecules. 00:01:11.497 --> 00:01:14.137 Molecules and beyond that even to the 00:01:14.137 --> 00:01:17.417 point where if you made a new molecule this 00:01:17.417 --> 00:01:20.217 week you'd be able to if it 00:01:20.217 --> 00:01:22.997 was within certain mass criteria which is very 00:01:22.997 --> 00:01:25.997 wide anyway between 30 daltons and 00:01:25.997 --> 00:01:29.057 300 daltons in molecular mass then 00:01:29.057 --> 00:01:32.037 you would almost certainly have some kind 00:01:32.037 --> 00:01:36.057 of response to this so it has 00:01:36.057 --> 00:01:39.177 this well i think one of the most amazing properties 00:01:39.177 --> 00:01:42.457 of the olfactory system is this semi-biotic or foreignness 00:01:42.457 --> 00:01:45.337 property that it's designed from the 00:01:45.337 --> 00:01:49.217 ground up to detect uh all 00:01:49.217 --> 00:01:52.337 sorts of uh foreign signals that uh 00:01:52.337 --> 00:01:55.957 may not be uh have some priors right 00:01:55.957 --> 00:02:01.437 so your prior in the system is very flat you don't know necessarily in advance 00:02:01.437 --> 00:02:06.937 which uh chemical signals out there in the world are going to be uh extremely 00:02:06.937 --> 00:02:09.337 relevant for you behaviorally and 00:02:09.337 --> 00:02:13.297 you need to have as broad a sampling as possible on this without making. 00:02:14.017 --> 00:02:15.717 Necessarily prior assumptions. 00:02:16.277 --> 00:02:22.137 Right, but now as a functional feature, that's not specific to olfaction, right? 00:02:22.197 --> 00:02:26.977 I mean, also my visual system can detect novel stimuli or my auditory system and. 00:02:27.991 --> 00:02:31.351 So, that's not necessarily outstanding, is it? 00:02:32.251 --> 00:02:39.331 No, but maybe in the sense of just the diversity of the basic stimuli, right? 00:02:39.451 --> 00:02:45.371 I mean, and also, of course, I mean, this is also inherent in the architecture of the system, because, 00:02:45.491 --> 00:02:50.431 I mean, basically you've got three receptors or four receptors for vision in 00:02:50.431 --> 00:02:52.251 certain animals and humans, 00:02:52.251 --> 00:02:54.991 humans um and you know which has 00:02:54.991 --> 00:02:57.831 a certain cost uh in terms of the genetic real 00:02:57.831 --> 00:03:00.751 estate but when you look at olfaction it's more 00:03:00.751 --> 00:03:03.891 than one percent of the genetic real estate 00:03:03.891 --> 00:03:06.691 is dedicated just for this reason so there has 00:03:06.691 --> 00:03:14.591 to be a good reason to deploy all of that uh diverse coding uh so that you can 00:03:14.591 --> 00:03:19.111 have a very broad sampling if there was a simple answer to it i'm sure it would 00:03:19.111 --> 00:03:24.851 have been being found in the sort of genetic architecture, yeah. 00:03:25.071 --> 00:03:31.511 So now in your talk, you mentioned five outstanding features like concentration 00:03:31.511 --> 00:03:35.751 dynamics, temporal dynamics, the specificity and sensitivity. 00:03:37.271 --> 00:03:43.291 Also the ability to exploit olfaction in active senses, an active perceptual 00:03:43.291 --> 00:03:49.031 system, and to deal with object recognition in a high-dimensional space. 00:03:49.531 --> 00:03:53.191 Right so these were the five um outstanding issues 00:03:53.191 --> 00:03:56.651 and we will discuss those in a little bit more more detail 00:03:56.651 --> 00:04:01.071 but the other thing that was sort of surprising is that you also emphasize this 00:04:01.071 --> 00:04:06.351 so-called retronasal olfaction so what what what's the difference sure yeah 00:04:06.351 --> 00:04:10.751 so i mean most of us are are used to sniffing things in the world through so-called 00:04:10.751 --> 00:04:16.331 orthonormal uh olfaction which which is through the normal route through your external naras, 00:04:16.351 --> 00:04:22.391 but maybe less people are aware that a lot of your olfactory signals is coming 00:04:22.391 --> 00:04:27.871 when you're eating through volatile compounds going up through the back of the nose, 00:04:28.591 --> 00:04:32.251 interacting with the nasal mucosa that way. 00:04:33.891 --> 00:04:37.411 And, well, the interesting point is that depending on the flow, 00:04:37.531 --> 00:04:42.271 it seems that you get a different experience, whether the compounds are coming 00:04:42.271 --> 00:04:45.491 from internally during ingestion or whether they're coming externally. 00:04:45.831 --> 00:04:51.591 Even if you control for the fact that obviously eating things has a gustatory 00:04:51.591 --> 00:04:53.131 input as well, which is combined. 00:04:53.951 --> 00:04:58.971 But even if you control for that, then you still have a very different olfactory 00:04:58.971 --> 00:05:02.231 percept depending upon the direction of the odor. 00:05:03.351 --> 00:05:06.971 So does this retronasal olfaction, 00:05:06.971 --> 00:05:13.911 does it give it another quality of other sensation or other processing or does 00:05:13.911 --> 00:05:20.811 it fall the same back of molecular recognition and processing as the orthonormal version. 00:05:22.572 --> 00:05:28.692 Yeah, it's a good question. There have been many studies to show that the percept is very different. 00:05:28.932 --> 00:05:33.152 So during retronasal olfaction, you will perceive different qualities. 00:05:33.252 --> 00:05:37.972 It hasn't really been specified precisely how that changes. 00:05:38.252 --> 00:05:45.212 But we also know that the neural dynamics, even at the receptor sheet, 00:05:45.312 --> 00:05:49.732 is very different patterns that are elicited by exactly the same chemicals. 00:05:50.652 --> 00:05:54.872 Just from the fact whether they're occurring retronasally or orthonasally. 00:05:55.032 --> 00:06:02.312 So the obvious explanation for this must be to do with detecting whether things 00:06:02.312 --> 00:06:07.632 you're eating are good things to eat or externally whether they're things that 00:06:07.632 --> 00:06:10.532 you need to find to eat or avoid, I guess, right? 00:06:10.652 --> 00:06:15.232 So there does seem to be good reasons why this might be the case. Right. 00:06:15.592 --> 00:06:24.532 So now, one interesting observation that you put forward is also that in some 00:06:24.532 --> 00:06:28.852 sense human olfaction might not be of equal sensitivity as many other animals, 00:06:29.032 --> 00:06:34.752 but you do believe that the basic forms of processing and strategies in which 00:06:34.752 --> 00:06:36.492 we use olfaction might be rather similar. 00:06:36.832 --> 00:06:42.752 Is that fair to say? sure yeah i mean i it's pretty clear people like nick strausfeld 00:06:42.752 --> 00:06:44.452 and so on they've gotten into incredible 00:06:44.452 --> 00:06:48.112 detail in the fossil records for different olfactory systems from. 00:06:48.812 --> 00:06:55.732 Mammals all the way down to uh very extremely simple uh forms of forms of life 00:06:55.732 --> 00:07:01.012 have shown that there are a number of key uh generic architectural features 00:07:01.012 --> 00:07:05.332 which are crucial to to building olfactory systems you know one of these being 00:07:05.332 --> 00:07:07.772 uh glomeruli where there's these 00:07:07.852 --> 00:07:11.892 convergence of sensory signals at the first stage of processing, 00:07:12.092 --> 00:07:15.832 which is completely common to all animals achieving this. 00:07:15.952 --> 00:07:21.032 There also seems to be a genuine necessity for a first stage of processing that 00:07:21.032 --> 00:07:26.632 has lateral inhibition at an early stage, 00:07:26.772 --> 00:07:28.472 which somehow is thought to 00:07:28.472 --> 00:07:33.152 sharpen tuning and also impose all sorts of dynamical properties on top. 00:07:33.152 --> 00:07:40.972 And um there's also this feature in most animals of a very uh a mixed specificity of certain, 00:07:41.632 --> 00:07:45.832 receptors being highly tuned to certain compounds and other receptors being 00:07:45.832 --> 00:07:49.972 very widely tuned to large groups of compounds it seems to be this also seems 00:07:49.972 --> 00:07:53.732 to be common so yeah there are a number of properties that it seems if you're 00:07:53.732 --> 00:07:57.352 going to build um it's a nice thought experiment if you say take a blank piece 00:07:57.352 --> 00:07:59.512 of paper and think if we're going to build um. 00:08:00.282 --> 00:08:04.142 An olfactory system without any prior knowledge of what animals do, 00:08:04.402 --> 00:08:10.962 given that you want to detect huge diversity of compounds with certain high 00:08:10.962 --> 00:08:15.742 specificity and high sensitivity to certain groups of compounds that are important to you behaviorally, 00:08:15.802 --> 00:08:20.322 you know, what sort of strategies would you have to do to achieve that? 00:08:20.842 --> 00:08:26.742 And I think these common architectural sort of motifs that you you see uh from 00:08:26.742 --> 00:08:31.322 the fossil record sort of underpin how you might approach that also from an 00:08:31.322 --> 00:08:36.362 engineering sense so but then and so that basically means whether we whether 00:08:36.362 --> 00:08:38.622 we go from drosophila to humans, 00:08:39.342 --> 00:08:44.342 in your mind we will find common design principles behind this olfactory system 00:08:44.342 --> 00:08:49.702 right so that that's also interesting so it's how it would suggest it's rather strongly conserved. 00:08:51.302 --> 00:08:54.182 So now in the human case so here we go, we have 00:08:54.182 --> 00:08:57.902 our nasal cavity I have my epithelium in 00:08:57.902 --> 00:09:01.042 there through which I have the sensilla sticking from 00:09:01.042 --> 00:09:06.602 the olfactory receptor neurons these olfactory receptor neurons are then projecting 00:09:06.602 --> 00:09:12.402 their axons through the skull into the olfactory bulb where they form these 00:09:12.402 --> 00:09:17.542 glomeruli that you described and then these glomeruli in turn get read out by 00:09:17.542 --> 00:09:20.882 these mitral cells and then and sends this information to other parts of the brain. 00:09:21.242 --> 00:09:25.802 Now, in the mammalian case, 00:09:27.388 --> 00:09:30.848 What do we know about the levels of processing along this hierarchy? 00:09:31.268 --> 00:09:36.688 So, for instance, would you say the olfactory percept in the human case is already 00:09:36.688 --> 00:09:38.388 defined in the olfactory bulb? 00:09:38.788 --> 00:09:43.368 Or does it require higher levels of processing? Does it require interaction 00:09:43.368 --> 00:09:44.388 between different levels? 00:09:46.068 --> 00:09:51.408 Yeah, this is a really good question. I mean, in humans, it's pretty clear that 00:09:51.408 --> 00:09:55.868 the first stage of processing the olfactory bulb is thought to do a number of things. 00:09:55.868 --> 00:09:58.868 One of the most key features seems to be sharpening the representation, 00:09:59.348 --> 00:10:05.208 so decorrelating the signals at an early stage is obviously very important for 00:10:05.208 --> 00:10:10.088 discrimination of different odor compounds. 00:10:11.768 --> 00:10:15.848 But I'm not sure I would say this is the place where the percept comes from, 00:10:15.908 --> 00:10:18.848 because it clearly has to be an active process. 00:10:18.848 --> 00:10:21.788 And it's also demonstrated that in 00:10:21.788 --> 00:10:24.908 higher processing centers like the piriform cortex 00:10:24.908 --> 00:10:28.548 they have much more sparse representations of 00:10:28.548 --> 00:10:32.628 these odors which again is sharpening the 00:10:32.628 --> 00:10:35.748 tunings even further that has 00:10:35.748 --> 00:10:38.548 been already demonstrated quite clearly to be 00:10:38.548 --> 00:10:42.048 related to odor memories and also learning and 00:10:42.048 --> 00:10:45.948 the other important aspect is that these higher centers as we see in many other 00:10:45.948 --> 00:10:54.168 sensory systems in the brain are enforcing a top-down processing on the olfactory 00:10:54.168 --> 00:10:58.848 bulb itself on the earliest stage to change the gain control and do various things. 00:10:59.048 --> 00:11:05.108 So it's also been clear that, for instance, in orbitofrontal cortex, 00:11:05.208 --> 00:11:08.988 that this is also an important aspect of the perceptual aspect. 00:11:09.148 --> 00:11:14.308 So I would say it's really at these more cortical levels that if you were to 00:11:14.308 --> 00:11:17.208 think about percepts, this is where you'd probably look. 00:11:17.868 --> 00:11:23.768 So in the human case, how many of these receptor neurons do we have sticking their sensilla in my... 00:11:24.288 --> 00:11:29.468 Yeah, so the total real estate at its peak in rats was about a thousand different 00:11:29.468 --> 00:11:34.188 diversity of receptor GPCR types. 00:11:34.708 --> 00:11:41.888 And the story is that humans, basically our sense of smell is deteriorating in evolutionary terms. 00:11:42.008 --> 00:11:44.988 It's not under sufficient selective pressure. 00:11:45.448 --> 00:11:49.128 Basically and the story is that uh something like 00:11:49.128 --> 00:11:51.888 about 600 of those thousand receptors have become 00:11:51.888 --> 00:11:56.948 fairly defunctional now they've got introns within them junk dna and so on that 00:11:56.948 --> 00:12:00.628 means that they're they're they're not functioning as olfactory receptors anymore 00:12:00.628 --> 00:12:05.028 so we're left with in humans the story is somewhere between three and four hundred 00:12:05.028 --> 00:12:10.308 um functional is there any idea when in evolution this sort of started to happen um, 00:12:11.131 --> 00:12:14.351 Yeah, it's a good question. I don't know the answer to that in evolutionary terms. 00:12:14.531 --> 00:12:20.411 Yeah, but there are very nice studies that have been done which show phylogenetic. 00:12:20.571 --> 00:12:26.691 So you can, I'm not an expert on this, but you can look at the phylogenetic similarities. 00:12:26.991 --> 00:12:31.571 There have been a number of papers where effectively you're looking at sort 00:12:31.571 --> 00:12:35.171 of dendrogram of similarity of receptors between different animals. 00:12:35.791 --> 00:12:38.511 Now, for instance, primates, do they also show the same deterioration? 00:12:39.311 --> 00:12:41.891 Okay. So, primates are in a similar class to us. 00:12:42.751 --> 00:12:49.611 So, in the human case, I have, let's say, 300 receptor neuron types expressed. 00:12:49.971 --> 00:12:55.971 Then how many individual receptor neurons would I have in my epithelium? 00:12:56.231 --> 00:13:00.631 Yeah, so the epithelium has between 10 and 100 million in total. For humans? 00:13:00.931 --> 00:13:07.391 Yeah. So, you've got a large number of the same type expressed many times. And red, how many? 00:13:09.611 --> 00:13:12.691 They probably have similar numbers, slightly more with a higher density because 00:13:12.691 --> 00:13:14.391 it's obviously crammed into a smaller area. 00:13:14.551 --> 00:13:19.051 Things like dogs and pigs would have at least an order of magnitude greater 00:13:19.051 --> 00:13:21.151 than that in terms of numbers. 00:13:21.651 --> 00:13:24.711 And probably slightly more different types as well. 00:13:25.331 --> 00:13:31.871 Right. Okay, so now we have these 300 different types of receptor neurons. 00:13:32.691 --> 00:13:37.811 As a population, they're active. We also know that they are all highly specific 00:13:37.811 --> 00:13:41.711 in how they get integrated at the next stage in this glomeruli. 00:13:41.831 --> 00:13:47.271 Glomeruli are very much specific to the olfactory receptor neuron types, roughly. 00:13:48.911 --> 00:13:52.971 And now with that, also in the case of humans, let's say we can at least detect 00:13:52.971 --> 00:13:55.491 about 10,000 different molecules with this. 00:13:56.291 --> 00:14:01.551 Well, yeah, it's not so much molecules. There's this number of 10,000 sort of 00:14:01.551 --> 00:14:03.391 flies around olfaction. 00:14:03.391 --> 00:14:06.731 You're always hearing this number of 10,000. It has kind of a funny story, 00:14:06.811 --> 00:14:12.251 actually, that originally this number of 10,000 came from perfumers, 00:14:12.251 --> 00:14:16.031 I think, who were asked how many different, say, 00:14:16.231 --> 00:14:24.331 different perceptual qualities or different perceptual nuances could you have for a trained perfumer? 00:14:24.551 --> 00:14:28.291 And some perfumer just sort of plucked this number of 10,000. 00:14:28.311 --> 00:14:32.171 They thought there might be a possibility of 10,000 different perceptions that 00:14:32.171 --> 00:14:33.131 you might be able to have. 00:14:33.391 --> 00:14:37.411 Um, and that's about the limit of the evidence for it really. 00:14:37.491 --> 00:14:42.351 And that number has been sort of bandied, bandied about in, uh, olfaction ever since. 00:14:42.371 --> 00:14:47.411 So it doesn't necessarily correspond to the number of chemicals. Uh, and in fact. 00:14:48.157 --> 00:14:57.357 You can have almost an infinite variety of molecules that you could have a smell response to. 00:14:57.557 --> 00:15:03.637 But I'm sure in the ethology of olfaction, people must have made some sort of 00:15:03.637 --> 00:15:08.917 taxonomy of the whole collection of molecules that are detectable in principle 00:15:08.917 --> 00:15:10.577 by our olfactory system. 00:15:10.697 --> 00:15:13.457 So how large would that set be? Sure, yeah. How is that organized? 00:15:13.457 --> 00:15:18.097 We know at least a few thousand different chemically active, 00:15:18.377 --> 00:15:23.517 olfactory active molecules that would give you a response in some way. 00:15:23.517 --> 00:15:26.457 And how big a subset is that of all molecules? 00:15:28.417 --> 00:15:32.057 Well, all molecules, of course, that's including going way up to, 00:15:32.137 --> 00:15:37.897 you know, massive proteins, 30,000, you know, Daltons, you know, 00:15:37.917 --> 00:15:41.477 30 kilodaltons, enormous molecular machines, right? 00:15:41.477 --> 00:15:44.997 Because we're only looking at quite a narrow bandwidth. 00:15:45.497 --> 00:15:52.117 There's actually an interesting story about that. At our recent NICE Odor Maps workshop, 00:15:52.357 --> 00:15:57.957 we heard that someone did an analysis of the lower end of this quite narrow 00:15:57.957 --> 00:16:01.957 spectrum between 30 and 300 Daltons of what we can appreciate. 00:16:02.937 --> 00:16:05.777 And they looked at all the molecules down at the bottom 00:16:05.777 --> 00:16:10.037 end in terms of um uh in 00:16:10.037 --> 00:16:12.917 terms of our response them and they they looked 00:16:12.917 --> 00:16:18.197 at the molecules just under the threshold so just under 30 30 daltons and they 00:16:18.197 --> 00:16:23.097 called these um infra smells okay and then they looked at the molecules just 00:16:23.097 --> 00:16:28.277 above the 300th uh threshold and they called these ultra smells you know in 00:16:28.277 --> 00:16:30.817 line with vision as you can imagine and um, 00:16:31.872 --> 00:16:38.132 And it showed some interesting things. In fact, there seems to be a kind of 00:16:38.132 --> 00:16:43.132 story that these infra-smells may share with them a principle of kind of toxicity. 00:16:44.192 --> 00:16:49.312 That maybe these lighter molecules are getting towards an area of toxicity. 00:16:49.412 --> 00:16:56.512 And this may also be a guiding principle in terms of how our perception of odors may be organized. 00:16:56.752 --> 00:17:00.232 That we're obviously detecting these to be safe or not. 00:17:00.292 --> 00:17:06.712 Right. But it should be telling us something about also the specific niche for 00:17:06.712 --> 00:17:09.292 which we have been optimized, right? 00:17:09.352 --> 00:17:14.752 So why – because it's interesting that we're not sensitive to all possible molecules 00:17:14.752 --> 00:17:18.972 because actually in the big picture, it's actually a very small subset of all 00:17:18.972 --> 00:17:20.392 molecules we could ever encounter. 00:17:20.692 --> 00:17:26.032 So why is this subset then behaviorally and also from a perspective of fitness, 00:17:26.212 --> 00:17:28.352 evolutionary fitness? It's really relevant. 00:17:28.612 --> 00:17:31.952 Really a good question. i mean i think i think it's probably limited if 00:17:31.952 --> 00:17:34.672 uh by by most things is that once you get 00:17:34.672 --> 00:17:37.752 beyond 300 daltons things uh don't become 00:17:37.752 --> 00:17:40.672 volatile anymore right they're heavier molecules uh they 00:17:40.672 --> 00:17:43.752 become less volatile so basically your chances of 00:17:43.752 --> 00:17:48.012 being able to uh get these into the air to even be able to smell them at all 00:17:48.012 --> 00:17:51.692 starts to become very limited right so you're sort of limited by the physics 00:17:51.692 --> 00:17:58.232 now below uh below 30 daltons i'm not quite sure why that lower limit is right 00:17:58.232 --> 00:18:03.272 because because you certainly can have lighter molecules than that. 00:18:03.272 --> 00:18:07.212 Could it be a limitation of biophysics of the receptor neuron? 00:18:08.892 --> 00:18:12.612 For the lighter molecules? Yeah, well, I'm not sure. 00:18:12.692 --> 00:18:16.432 It may also be that many of these lighter molecules, 00:18:17.152 --> 00:18:21.412 you don't necessarily want to be detecting, like extremely light molecules, 00:18:21.832 --> 00:18:27.952 oxygen, CO2, and so on, are so common in the world that potentially maybe they 00:18:27.952 --> 00:18:29.052 would flood the whole system. 00:18:29.132 --> 00:18:32.232 So maybe you need to be more selective to these compounds. 00:18:32.732 --> 00:18:37.092 Although the common story in olfaction is that carbon dioxide doesn't have a smell. 00:18:37.132 --> 00:18:41.252 In fact, it does have a slight smell, and that may be through interacting with 00:18:41.252 --> 00:18:42.112 other molecular things. 00:18:42.292 --> 00:18:48.312 But probably these extremely light molecules, maybe there just isn't a behavioral 00:18:48.312 --> 00:18:52.092 reason for responding to them on the basis that they're there all the time. 00:18:52.092 --> 00:18:53.452 Yeah, but that's of course with the circular argument, right? 00:18:53.452 --> 00:18:55.992 Because you're saying, look, well, we don't smell them because we don't smell 00:18:55.992 --> 00:18:58.472 them. I mean, it's illogical in that way. 00:18:58.772 --> 00:19:02.192 But at the end of, but no, the argument may be, I mean, if they're there all 00:19:02.192 --> 00:19:05.792 the time, they don't give you any information, right? So in the end, it's okay. 00:19:06.657 --> 00:19:09.517 But also that argument is not so convincing, right? 00:19:09.597 --> 00:19:15.137 But for instance, is it possible that we have molecules binding to our receptors 00:19:15.137 --> 00:19:20.037 but not leading to sufficient activation to create percepts? 00:19:20.997 --> 00:19:27.337 That certainly seems to be true. I mean, part of the story in my talk was that 00:19:27.337 --> 00:19:33.197 there really can be two parameters for a molecule interacting with a receptor. 00:19:33.197 --> 00:19:38.717 Receptor, for a long time everyone's considered that the main parameter for 00:19:38.717 --> 00:19:45.537 a ligand is its affinity to a receptor, which actually tells you how strongly it wants to bind to it. 00:19:45.717 --> 00:19:51.477 And we know that ligands and receptors have a very wide distribution of these affinities. 00:19:52.417 --> 00:19:57.737 But it's also very clear that in more recent evidence is that there's a second parameter, 00:19:57.937 --> 00:20:01.257 which in pharmacology has been called efficacy and 00:20:01.257 --> 00:20:04.357 this tells you once a ligand has 00:20:04.357 --> 00:20:07.337 bound to a receptor how much kind of activity does it give downstream 00:20:07.337 --> 00:20:10.517 to an orn right and so that's related 00:20:10.517 --> 00:20:13.657 very much to your question that if you have for instance uh molecules 00:20:13.657 --> 00:20:18.537 that are binding there uh but they don't necessarily contribute much to uh to 00:20:18.537 --> 00:20:25.297 a um to a uh to a neural response then this is a classic so-called antagonist 00:20:25.297 --> 00:20:33.017 which is effectively filling up a space and sort of creating a blind spot, right? 00:20:33.497 --> 00:20:41.577 And up until now, until the last five years, we didn't really know about these antagonisms. 00:20:43.057 --> 00:20:49.697 And as you can imagine, they're kind of a bit scary experimentally. 00:20:50.499 --> 00:20:55.399 Because in order to be able to see them, you have to look at all possible mixtures, 00:20:55.439 --> 00:20:59.759 and you have to see how certain molecules may be blocking other molecules. 00:21:00.879 --> 00:21:04.519 And it's only really in the last five years that we've started to appreciate 00:21:04.519 --> 00:21:10.899 that, in fact, these types of antagonisms could indeed be really widespread in a faction. 00:21:11.079 --> 00:21:17.219 That basically our neural response is nowhere near the linear addition of just 00:21:17.219 --> 00:21:19.099 adding on lots of different odors. 00:21:19.099 --> 00:21:24.099 There's all sorts of these nonlinear competitions and interactions that are 00:21:24.099 --> 00:21:26.899 taking place, and this is probably what defines olfaction. 00:21:27.479 --> 00:21:32.199 So that might also mean that the response of glomeruli, 00:21:32.359 --> 00:21:38.759 mitral cells, and so on, is reflecting a much more complex flux of binding and 00:21:38.759 --> 00:21:44.099 unbinding of very broadly shaped ligands. 00:21:44.759 --> 00:21:50.239 Certainly. And how they interact. And we haven't even said anything about the 00:21:50.239 --> 00:21:53.879 so-called odorant binding properties which live in the mucosa. 00:21:54.299 --> 00:21:57.919 And these OBPs are doing two things. 00:21:58.039 --> 00:22:03.639 One thing they're doing is, in quite a clever way, shifting the sorption spectra 00:22:03.639 --> 00:22:08.779 for odors out there in the air phase so that we can detect odors that are hydrophobic. 00:22:08.779 --> 00:22:10.799 They don't want to be in the liquid phase. Yes. 00:22:11.240 --> 00:22:15.800 We need to be able to detect those as well. So that's a solution from nature 00:22:15.800 --> 00:22:18.740 to get those into the liquid phase so that we can interact with them. 00:22:19.040 --> 00:22:24.060 And another thing they do, they're very large proteins. They themselves have 00:22:24.060 --> 00:22:26.880 selective bindings to different chemicals. 00:22:26.880 --> 00:22:33.220 And so it's actually an extra layer of building specificity and tuning into 00:22:33.220 --> 00:22:34.900 the system to create more information. 00:22:35.100 --> 00:22:41.660 It's really quite an exquisitely layered sequence of events that deliver the 00:22:41.660 --> 00:22:46.380 certain subsets of molecules in certain places in the epithelium to be processed. 00:22:46.380 --> 00:22:53.820 But these odor-binding molecules, is there a belief that they in turn set up 00:22:53.820 --> 00:22:58.460 a local dynamic in the mucus layer, or are they just like transporters? 00:22:59.400 --> 00:23:02.980 I think you can think of them more like transporters because the mucosal layer 00:23:02.980 --> 00:23:04.680 is very thin. It's only a few microns. 00:23:04.840 --> 00:23:08.720 And their main job is to basically shift the sorption spectrum to get those 00:23:08.720 --> 00:23:14.120 hydrophobic compounds into the liquid phase and transport them to the binding 00:23:14.120 --> 00:23:18.380 site so that they can be… You see, they're probably not traveling lengthways. 00:23:18.540 --> 00:23:23.560 There probably isn't time really for them to, say, move them to different portions 00:23:23.560 --> 00:23:26.780 of the sheet, for instance, or something complicated like this, right? Right. 00:23:26.960 --> 00:23:30.120 So there are other mechanisms for that. For instance, 00:23:30.240 --> 00:23:35.920 the fact that you've got preferential sorption means that some odors are staying 00:23:35.920 --> 00:23:38.120 in longer in the air phase and some shorter, 00:23:38.220 --> 00:23:43.000 which means that you've already got a nice sort of almost like postal service 00:23:43.000 --> 00:23:49.800 selective delivery of molecules in certain places depending upon how much they want to be sorbed. 00:23:49.800 --> 00:23:55.340 Yeah, but I was more thinking you could also imagine that these binding molecules 00:23:55.340 --> 00:23:59.800 in turn, for instance, set up a competitive relationship. 00:23:59.820 --> 00:24:04.620 For instance, if you have a binding with a molecule of a certain kind that it 00:24:04.620 --> 00:24:09.240 leads to, let's say, signaling systems becoming active. 00:24:09.908 --> 00:24:13.248 That would change again probabilities to bind to other molecules or not. 00:24:13.508 --> 00:24:17.008 Yeah, it's quite possible that there's some sort of interaction between those 00:24:17.008 --> 00:24:19.208 at the receptors. I don't think it's really well understood. 00:24:19.368 --> 00:24:25.588 And for instance, even antagonism is not understood at all well because the 00:24:25.588 --> 00:24:30.388 simple-minded naive theory would be that an antagonist sort of fills the binding 00:24:30.388 --> 00:24:36.648 site on a receptor so that another ligand with higher efficacy cannot bind in that same place. 00:24:36.648 --> 00:24:44.088 But it seems quite likely that, in fact, there are more complicated mechanisms 00:24:44.088 --> 00:24:50.188 which are called dimers, which effectively means that they may not fill the main site. 00:24:50.288 --> 00:24:56.168 The main site may still be available for preferentially or high affinity binded 00:24:56.168 --> 00:25:01.448 compounds, but that these antagonists may actually bind onto the receptor in different places. 00:25:01.628 --> 00:25:05.328 And by doing that, actually change how the conformation of the protein. 00:25:05.328 --> 00:25:11.188 So in extremely complex ways that the protein can change its folding dynamics 00:25:11.188 --> 00:25:13.768 according to the binding to signal different stuff. 00:25:13.928 --> 00:25:18.528 So there's all sorts of complexities there. But in terms of your point about the dynamics, 00:25:18.668 --> 00:25:22.728 that's well taken because there's also this other group of compounds in the 00:25:22.728 --> 00:25:27.708 mucosa called odor degrading enzymes because a very good question is how do 00:25:27.708 --> 00:25:29.988 you get rid of these odors after binding? 00:25:29.988 --> 00:25:35.408 There needs to be a whole set of other molecules out there that are effectively 00:25:35.408 --> 00:25:39.888 basically removing these signals and terminating the signals. 00:25:39.968 --> 00:25:44.308 So it's a very exquisite balance of all of these molecular processes going on 00:25:44.308 --> 00:25:50.168 to both initiate the signal and then terminate it as well in a timely fashion. 00:25:50.168 --> 00:25:53.428 It's actually, it's already amazing to see, right, that here we're looking at, 00:25:53.508 --> 00:26:00.568 let's say, a very advanced variation on, in some sense, the most primitive form of sensation we know. 00:26:00.668 --> 00:26:05.628 Because for single cellular organisms, you have, you know, mechanosensing or 00:26:05.628 --> 00:26:08.888 chemical sensing, and that's what it started with, right? 00:26:08.988 --> 00:26:13.488 Yeah. So, and if you just look at it already at the level of this mucus layer, 00:26:13.628 --> 00:26:19.608 the very first point where olfactory release starts, the exquisite orchestration 00:26:19.608 --> 00:26:23.448 of all these different pathways and interactions is very impressive. 00:26:23.948 --> 00:26:28.668 But then now, of course, we want to go step up and start to think about, 00:26:28.768 --> 00:26:33.368 okay, how does this in the end lead to the encoding and representations and 00:26:33.368 --> 00:26:35.968 detection of odors in the environment? 00:26:36.468 --> 00:26:41.948 And so the first thing, of course, we want to understand is what's really this molecular language? 00:26:42.228 --> 00:26:47.428 How would an olfactory system in the end really, if you want, decompose molecules? 00:26:47.848 --> 00:26:52.348 What are the key features of molecules our olfactory system is sensitive to? 00:26:52.948 --> 00:26:58.108 Yeah, so there's some very nice data sets. I think finally olfaction is coming 00:26:58.108 --> 00:27:03.948 into the 21st century of sort of data sharing and there's some very nice data 00:27:03.948 --> 00:27:09.228 sets with many different odor conditions and optical imaging on the receptor surface. 00:27:09.228 --> 00:27:14.808 Using these methods, we can directly see actually what the, for instance, 00:27:14.968 --> 00:27:19.668 calcium levels of activity are, different portions of the receptor sheet. 00:27:19.908 --> 00:27:26.788 And large data sets which have shown how different portions of the sheet are 00:27:26.788 --> 00:27:29.248 responding in different ways to different groups of chemicals. 00:27:29.248 --> 00:27:34.188 And it's very clear when you look at these sorts of databases that different 00:27:34.188 --> 00:27:40.388 portions of the receptor sheet are responding to different groups of compounds. 00:27:40.608 --> 00:27:49.448 So there are sort of, you can see common properties of molecules that may be 00:27:49.448 --> 00:27:54.908 preferentially tuning or activating different parts of the receptor sheet. 00:27:54.908 --> 00:27:57.968 And so this is 00:27:57.968 --> 00:28:01.048 leading to the idea that there is a kind of chemotopic mapping 00:28:01.048 --> 00:28:05.148 of the receptor sheet rather than the receptor sheet 00:28:05.148 --> 00:28:08.148 being mapped onto physical space as 00:28:08.148 --> 00:28:11.708 it is in say vision on the retina the 00:28:11.708 --> 00:28:16.088 mapping seems to be a chemotopic mapping so we talk about different molecular 00:28:16.088 --> 00:28:19.948 features being mapped to different areas and this may be for a couple of reasons 00:28:19.948 --> 00:28:24.328 a number of reasons actually one reason is this sorption thing that I talked 00:28:24.328 --> 00:28:28.968 about that compounds with higher sorption parameters, 00:28:29.188 --> 00:28:34.148 so they like to be in the liquid phase, they're more hydrophilic. 00:28:34.448 --> 00:28:41.448 Those will basically absorb into the earlier parts of the receptor sheet as you sniff. 00:28:42.262 --> 00:28:48.662 So they'll be sort of more delivered to the front end of the receptors where they live in the mucosa. 00:28:48.742 --> 00:28:54.322 And other compounds that are more hydrophobic would be delivered later, 00:28:54.482 --> 00:28:58.142 further back in the receptor sheet as you sniff. 00:28:58.542 --> 00:29:03.522 And there are other reasons is that this family of receptors that I talked about, 00:29:03.622 --> 00:29:07.782 in fact, it turns out that there's at least four classes of these, 00:29:07.782 --> 00:29:13.122 although this is currently under debate, but there's two very, very clear classes, 00:29:13.522 --> 00:29:19.222 class 1 and class 2, which correspond to fishoid receptors, which in fact are 00:29:19.222 --> 00:29:23.102 leftover that we have from when we were fish. 00:29:23.302 --> 00:29:28.042 They're dedicated to detecting non-volatile compounds largely, 00:29:28.282 --> 00:29:32.802 and they are exclusively expressed in certain zones of the receptor sheet. 00:29:32.802 --> 00:29:37.682 And so, as you can imagine, this leads to another form of chemotopic mapping 00:29:37.682 --> 00:29:40.742 where you get a preferential response. 00:29:40.742 --> 00:29:45.962 And these FISH-derived receptors wouldn't go for the higher Dalton molecules 00:29:45.962 --> 00:29:48.922 because these are the ones that are less volatile? 00:29:49.442 --> 00:29:55.382 No, not necessarily because whether a compound is hydrophilic or hydrophobic 00:29:55.382 --> 00:30:03.722 depends more on actually the charge on the molecule. So it's more to do with the sort of asymmetry. 00:30:03.802 --> 00:30:07.442 No, but you said they went for the non-volatiles or the less volatile. 00:30:07.822 --> 00:30:14.262 Yeah, yeah, yeah. So I was wondering then whether these fish-derived receptors 00:30:14.262 --> 00:30:20.202 are also sensitive to molecules with higher dose. Well, there's two things that 00:30:20.202 --> 00:30:21.962 determine sort of volatility. 00:30:22.242 --> 00:30:25.462 One is the molecular mass, and the other thing is whether they like the air 00:30:25.462 --> 00:30:29.462 phase or the liquid phase, which is the sorption profile. So it's really those 00:30:29.462 --> 00:30:30.982 two things together that determine. 00:30:31.942 --> 00:30:38.102 So in a fishoid receptor, they would be exclusively compounds that they have 00:30:38.102 --> 00:30:42.442 sorption parameters that they can never make it into the air phase really at 00:30:42.442 --> 00:30:47.022 room temperatures unless you started boiling them or had a lot higher energy. 00:30:47.022 --> 00:30:51.722 Okay, but so now we have a zoning, if you want, of the epithelium. 00:30:52.802 --> 00:30:56.802 And the zoning might correspond to different properties of molecules like their 00:30:56.802 --> 00:31:01.782 cyclization, their carbon number, bond saturation, branching, 00:31:01.802 --> 00:31:04.442 substitution pattern, functional groups. 00:31:05.802 --> 00:31:09.982 So are these the key chemical features of these molecules that you would map onto these zones? 00:31:10.322 --> 00:31:17.002 These are at least a number of key features. Those ones you've just mentioned are very important. 00:31:18.262 --> 00:31:21.462 This has been another area of a crucial study. 00:31:21.562 --> 00:31:25.942 There's databases. databases uh there's some very nice software that you can 00:31:25.942 --> 00:31:28.542 download for public use called dragon, 00:31:29.302 --> 00:31:36.002 and for pretty much any molecule you might care about you can go to this package 00:31:36.002 --> 00:31:37.562 and it will give you something like um. 00:31:39.148 --> 00:31:44.468 I think it's about 400 or 500 different exotic descriptors for a molecule, 00:31:44.548 --> 00:31:47.788 branching, all sorts of things that you can't imagine. 00:31:47.928 --> 00:31:50.808 So a huge number of possible descriptors, because, of course, 00:31:50.848 --> 00:31:58.768 you can describe a molecular structure in an almost infinite set of ways that 00:31:58.768 --> 00:32:01.048 you might choose to describe that, right? 00:32:01.148 --> 00:32:04.248 So these databases and chemists 00:32:04.248 --> 00:32:07.548 have been very carefully characterizing different descriptor 00:32:07.548 --> 00:32:11.128 or what you might call molecular determinants and then 00:32:11.128 --> 00:32:14.388 of course you can play very interesting games like um 00:32:14.388 --> 00:32:17.388 look at all those odorous compounds and look 00:32:17.388 --> 00:32:20.248 at all the descriptors are there are there certain descriptors that 00:32:20.248 --> 00:32:23.228 may be more important for certain areas of the receptor sheet 00:32:23.228 --> 00:32:25.948 and others that aren't and so on you can 00:32:25.948 --> 00:32:28.808 play games like this and you find that um indeed there are 00:32:28.808 --> 00:32:31.788 um but you also find that um all of 00:32:31.788 --> 00:32:35.968 these descriptors are highly redundant and what i mean by that is that um you 00:32:35.968 --> 00:32:40.428 know you don't find for instance that there's one and two one or two descriptors 00:32:40.428 --> 00:32:46.488 that sort of uh that sort of tell you very orthogonal or decorrelated information 00:32:46.488 --> 00:32:49.688 they all tend to be measuring a similar thing. 00:32:51.168 --> 00:32:54.168 And uh and so if you take for instance a pca 00:32:54.168 --> 00:32:57.308 or principal component analysis of the contribution of all 00:32:57.308 --> 00:33:00.868 of these discriminants to what you might think of as a perceptual description 00:33:00.868 --> 00:33:03.808 of an odor you find that there are large numbers 00:33:03.808 --> 00:33:07.068 you know because they're highly redundant so right exactly you 00:33:07.068 --> 00:33:09.888 can no but so this is very interesting right because 00:33:09.888 --> 00:33:12.848 so here we go if you want i 00:33:12.848 --> 00:33:17.108 mean this is a little bit proto-science right where we just develop descriptors 00:33:17.108 --> 00:33:21.428 of reality so the same holds for chemistry so here we are for this huge collection 00:33:21.428 --> 00:33:27.288 of descriptors um but now on the other hand you'll see also you see two things 00:33:27.288 --> 00:33:32.588 right so on the one there's a zoning a zoning in the olfactory system that would suggest, 00:33:33.168 --> 00:33:40.608 that the properties of these molecules are in some sense also mapped onto this chemotopic map. 00:33:40.908 --> 00:33:45.368 And that, of course, would also give you, let's say, this should give you some 00:33:45.368 --> 00:33:48.988 sort of hierarchy of which descriptors actually are helpful or are, 00:33:49.048 --> 00:33:53.908 from a biological perspective, relevant and which are indeed, 00:33:53.968 --> 00:33:55.348 in that sense, redundant. That's one thing. 00:33:56.288 --> 00:33:58.308 But then you showed, which is also very interesting, Interesting. 00:33:59.223 --> 00:34:03.783 And if you analyze in more detail these descriptors and the odor space that 00:34:03.783 --> 00:34:09.023 they define, that actually it's much lower dimensional than these descriptors would suggest. 00:34:10.023 --> 00:34:13.863 Yeah, well, that's another very interesting story. Although that one that I 00:34:13.863 --> 00:34:19.543 showed in the talk was actually about analyzing perceptual descriptors for a corresponding odor. 00:34:19.543 --> 00:34:25.103 So you can do another thing, which is quite a lot of fun, which is to take a 00:34:25.103 --> 00:34:29.883 large number of molecules, such as in the Dravnik's database, 00:34:30.063 --> 00:34:33.303 where about 150 different molecules were taken. 00:34:33.303 --> 00:34:40.143 And they then used a large pool of trained olfactory specialists, 00:34:40.523 --> 00:34:51.063 and they asked those specialists to describe each of these 150 odors using a 00:34:51.063 --> 00:34:55.343 panel of approximately 150 different descriptors, 00:34:55.343 --> 00:34:58.183 such as musky, 00:34:58.343 --> 00:35:00.743 grassy, nutty, so on. 00:35:01.303 --> 00:35:06.203 So large numbers of these and then you can also look at the space for similarity 00:35:06.203 --> 00:35:08.843 of those for different molecules so you can plot individual, 00:35:09.423 --> 00:35:14.103 molecules in a high dimensional space and you can look how similar or dissimilar 00:35:14.103 --> 00:35:17.283 are these discriminators so you can make a sort of odour map. 00:35:17.984 --> 00:35:20.804 Of perceptual odor map in terms of 00:35:20.804 --> 00:35:24.244 what you would expect is that you'd expect those points 00:35:24.244 --> 00:35:27.604 uh close in this map to have similar perceptual 00:35:27.604 --> 00:35:31.064 properties and therefore you'd expect them to have similar molecular properties 00:35:31.064 --> 00:35:34.004 right so it enables you to play this nice game where 00:35:34.004 --> 00:35:37.504 you can look at different portions of this map and see are 00:35:37.504 --> 00:35:40.544 is there a nice continuum here in certain molecular 00:35:40.544 --> 00:35:43.424 properties that are moving this perception in 00:35:43.424 --> 00:35:46.564 this direction or whatever But what 00:35:46.564 --> 00:35:49.724 you find which is very interesting when you do this Which was studied by Kolokoff 00:35:49.724 --> 00:35:53.204 paper about three years ago Was 00:35:53.204 --> 00:35:57.104 that when you analyse all of these 150 descriptors You 00:35:57.104 --> 00:36:00.524 find that they're placed in a 00:36:00.524 --> 00:36:04.584 much lower dimensional space 00:36:04.584 --> 00:36:08.164 So it means that effectively they're highly 00:36:08.164 --> 00:36:11.324 correlated right so and and um 00:36:11.324 --> 00:36:15.044 they should they found that there was a uh at 00:36:15.044 --> 00:36:18.084 least in a two-dimensional manifold was able 00:36:18.084 --> 00:36:21.104 to explain something like um 60 or 00:36:21.104 --> 00:36:26.444 70 percent of the variance of this of this 150 dimensional structure can actually 00:36:26.444 --> 00:36:32.084 be explained quite adequately reasonably well in this two-dimensional manifold 00:36:32.084 --> 00:36:38.244 which is spread it looks a bit like a potato chip sort of shape within this. 00:36:38.884 --> 00:36:39.744 Higher dimensional space. 00:36:40.184 --> 00:36:46.364 And so I think it's very exciting actually. I think we will find in the future, 00:36:46.964 --> 00:36:49.724 olfactory studies, we're going to find that there are all sorts of these. 00:36:51.504 --> 00:36:57.704 Lower dimensional manifolds, which is kind of the answer to solving the olfactory 00:36:57.704 --> 00:37:02.304 code to find out there has to be some underlying similarities and structures in these things. 00:37:02.484 --> 00:37:05.904 This is of course where we want to get to, right? One thing that's, 00:37:05.924 --> 00:37:09.504 of course, interesting here is that now, with this last analysis, 00:37:09.724 --> 00:37:15.304 you could argue, well, maybe human experience of olfaction is relatively low-dimensional. 00:37:15.464 --> 00:37:18.484 But what's complex are all these descriptions we glue onto. 00:37:19.424 --> 00:37:22.764 It's also the language, right? Because this is limited by language. 00:37:22.924 --> 00:37:25.864 The language just complexifies the experience. 00:37:26.244 --> 00:37:29.744 But what's amazing about olfaction is that we have no other way to describe 00:37:29.744 --> 00:37:34.104 certain olfactory cues other than it's like other things. 00:37:34.284 --> 00:37:38.584 So this is kind of the interesting restrictions in the language because we say 00:37:38.584 --> 00:37:42.064 it's like grass or it's like this or like that. 00:37:42.464 --> 00:37:48.124 It doesn't seem that we have any kind of independent, really independent descriptions 00:37:48.124 --> 00:37:52.104 of owners other than just likening them to other things. Right. 00:37:52.895 --> 00:37:57.235 So there are no intrinsic qualities that we can use. Well, it turns out when 00:37:57.235 --> 00:38:01.595 you look at this map, because you can look at this manifold and then you can 00:38:01.595 --> 00:38:07.035 look at what are the actual qualities and how does this manifold actually represent the odors. 00:38:07.155 --> 00:38:11.355 When you do that, you find some interesting things. So you see that the main 00:38:11.355 --> 00:38:16.335 dimension of this map very strongly correlates to pleasantness and unpleasantness. 00:38:16.335 --> 00:38:22.815 So maybe this is the only true underlying parameter, whether we actually have 00:38:22.815 --> 00:38:27.755 a hedonic or whatever the valence is for the actual. Exactly right, yes. 00:38:28.195 --> 00:38:31.495 Which would be interesting because then you would have, let's say, valence and intensity. 00:38:32.295 --> 00:38:38.015 And this would be very compatible with how we can describe the emotional states. 00:38:38.375 --> 00:38:41.735 We would think of valence and arousal, something like this. 00:38:41.735 --> 00:38:44.395 And then within such a two-dimensional space you can 00:38:44.395 --> 00:38:49.455 find all sort of very complex emotional descriptors yeah but but so but the 00:38:49.455 --> 00:38:54.335 point i want to get to is then um how what does this now mean for this idea 00:38:54.335 --> 00:39:00.315 of zoning of of of the receptor neurons in the epithelium would that match would 00:39:00.315 --> 00:39:04.135 it match in any way the structure of this descriptor space or not, 00:39:05.335 --> 00:39:08.375 um yeah that's a really good question um. 00:39:10.000 --> 00:39:18.700 Yes, I think it probably will. I mean, there are another approach to describing 00:39:18.700 --> 00:39:20.520 the chemotopic mapping, 00:39:20.600 --> 00:39:24.980 and this is work by Kaniezo Mori, who's been very influential in the field, 00:39:25.020 --> 00:39:30.100 has made a very interesting argument that this space, 00:39:30.320 --> 00:39:32.600 although it's a chemotopic space, 00:39:33.040 --> 00:39:36.420 may also be structured behaviorally. 00:39:36.420 --> 00:39:42.080 So certain behaviorally relevant compounds may be represented in certain parts 00:39:42.080 --> 00:39:47.620 of the bulb and other differently behaviorally relevant compounds in other parts. 00:39:47.740 --> 00:39:52.820 And you would expect those compounds to have similar perceptual kind of descriptions. 00:39:52.820 --> 00:39:56.200 Right so it does seem that the 00:39:56.200 --> 00:39:59.740 the current thinking of uh mapping of 00:39:59.740 --> 00:40:04.900 the first stage of odor representation in the olfactory system comes down to 00:40:04.900 --> 00:40:10.460 both chemotopic in terms of molecular properties uh and probably uh behaviorally 00:40:10.460 --> 00:40:15.280 behavioral relevance in different ways uh which will then of course have some 00:40:15.280 --> 00:40:17.920 type of relationship to perceptual quality right exactly, 00:40:18.600 --> 00:40:21.500 and this is cool right so now we've had a bit an idea of the structuring of 00:40:21.500 --> 00:40:25.000 this of this sort of The receptor layer, there's a structure. 00:40:25.820 --> 00:40:30.100 And now if you measure, and these are classic experiments that you also showed, 00:40:30.740 --> 00:40:35.940 if you measure the response of single receptor neurons to different kinds of 00:40:35.940 --> 00:40:40.700 odorants, different kinds of molecules, you see that actually they're not so specifically tuned. 00:40:41.680 --> 00:40:46.520 So what's the deal there exactly? Yeah, no, this is a very good point. 00:40:46.600 --> 00:40:51.320 What you find when you look in detail at each one of these receptor types is 00:40:51.320 --> 00:40:56.640 that they tend to have what we call broad tuning, as you say, 00:40:56.780 --> 00:40:59.100 to large numbers of compounds. 00:40:59.480 --> 00:41:10.280 But yet there are also a sort of smaller subset, I guess, which are more narrowly tuned. 00:41:10.520 --> 00:41:15.380 So it seems that the current story is that in fact the olfactory system has 00:41:15.380 --> 00:41:18.860 a variety of broadly tuned and more specific tuning. 00:41:18.860 --> 00:41:22.100 And one of the things I showed in the talk, which was some work that I did with 00:41:22.100 --> 00:41:25.660 Manuel Sanchez quite a while ago, actually 10 years, 00:41:25.820 --> 00:41:34.880 where we applied an information theory approach to basically describe the accuracy 00:41:34.880 --> 00:41:40.360 that a neuronal code at the epithelium would be able to reconstruct the original stimulus. 00:41:40.360 --> 00:41:45.000 And when you do this for very high dimensional stimulus like we have here with 00:41:45.000 --> 00:41:46.080 large numbers of compounds, 00:41:46.360 --> 00:41:51.080 you find that in fact this is exactly the perfect strategy that you need because 00:41:51.080 --> 00:41:58.680 you want to have certain receptors that are very broadly discriminating between 00:41:58.680 --> 00:42:01.300 groups of compounds. So they're very broadly tuned. 00:42:01.900 --> 00:42:06.100 And then you want other receptors which are doing more fine things. 00:42:06.651 --> 00:42:09.891 Uh discriminations between you know more specific sets 00:42:09.891 --> 00:42:12.631 of compounds and that if you have this 00:42:12.631 --> 00:42:16.051 sort of two layers and some continuum between them uh 00:42:16.051 --> 00:42:22.451 in information theory terms this gives you the best possible ability to reconstruct 00:42:22.451 --> 00:42:28.951 the stimulus so you you get in a high dimensional space now if your problem 00:42:28.951 --> 00:42:33.631 the reason why this is interesting is that uh if you restrict the problem to 00:42:33.631 --> 00:42:34.871 a lower a dimensional space, 00:42:35.011 --> 00:42:40.311 such as I know Bill Hanson, Peter Mombats were here last week. 00:42:40.851 --> 00:42:45.311 And for instance, Bill will be working, or the animals that Bill studies will 00:42:45.311 --> 00:42:51.851 be working in much more restricted dimensional space because ecologically, maybe there's fewer. 00:42:53.411 --> 00:42:58.511 Compounds which are more relevant to, you know, ethologically and ecologically for the animal. 00:42:58.511 --> 00:43:01.491 Then in those types of olfactory systems 00:43:01.491 --> 00:43:04.511 you tend to find that there are more specific receptors 00:43:04.511 --> 00:43:07.551 because you don't get this advantage of 00:43:07.551 --> 00:43:11.031 having lots of broad tune because you don't have the similar very 00:43:11.031 --> 00:43:13.691 high dimensionality so your your prediction is that 00:43:13.691 --> 00:43:16.751 the ratio of broad or specifically tuned 00:43:16.751 --> 00:43:19.711 receptor neurons would scale with the outer space 00:43:19.711 --> 00:43:23.011 you have to classify absolutely and the complexity of 00:43:23.011 --> 00:43:27.711 the world within you within which you have to operate right because that this 00:43:27.711 --> 00:43:30.151 Just coming back to the point I made at the beginning that if you need this 00:43:30.151 --> 00:43:37.371 very xenobiotic property where you have to be able to basically associate any 00:43:37.371 --> 00:43:42.111 potential molecule out there in the world with a particular behavior or whatever, 00:43:42.291 --> 00:43:48.551 then you very much need this ability to deal with very high dimensional stimuli. 00:43:48.851 --> 00:43:52.571 Because if you're an insect, there's probably no real requirement for these 00:43:52.571 --> 00:43:56.331 sorts of things in certain settings. Right. 00:43:56.811 --> 00:44:01.711 But now, let's go back to this theoretical study you just described. 00:44:01.991 --> 00:44:07.511 And sometimes you have a fairly linear decomposition of the olfactory system. 00:44:08.137 --> 00:44:10.657 A perception problem, right? Because you basically, you assumed, 00:44:10.817 --> 00:44:14.197 look, I have a bunch of, I have a stimulus, stimulus comes in, 00:44:14.237 --> 00:44:19.537 I have my receptor neurons, they feed directly into some sort of response. 00:44:21.157 --> 00:44:23.557 Presumably in the glomeruli or the mitral cells. 00:44:24.277 --> 00:44:29.977 And then I'm going to use some estimator. You don't need to make any commitments 00:44:29.977 --> 00:44:32.937 where this estimator resides in the brain because now we're doing it for an 00:44:32.937 --> 00:44:33.957 information theory game. 00:44:34.317 --> 00:44:37.077 And now I get an estimated stimulus. 00:44:38.137 --> 00:44:41.757 And then the game is that you really want to understand the parameters under 00:44:41.757 --> 00:44:46.957 which this estimated stimulus is as equal or similar as possible to the real 00:44:46.957 --> 00:44:49.937 stimulus, so you have a minimum. The error is minimal. Exactly. All right? 00:44:50.757 --> 00:44:55.297 So then, of course, the usual game is, well, the response will have some noise, 00:44:55.497 --> 00:44:57.217 and this noise will have certain properties. 00:44:59.497 --> 00:45:04.437 So because of this noise, I cannot really inversely map anymore my response 00:45:04.437 --> 00:45:05.817 to the stimulus. Exactly. 00:45:06.177 --> 00:45:09.917 Right? But okay, when the noise is big. It corrupts the original. Exactly. 00:45:10.677 --> 00:45:16.277 So now you use this Fisher information approach to decompose this problem and 00:45:16.277 --> 00:45:17.797 understand what's the optimal strategy. 00:45:18.217 --> 00:45:22.597 Exactly. So how does this Fisher information approach help you with this? 00:45:23.117 --> 00:45:30.037 Well, Fisher information effectively characterizes and quantifies very precisely 00:45:30.037 --> 00:45:34.657 two things, which is both your sensitivity to a stimulus. 00:45:34.657 --> 00:45:38.577 So the sensitivity that a receptor 00:45:38.577 --> 00:45:47.117 has to a stimulus and also how that sensitivity is corrupted by noise. 00:45:47.857 --> 00:45:54.697 And then it also gives you another very nice thing, which is how does that add across a population? 00:45:54.937 --> 00:46:00.297 Because you might know that for one receptor, but how do you know how all that 00:46:00.297 --> 00:46:04.197 sort of information adds when you have a population? And so Fischer information 00:46:04.197 --> 00:46:06.497 precisely quantifies that effectively. 00:46:06.677 --> 00:46:13.217 So for an array of receptors, it quantifies very precisely for each receptor 00:46:13.217 --> 00:46:18.977 what the contribution is to the sensitivity for each one of the dimensions and 00:46:18.977 --> 00:46:21.617 also how that is corrupted by its own noise. 00:46:21.617 --> 00:46:27.897 And effectively, when you characterize that, there are some very nice results, 00:46:28.197 --> 00:46:34.917 particularly by Kramer-Rayo, which tells you that the inverse of this matrix, 00:46:35.137 --> 00:46:39.037 which contains all of this information about the noise and the sensitivities, 00:46:39.217 --> 00:46:48.257 basically the inverse of this directly tells you what the best estimator can 00:46:48.257 --> 00:46:51.197 do in terms of being able to, 00:46:52.017 --> 00:46:53.637 reconstruct the original stimulus. 00:46:54.517 --> 00:46:58.977 And so therefore it tells you that it doesn't matter what the biological system 00:46:58.977 --> 00:47:05.677 is, it will not be able to exceed this limit and so it's nice very nice I think in the sense that it, 00:47:06.220 --> 00:47:09.460 provides a very defined limit on 00:47:09.460 --> 00:47:12.560 a psychophysical test that you might do to tell 00:47:12.560 --> 00:47:17.300 you for instance if you had a stimulus and you changed it a small amount so 00:47:17.300 --> 00:47:22.120 the just noticeable difference the just noticeable difference or the jnd in 00:47:22.120 --> 00:47:27.720 the psychophysics experiment would relate directly to this uh inverse or this 00:47:27.720 --> 00:47:31.600 kramer around which comes so it makes a very nice firm prediction 00:47:31.920 --> 00:47:37.280 for the architecture of a neural system and uh and and what the limits would 00:47:37.280 --> 00:47:41.720 be for an animal right have you been able to to validate any of this any of these predictions. 00:47:42.820 --> 00:47:46.580 Um well in terms of relating it to psychophysics 00:47:46.580 --> 00:47:49.640 um that's something 00:47:49.640 --> 00:47:52.480 that's something we should we should push forward but 00:47:52.480 --> 00:47:56.960 in this experiment we're only really doing it as a game rather than because 00:47:56.960 --> 00:48:01.520 it doesn't really matter what the precise value is because all we were really 00:48:01.520 --> 00:48:07.420 interested to know is what are the actual features of the receptors which either 00:48:07.420 --> 00:48:09.640 give you a good or bad ability to do this. 00:48:09.860 --> 00:48:15.200 Right, but now you do assume that in this case you only have a linear interaction 00:48:15.200 --> 00:48:18.840 between your receptor and your actual linear combination, right? 00:48:18.900 --> 00:48:20.440 And is that not a very strong assumption? 00:48:21.060 --> 00:48:26.240 It is quite a strong assumption. I mean, what you're saying is that a receptor 00:48:26.240 --> 00:48:31.220 is giving a linear response to a set of combinations. 00:48:31.380 --> 00:48:36.240 So, for instance, things like antagonism are kind of out. As we just discussed, right? 00:48:36.440 --> 00:48:45.180 Yeah. So, you have to go to a sort of second order. Yeah. 00:48:45.451 --> 00:48:48.311 You could extend it to a second order, which we haven't done. 00:48:49.531 --> 00:48:52.551 But really it was a quick study to just show what the features are. 00:48:52.711 --> 00:48:58.531 And I think the main sort of interest in it is that the simple properties of 00:48:58.531 --> 00:49:02.711 the simple-minded approach is that you get this sort of mixture of diversity 00:49:02.711 --> 00:49:09.051 in the outputs together with specific receptors that very clearly matches what 00:49:09.051 --> 00:49:12.371 you see in all experiments from insects and mammals. 00:49:12.371 --> 00:49:16.891 In some sense, you try to make sense of this idea of population coding with 00:49:16.891 --> 00:49:21.191 broadly tuned receptor neurons that have, again, variability in this tuning. 00:49:21.291 --> 00:49:22.791 This is the key point of this study. 00:49:23.051 --> 00:49:29.091 Exactly. It's just to show that what the biology is doing is a sensible thing 00:49:29.091 --> 00:49:32.751 in terms of when you're facing a very high-dimensional input. 00:49:33.031 --> 00:49:38.411 Right, which is interesting, right? Because very often you hear people making 00:49:38.411 --> 00:49:43.491 these wild claims about how suboptimal, inefficient natural systems are. 00:49:43.631 --> 00:49:47.831 But I think, certainly with this example, you'll be hard-pressed to do it better. 00:49:48.411 --> 00:49:52.991 Would you agree with that? Yeah, within your space, within your defined space 00:49:52.991 --> 00:49:54.371 of the problem you're dealing with. 00:49:54.651 --> 00:49:58.471 Right, exactly. What would be interesting to compare, which we haven't done, 00:49:58.651 --> 00:50:01.811 is to then compare these measures maybe across animals, right? 00:50:01.931 --> 00:50:05.551 Sure. To show that different levels of specificity, 00:50:05.551 --> 00:50:09.011 like we were discussing, processing may be more or less relevant to 00:50:09.011 --> 00:50:12.091 the dimensionality of what you're dealing with in your in your environment 00:50:12.091 --> 00:50:14.831 but now also the other thing though with your with your 00:50:14.831 --> 00:50:18.231 model are there two things on the one hand you do assume that 00:50:18.231 --> 00:50:21.631 after let's say one stage of of processing 00:50:21.631 --> 00:50:27.451 one of transformation you can reconstruct your stimulus right that's your quality 00:50:27.451 --> 00:50:31.491 measure well we don't assume that it actually does that all we're saying is 00:50:31.491 --> 00:50:35.891 that this places a bound on what any subsequent neural processing would not 00:50:35.891 --> 00:50:40.171 be able to go beyond this effectively any estimators. 00:50:40.870 --> 00:50:46.010 Provided that any subsequence step is fully isolated from any preceding step. 00:50:46.130 --> 00:50:47.690 It has to be a cleanly modular system. 00:50:49.070 --> 00:50:53.370 No, I mean it just provides a limit on that. But what it's really saying is 00:50:53.370 --> 00:50:55.890 it provides a limit given the noise that was introduced. 00:50:56.150 --> 00:51:00.250 So it's based on the assumption of what the noise was at the periphery. 00:51:00.270 --> 00:51:05.390 But for instance, if I would have a subset of projections that would also percolate 00:51:05.390 --> 00:51:12.550 up this processing hierarchy, bypassing noise stages or following a different kind of noise dynamics. 00:51:13.350 --> 00:51:17.610 Oh, okay. Well, the noise we're thinking about is only at this stage the noise 00:51:17.610 --> 00:51:20.990 at the receptors, because all we're characterizing is the receptor population. 00:51:21.470 --> 00:51:25.530 We're not saying anything about subsequent processing. So all we're really saying 00:51:25.530 --> 00:51:30.830 is the question we wanted to answer was what should the tuning of receptors be, 00:51:30.950 --> 00:51:33.810 not necessarily the subsequent neural processing, which is 00:51:33.810 --> 00:51:36.510 a a different story the reason why you can't really do that with 00:51:36.510 --> 00:51:39.270 this measure is that the subsequent neural processing has all 00:51:39.270 --> 00:51:42.770 sorts of complex time dynamics as you say quite rightly 00:51:42.770 --> 00:51:45.650 there's all sorts of maybe noise properties there in complex 00:51:45.650 --> 00:51:48.830 ways there's all sorts of attentional processing and 00:51:48.830 --> 00:51:51.990 so on so uh at least in 00:51:51.990 --> 00:51:54.930 this formulation of uh let's say 00:51:54.930 --> 00:51:57.670 thinking about the fisher information it would not make a 00:51:57.670 --> 00:52:00.530 whole lot of sense to okay try to extend it beyond 00:52:00.530 --> 00:52:03.350 uh subsequent neural processing thing we're just 00:52:03.350 --> 00:52:06.230 really trying to limit it to what the receptors themselves 00:52:06.230 --> 00:52:08.990 are doing right but now what so what's the signal to 00:52:08.990 --> 00:52:12.170 noise level you find in these receptors signal to 00:52:12.170 --> 00:52:18.730 noise um uh well obviously it's proson we find proson firing uh quite close 00:52:18.730 --> 00:52:25.410 uh to sort of cortex uh sort of proson i mean as you know in cortex i think 00:52:25.410 --> 00:52:29.750 the fano factor is slightly greater than in Poisson, so they show more variability. 00:52:30.010 --> 00:52:35.190 The story with the variability of firing, you sort of see that also in receptors. 00:52:35.410 --> 00:52:39.150 So you see large amount of variability in the firing, beyond, 00:52:39.390 --> 00:52:41.090 slightly beyond Poisson firing. 00:52:41.310 --> 00:52:47.430 Okay, but the other thing. That's the noise, as far as the receptor population's 00:52:47.430 --> 00:52:51.010 concerned, it's both the noise and the firing. Right, exactly. 00:52:51.590 --> 00:52:54.330 But now the other thing is that in this case, 00:52:55.327 --> 00:52:59.207 This is under, I assume, a constant drive with a single molecule. 00:53:00.267 --> 00:53:04.707 But the real system will be operating in the face of a continuous fluctuating 00:53:04.707 --> 00:53:11.407 stream of molecules bombarding these different receptors with varying binding kinetics and so on. 00:53:11.467 --> 00:53:20.727 So would this statement on this upper bound of processing capability also hold 00:53:20.727 --> 00:53:23.787 if we start to generalize to this more realistic input condition? 00:53:24.407 --> 00:53:27.287 Sure i mean it may not uh there's all 00:53:27.287 --> 00:53:30.067 sorts of complex dynamical processes going on there and 00:53:30.067 --> 00:53:33.227 the fact that it's a first order description is is 00:53:33.227 --> 00:53:36.047 just the simplest minded thing you can do in order to right in 00:53:36.047 --> 00:53:39.167 order to uh just look at the properties of 00:53:39.167 --> 00:53:44.087 the system so i don't think we make any claim that um that 00:53:44.087 --> 00:53:47.367 maybe it has a great expectation 00:53:47.367 --> 00:53:50.107 to match very precisely with the 00:53:50.107 --> 00:53:53.507 psychophysics right because also you don't 00:53:53.507 --> 00:53:56.567 know how much uh the information either gets 00:53:56.567 --> 00:53:59.487 selected or not selected downstream right because 00:53:59.487 --> 00:54:03.607 you have all sorts of selection processes as well right so when you come to 00:54:03.607 --> 00:54:07.867 perceptual processes um you've got all sorts of selection of the information 00:54:07.867 --> 00:54:14.407 going on so um yeah and you don't have complete information at the periphery 00:54:14.407 --> 00:54:19.427 to to be able to say the exact state of the system to ever really be able to test very precisely 00:54:19.747 --> 00:54:25.807 how exactly accurately this would be characterized. 00:54:25.947 --> 00:54:29.747 Right. But then the other issue that actually we haven't touched upon at all 00:54:29.747 --> 00:54:32.987 is that, okay, we have sort of talked about these molecules. 00:54:34.487 --> 00:54:37.967 The ability to detect them, the reliability of the system. 00:54:40.187 --> 00:54:46.007 But actually a key issue for olfaction is concentration, right? 00:54:46.007 --> 00:54:53.047 So how does an olfactory system really deal with varying levels of concentration? 00:54:53.467 --> 00:54:58.607 I mean, it can go from really minute, that's a homeopathic quantity, to full saturation. 00:54:59.047 --> 00:55:04.627 So it has to operate in a huge dynamic range. So how does it manage to do that? 00:55:04.707 --> 00:55:06.387 And what kind of sensitivity do we get? 00:55:07.667 --> 00:55:11.467 Yeah, I think there's a number of tricks for this. 00:55:11.467 --> 00:55:17.627 This uh the the system as you say is in a it's about the most remarkable i think in the, 00:55:18.247 --> 00:55:22.867 in the nervous system and having to deal with such incredible uh dynamic range 00:55:22.867 --> 00:55:29.327 you've got about probably at least 10 orders of magnitude in concentration which 00:55:29.327 --> 00:55:31.787 is phenomenal phenomenal uh. 00:55:32.885 --> 00:55:36.525 Uh degree of variability uh and 00:55:36.525 --> 00:55:39.365 it's pretty clear that the olfactory system has to use all sorts 00:55:39.365 --> 00:55:42.445 of tricks to deal with this one of the tricks is going 00:55:42.445 --> 00:55:46.985 back to the original point if you have a very wide variety of affinity distributions 00:55:46.985 --> 00:55:54.545 then if you do this um you can still have some receptors binding um or their 00:55:54.545 --> 00:55:58.365 binding is changing even when the concentration is very high because you might 00:55:58.365 --> 00:56:01.785 Even though you probably have, for a given odor, 00:56:02.165 --> 00:56:07.925 you'll have some high affinity receptors, which will get bound very quickly, lower concentrations, 00:56:08.385 --> 00:56:12.945 and they will quickly become saturated in their response, which doesn't inform 00:56:12.945 --> 00:56:16.745 you anymore and get any more information out of those because they're just saturated. 00:56:17.785 --> 00:56:23.705 But then, if you've also got this population with lower affinity receptors, 00:56:24.145 --> 00:56:28.185 then even though the concentration is increasing a lot, you're still getting 00:56:28.185 --> 00:56:33.405 some information from those lower affinity receptors. 00:56:33.405 --> 00:56:41.485 So by using this very wide range of receptor affinities, you can still get information 00:56:41.485 --> 00:56:44.145 through to the system that there's a change, 00:56:44.465 --> 00:56:54.325 which you can, that then gives you choices in terms of the coding to still be 00:56:54.325 --> 00:56:57.385 able to at least receive the information over a large range. 00:56:57.385 --> 00:56:58.625 Now that's one strategy. 00:56:59.305 --> 00:57:05.545 Our study also shows by looking at these antagonisms, and we made a, 00:57:05.545 --> 00:57:11.005 again with Manuel, this was work with Manuel in Madrid, 00:57:11.385 --> 00:57:15.565 we looked at these receptor antagonisms. 00:57:15.565 --> 00:57:18.545 And there are some very nice and simple 00:57:18.545 --> 00:57:23.265 pharmacological equations that we could apply to olfaction which hasn't really 00:57:23.265 --> 00:57:29.305 been done at all before using these efficacy models and there's something called 00:57:29.305 --> 00:57:35.225 an operational model in pharmacology which is used effectively to describe drug 00:57:35.225 --> 00:57:37.505 binding on cells and so on. 00:57:38.605 --> 00:57:44.705 So in these studies with Manuel Sanchez-Montanés you started to develop more 00:57:44.705 --> 00:57:45.885 this notion of a ratio code. 00:57:46.285 --> 00:57:50.705 Yeah, exactly. So what does it exactly mean? So what you find is when you apply 00:57:50.705 --> 00:57:53.205 these very simple description. 00:57:55.165 --> 00:58:01.205 Then what you get is that you get effectively two parameters for each ligand, 00:58:01.265 --> 00:58:05.045 both an efficacy, which tells you how much it contributes to a cellular response, 00:58:05.305 --> 00:58:07.065 and its affinity of binding. 00:58:07.065 --> 00:58:13.025 And when you operate in this way or think about olfaction in this way, 00:58:13.105 --> 00:58:18.345 you get this nice property that something with a low efficacy could effectively 00:58:18.345 --> 00:58:24.505 block a space and you can get, when the receptors start to become filled, 00:58:24.685 --> 00:58:27.145 you can get effectively a competitive binding regime. 00:58:29.563 --> 00:58:33.103 Where you get certain sites are being filled but not contributing, 00:58:33.263 --> 00:58:37.723 for instance, or other sites that are filled and contributing a lot. 00:58:38.023 --> 00:58:41.603 And what you find when you look at the equation is that it very simply drops 00:58:41.603 --> 00:58:49.223 out that instead of the response of the cell being determined by the distribution 00:58:49.223 --> 00:58:53.863 of the concentrations of the different ligands within a mixture. 00:58:54.943 --> 00:58:58.683 In the high regime when there's this competition, it actually turns out that 00:58:58.683 --> 00:59:03.423 it then becomes depending upon the ratio of the concentration. 00:59:03.803 --> 00:59:07.443 So it's no longer concentration-dependent, and 00:59:07.443 --> 00:59:12.983 it's in this mode that we call ratio mode operation of the neuron when it's 00:59:12.983 --> 00:59:20.523 operating in this high occupancy regime that you lose the concentration dependence 00:59:20.523 --> 00:59:24.103 and you gain a sort of concentration invariance property, 00:59:24.103 --> 00:59:30.483 which we've shown holds over very large numbers, orders, and magnitudes. 00:59:30.683 --> 00:59:36.703 So although you might think it needs some very complex neural processing to 00:59:36.703 --> 00:59:37.723 maybe solve this problem, 00:59:38.383 --> 00:59:42.623 it's kind of a nice result in the sense that maybe just directly at the periphery 00:59:42.623 --> 00:59:47.423 in terms of competing for sites may actually solve this problem for you almost for free, right? 00:59:48.203 --> 00:59:52.483 Well, nothing for free, right? Well, for free in terms of the neural processing, right? 00:59:52.543 --> 00:59:58.003 Because you don't have to. Okay, that's fair enough, but this works under certain 00:59:58.003 --> 01:00:01.543 assumptions of binding and receptor distribution and so on, right? 01:00:01.623 --> 01:00:06.383 So you could also imagine that in these highly saturated regimes, 01:00:06.663 --> 01:00:12.243 you actually have so few unoccupied receptors left that the probability to actually 01:00:12.243 --> 01:00:18.343 exploit this invariant or concentration invariant coding, that probability is then very low. 01:00:19.391 --> 01:00:22.991 Um how do you mean exploit well that 01:00:22.991 --> 01:00:25.771 so you're saying look the trick is 01:00:25.771 --> 01:00:28.871 that um at some point you have saturated your 01:00:28.871 --> 01:00:32.511 receptor sheet but still there are few receptors available that you can get 01:00:32.511 --> 01:00:37.431 you can use yeah and those can then tell you uh about the presence or absence 01:00:37.431 --> 01:00:41.931 of a certain odor independent of the concentration exactly right but now i'm 01:00:41.931 --> 01:00:47.031 saying well that's that's fine But if I have saturated my receptor sheet for, let's say, 99%, 01:00:47.651 --> 01:00:52.531 then the probability for these unbound molecules to still hit those receptors 01:00:52.531 --> 01:00:58.191 might be so low that actually I will not be able to exploit my concentration invariant encoding. 01:00:58.871 --> 01:01:02.811 Yeah, no, that's a good point. But you have to keep in mind that you have a 01:01:02.811 --> 01:01:05.351 very wide variety of affinities as well. 01:01:05.471 --> 01:01:12.331 So whilst there may be a subpopulation of receptors that are forced into this high occupancy regime, 01:01:12.331 --> 01:01:15.091 gene it's quite likely that it's the 01:01:15.091 --> 01:01:17.911 case it's probably only a small subpopulation out of 01:01:17.911 --> 01:01:21.291 the total and that there will be large 01:01:21.291 --> 01:01:24.671 numbers of other receptors with lower affinities that 01:01:24.671 --> 01:01:27.831 will not therefore be in a they will be in a relatively low. 01:01:27.831 --> 01:01:30.851 Occupancy and so they are still very much sensitive 01:01:30.851 --> 01:01:34.411 to chemical changes so if this. 01:01:34.411 --> 01:01:37.851 Is true and we're still look we've fitted it to the data. 01:01:37.851 --> 01:01:41.271 That we've received and it matches the data that we've received. 01:01:41.271 --> 01:01:44.151 On antagonism for instance from tahara and an earlier 01:01:44.151 --> 01:01:47.331 paper by anderson on insect receptors 01:01:47.331 --> 01:01:50.891 all the data we fitted works but if 01:01:50.891 --> 01:01:53.971 it's true it predicts that there's a 01:01:53.971 --> 01:01:58.051 very strong prediction that there would be likely for any 01:01:58.051 --> 01:02:01.511 natural odors at 01:02:01.511 --> 01:02:04.571 naturally relevant or behaviorally relevant 01:02:04.571 --> 01:02:08.031 conditions there will be a subpopulation in 01:02:08.031 --> 01:02:11.031 this high occupancy regime that that 01:02:11.031 --> 01:02:14.711 will give you uh if you like the relational features 01:02:14.711 --> 01:02:18.131 between a complex odor so for instance in a coffee uh 01:02:18.131 --> 01:02:20.891 they'll be you you know the quality of the coffee is very much 01:02:20.891 --> 01:02:23.731 about what is the ratio of of 01:02:23.731 --> 01:02:26.911 this peak of this compound compared to another compound as 01:02:26.911 --> 01:02:29.731 to whether that's a pleasant coffee for you and you don't 01:02:29.731 --> 01:02:33.011 necessarily need to need a certain concentration to 01:02:33.011 --> 01:02:36.051 know that that sort of that works across many orders of 01:02:36.051 --> 01:02:39.311 concentration but on the other hand you 01:02:39.311 --> 01:02:42.071 want probably other parts of the olfactory system that are able 01:02:42.071 --> 01:02:45.451 to tell you about concentration right and it's already been demonstrated that 01:02:45.451 --> 01:02:48.071 in fact the olfactory system this is one of 01:02:48.071 --> 01:02:54.031 its amazing properties right that in fact that you can separate um you can separate 01:02:54.031 --> 01:02:59.711 both concentration information for certain ligands and also it's sort of what 01:02:59.711 --> 01:03:04.571 we call configurable properties which is basically the sort of combined percepts 01:03:04.571 --> 01:03:07.311 so So on the one hand, you've got to sort of gestalt. 01:03:08.319 --> 01:03:11.039 Sort of uh odor or flavor which is 01:03:11.039 --> 01:03:14.699 to do with all these sort of ratio properties which are invariant across 01:03:14.699 --> 01:03:17.479 concentrations and on the other hand you've got this sort 01:03:17.479 --> 01:03:20.819 of opposite of gestalt which is able to sort of tease them 01:03:20.819 --> 01:03:24.639 apart and look at the separate component concentrations and 01:03:24.639 --> 01:03:29.179 with this mechanism you you then have a population when you combine this mechanism 01:03:29.179 --> 01:03:34.599 with a wide range of affinities which we know is in the olfactory system you 01:03:34.599 --> 01:03:38.259 have this very nice feature that you You would have these sort of subpopulations 01:03:38.259 --> 01:03:43.519 doing different things to either support either gestalt processing or non-gestalt processing. 01:03:43.939 --> 01:03:49.199 But now, with that separation, would you imagine that there is a possibility 01:03:49.199 --> 01:03:53.359 for the brain to control this in a more top-down fashion, so you get something 01:03:53.359 --> 01:03:54.999 like an olfactory attention? 01:03:56.039 --> 01:04:01.259 This is a very good point. I think probably not at the binding level, right? Right. 01:04:01.279 --> 01:04:06.159 But we do know that certainly the feed forward gain in these different pathways 01:04:06.159 --> 01:04:08.839 from the different receptors, we know very well that this is... 01:04:08.839 --> 01:04:09.819 That's a level of receptor neurons though. 01:04:10.119 --> 01:04:13.299 Not necessarily the receptor. Well, there is some evidence that there's even 01:04:13.299 --> 01:04:15.119 feedback actually to the receptors. 01:04:15.379 --> 01:04:22.039 Okay. It's still not really well studied, but we know very, very surely that 01:04:22.039 --> 01:04:23.579 the readout of the glomeruli themselves, 01:04:23.759 --> 01:04:28.879 which is effectively at an early stage of that processing, is reflecting the 01:04:28.879 --> 01:04:33.419 excitatory drive of the common set of olfactory receptors. 01:04:33.699 --> 01:04:38.859 We know that those are under very complex inhibitory control and gain control, 01:04:39.039 --> 01:04:44.859 top-down imposed from the piriform cortex and through centrifugal feedback coming. 01:04:44.859 --> 01:04:49.439 I mean, coming back to the system, we know that that has a very strong influence 01:04:49.439 --> 01:04:52.199 on the type of codes that go forward to the higher level. 01:04:52.239 --> 01:04:57.699 So it's very likely that there's an additional sorts of then selection mechanisms. 01:04:57.779 --> 01:05:03.719 If you could imagine there are subsets of these different receptors, 01:05:04.079 --> 01:05:08.979 you could imagine that then downstream you could select between those that are 01:05:08.979 --> 01:05:14.739 either in a ratio mode or in a concentration mode, for instance. 01:05:14.859 --> 01:05:18.859 So that might be another way to deal with that problem that's not only feed 01:05:18.859 --> 01:05:22.859 forward and dependent on the receptors themselves. So this could be synergistic mechanisms. 01:05:23.299 --> 01:05:29.119 All right. So now we have a bit of an idea how we get our first responses in 01:05:29.119 --> 01:05:31.579 and then in the modeling work that you have been doing. 01:05:32.238 --> 01:05:36.078 You try to also get a clear idea about the actual encoding of now these odors 01:05:36.078 --> 01:05:37.838 at the level of olfactory bulb. 01:05:38.038 --> 01:05:41.578 So what can you tell us about the encoding that then happens of these odors? 01:05:41.978 --> 01:05:45.478 Sure, yeah. So the main feature in the olfactory bulb, which is really the first 01:05:45.478 --> 01:05:47.018 site of neuronal processing, 01:05:47.098 --> 01:05:55.238 of all this stuff coming in from 10 million receptors in the olfactory mucosa, 01:05:55.318 --> 01:05:57.538 as it comes through the skull, 01:05:57.718 --> 01:06:00.538 the cribriform plate, to make contact with these glomeruli. 01:06:00.538 --> 01:06:05.118 And the first thing that happens is they converge into these glomeruli structures, 01:06:05.458 --> 01:06:10.398 and they are innervated by these mitral tufted cells. 01:06:10.678 --> 01:06:16.438 And basically the story is that one single mitral tufted cell is innervating a single glomerulus. 01:06:16.698 --> 01:06:22.398 So they're acting as a readout, if you like, of a single chemotopic feature, 01:06:22.598 --> 01:06:28.838 because they're only innervating those common GPCR or receptor types. Okay. 01:06:29.038 --> 01:06:33.038 And then all of those mitral receptor population, 01:06:33.458 --> 01:06:37.338 which acting as the main readout of the bulb to the higher centers like piriform 01:06:37.338 --> 01:06:44.618 cortex and so on, are under very quite complex lateral control. 01:06:44.618 --> 01:06:53.498 So for instance, a mitral cell is under control from what are called granule 01:06:53.498 --> 01:06:56.038 cells, which make lateral connections across the bulb. 01:06:56.658 --> 01:07:02.098 And it's even been demonstrated that granule cells may even connect across the 01:07:02.098 --> 01:07:05.498 whole length of the bulb, with quite specific connections, in fact. 01:07:05.498 --> 01:07:08.738 So there's this network of 01:07:08.738 --> 01:07:12.458 quite specific connections that are going laterally they're 01:07:12.458 --> 01:07:15.958 forcing an inhibitory drive on the mitral cells 01:07:15.958 --> 01:07:22.338 and there's at least two effects of this well three really one effect is that 01:07:22.338 --> 01:07:29.678 it's been demonstrated quite clearly that this lateral inhibition has a sort 01:07:29.678 --> 01:07:35.718 of on-center-off-surround type property So because it's also, 01:07:35.818 --> 01:07:42.978 it's a complex type of inhibition and it's a unique synapse called a dendro 01:07:42.978 --> 01:07:47.218 dendritic synapse, which is actually pretty unique in the whole brain in these granule cells. 01:07:47.938 --> 01:07:52.298 You don't really see it anywhere else. And the effect of this dendrodendritic 01:07:52.298 --> 01:07:56.818 synapse is that a mitral cell will activate one of these lateral granule cells, 01:07:57.158 --> 01:08:02.658 which will then in turn inhibit a distant mitral cell target, 01:08:02.858 --> 01:08:06.738 but it will also excite itself back because it's a two-way process. 01:08:07.293 --> 01:08:10.233 Didactic synapse and the 01:08:10.233 --> 01:08:13.193 effect of all of this anyway is that you get a sort of on center 01:08:13.193 --> 01:08:16.993 of surround type property because it's sort of inhibiting with 01:08:16.993 --> 01:08:19.993 a sustain with some nice time 01:08:19.993 --> 01:08:23.053 dynamics exactly and it's been demonstrated that 01:08:23.053 --> 01:08:26.733 this very nicely decorrelates and sharpens the representation 01:08:26.733 --> 01:08:29.673 of these odors at this stage in 01:08:29.673 --> 01:08:33.193 the olfactory bulb so that's one clear thing that happens another clear 01:08:33.193 --> 01:08:36.533 thing that happens is because of this um very nice 01:08:36.533 --> 01:08:39.353 balance of excitation drive coming in from 01:08:39.353 --> 01:08:42.353 the receptors via the glomeruli with this 01:08:42.353 --> 01:08:45.253 lateral inhibition that you're getting some very complex 01:08:45.253 --> 01:08:52.213 dynamics right so in fact the olfactory bulb was the i think it's fair to say 01:08:52.213 --> 01:08:57.473 actually the very first neuronal recordings in in the brain by lord adrian at 01:08:57.473 --> 01:09:02.353 cambridge measured also the first oscillations in the olfactory bulb. 01:09:02.513 --> 01:09:09.153 And oscillations are the very key feature and they come about due to this balance 01:09:09.153 --> 01:09:15.013 of lateral inhibition and excitatory drive. 01:09:15.293 --> 01:09:19.673 So there's these very complex sort of dynamics going on together with these 01:09:19.673 --> 01:09:20.813 sort of sharpening things. 01:09:21.693 --> 01:09:24.513 And the third feature is that we, as we mentioned before, that these 01:09:24.513 --> 01:09:29.173 granule cells also act as targets for the centrifugal feedback coming back from 01:09:29.173 --> 01:09:35.453 the higher centers to selectively change this lateral inhibitory mechanism to 01:09:35.453 --> 01:09:38.013 effectively selectively change 01:09:38.013 --> 01:09:41.733 the gain in different channels so that you may listen, for instance, 01:09:41.873 --> 01:09:45.813 more to some types of chemical information and suppress others, right? 01:09:45.913 --> 01:09:47.573 Right, exactly. So this has already been shown to be very important. 01:09:47.593 --> 01:09:52.333 But now, do you see these mitral cells and linked to this complex network of 01:09:52.333 --> 01:09:56.073 granule cells and they are sampling the glomeruli, 01:09:57.133 --> 01:10:03.713 are they just adding up activity and emitting spikes or is the encoding of these odors more involved? 01:10:04.433 --> 01:10:08.453 Like, for instance, does it depend on population responses? Does it depend on 01:10:08.453 --> 01:10:10.273 the temporal structuring of these responses? 01:10:11.573 --> 01:10:16.313 Yeah, I think evidence shows, for instance, Detlef Schildert-Göttingen has shown 01:10:16.313 --> 01:10:21.313 that the latency of the mitral cell responses is very crucial to different compounds, 01:10:22.373 --> 01:10:25.193 which you might expect right because if you've got 01:10:25.193 --> 01:10:27.973 this dendro dendritic thing you can you have 01:10:27.973 --> 01:10:30.833 these periods of inhibition so what can 01:10:30.833 --> 01:10:33.933 happen is that some mitral cells can selectively shut 01:10:33.933 --> 01:10:39.053 down other mitral cells in these channels for certain periods of time and so 01:10:39.053 --> 01:10:44.433 what you seem to find is that if you look at the population of um olfactory 01:10:44.433 --> 01:10:48.153 bulb mitral cell responses that you get these different latencies in the different 01:10:48.153 --> 01:10:51.613 channels and it's been demonstrated that these latencies are very important 01:10:51.613 --> 01:10:53.133 for coding the different odors, right? 01:10:53.253 --> 01:10:56.453 Do these latencies also express the binding kinetics? 01:10:58.433 --> 01:11:03.633 You could imagine that a receptor with high affinity to a ligand will trigger 01:11:03.633 --> 01:11:05.853 a short latency response in its target. 01:11:06.928 --> 01:11:13.768 Yeah, I think that's quite possible, but I think it's also reflecting the timing 01:11:13.768 --> 01:11:19.088 of these lateral inhibitions and the timing of the dendritic synapses, 01:11:19.088 --> 01:11:20.768 so I think it's probably both of those together. 01:11:20.768 --> 01:11:27.668 The other very interesting feature, which was shown very nicely by Rainer Freydig. 01:11:27.908 --> 01:11:36.628 Was to show that the spike timing of individual mitral cells in this oscillation 01:11:36.628 --> 01:11:43.768 in the olfactory bulb may well be very important for coding different aspects of a complex odor. 01:11:43.768 --> 01:11:50.248 So, for instance, if you're firing early on in the cycle with another set of 01:11:50.248 --> 01:11:52.468 receptors, this may have one meaning. 01:11:52.588 --> 01:11:58.548 And if you're firing with another subnetwork within this olfactory bulb space 01:11:58.548 --> 01:12:02.908 later on in the oscillatory cycle, this may mean another thing. 01:12:02.908 --> 01:12:09.268 And so he was crucial in introducing this idea of multiplexed odor codes for mixtures. 01:12:09.288 --> 01:12:15.148 So maybe you have a mixture where you have different synchronous firing between 01:12:15.148 --> 01:12:17.568 different subpopulations at different points in time. 01:12:17.568 --> 01:12:23.508 And that these may well be telling higher centers all sorts of information about 01:12:23.508 --> 01:12:29.628 grouping or binding of different chemical properties to tell you about, 01:12:29.748 --> 01:12:34.068 you know, much more complex chemical stimuli that we know after all is like 01:12:34.068 --> 01:12:36.568 coffee with 400 compounds or whatever. 01:12:37.148 --> 01:12:41.028 Okay, so now we have a bit of an idea of the biology of olfaction. 01:12:41.828 --> 01:12:46.748 In sort of a parallel existence, you also worry a lot about building artificial 01:12:46.748 --> 01:12:51.568 olfaction, artificial noses, olfactory machines, olfactory robots, 01:12:51.708 --> 01:12:53.988 and so on. So what are the challenges there exactly? 01:12:54.861 --> 01:13:00.521 Yeah, so it's a kind of parallel life to kind of build technology. 01:13:00.661 --> 01:13:06.721 I mean, similar to you, actually, that I think we both follow this idea that 01:13:06.721 --> 01:13:09.561 if we really understand it, we can build it, right? 01:13:09.561 --> 01:13:13.161 So let's go and see the proof by, can we take these principles, 01:13:13.401 --> 01:13:17.401 put them into some type of operational system, test what they do. 01:13:18.161 --> 01:13:24.601 And it introduces a whole bunch of challenges that a lot of them actually have 01:13:24.601 --> 01:13:26.121 very little to do with the biology. 01:13:26.281 --> 01:13:31.801 Because, for instance, just getting a good set of sensory signals to start with 01:13:31.801 --> 01:13:37.921 and reliable sensory responses is already a big challenge in chemical sensing. 01:13:37.921 --> 01:13:45.841 The idea of so-called building chemical analysis systems along the lines of 01:13:45.841 --> 01:13:49.021 following the biological system is not a new idea. 01:13:49.021 --> 01:13:57.321 There was a paper in Nature in 1982 by Krishna Persaud and George Dodd, 01:13:57.321 --> 01:14:04.161 who introduced this idea of putting together groups of different tuning chemical 01:14:04.161 --> 01:14:08.501 sensors in arrays and to look at population codes of these responses, 01:14:08.981 --> 01:14:15.181 which is clearly similar to how this is solved in the biological system. 01:14:15.181 --> 01:14:18.161 System um and you know 01:14:18.161 --> 01:14:21.741 so then what you need to do is you need to you know you face a challenge of 01:14:21.741 --> 01:14:26.041 finding different chemical technologies that give you nicely orthogonal and 01:14:26.041 --> 01:14:32.361 decorrelated sort of sensor responses um and that are sort of robust and give 01:14:32.361 --> 01:14:38.081 you reliable signals um that have sort of have reasonable noise properties, 01:14:38.441 --> 01:14:41.881 can be easily deposited. 01:14:42.581 --> 01:14:49.401 Can maybe have a wide range of different molecular interactions and have sort 01:14:49.401 --> 01:14:51.281 of broad tunings to different things. 01:14:52.761 --> 01:14:54.681 Maybe reversible responses. 01:14:54.701 --> 01:14:59.061 So you don't really necessarily want sensors that don't ever recover from an interaction, right? 01:14:59.181 --> 01:15:03.261 So there's a whole bunch of issues with chemical sensors. 01:15:03.481 --> 01:15:11.181 And I And I think whilst things are improving, there's actually a massive range 01:15:11.181 --> 01:15:13.221 of chemical sensor technologies you can use. 01:15:14.121 --> 01:15:17.581 There are clearly many issues. For instance, one of the main issues, 01:15:17.641 --> 01:15:23.321 I think, in chemical sensors is this concept of sensor drift. 01:15:23.621 --> 01:15:28.301 So we know that pretty much any chemical sensor, its properties will be quite 01:15:28.301 --> 01:15:29.501 non-stationary over time. 01:15:30.141 --> 01:15:33.741 So it may have one set of tunings at one particular point in time. 01:15:33.981 --> 01:15:37.501 You may come back there in a month's time and have a completely different set 01:15:37.501 --> 01:15:39.781 of tunings, baseline parameters and so on. 01:15:40.141 --> 01:15:45.321 These can change in very complex and difficult to predict ways. Uh-huh. 01:15:47.768 --> 01:15:51.848 On one level, okay, so this can be quite frustrating perhaps when you try and build these systems. 01:15:51.948 --> 01:15:57.188 But on another level, we also have to appreciate that every individual person 01:15:57.188 --> 01:16:01.148 has pretty much a unique, we haven't discussed this, but we all have a unique 01:16:01.148 --> 01:16:06.908 genetic fingerprint of the subset of thousand olfactory receptors that we express individually. 01:16:07.428 --> 01:16:11.128 So coffee actually smells completely different to you than it does to me. 01:16:12.028 --> 01:16:13.888 Well, but maybe not because earlier 01:16:13.888 --> 01:16:17.208 you said that there's actually a low dimensional manifold. default. 01:16:17.608 --> 01:16:21.728 Maybe we map it onto something that we can describe in similar terms, 01:16:21.888 --> 01:16:26.468 but what you can guarantee is that the receptor information that comes into 01:16:26.468 --> 01:16:29.728 your system is completely different to that that comes into mine, right? 01:16:29.968 --> 01:16:33.128 But you're right that this has to be somehow mapped. 01:16:33.348 --> 01:16:37.228 That's kind of my point really, is that somehow the olfactory system has to 01:16:37.228 --> 01:16:42.348 take care of this and map it all onto a stable representation of something that 01:16:42.348 --> 01:16:45.248 means coffee. And one that's It's unitary, of which you can say that's coffee, 01:16:45.408 --> 01:16:47.248 not that you say, oh, it's 400 compounds. 01:16:47.588 --> 01:16:51.988 Yeah, exactly. And the other point is that actually in a month's time, 01:16:52.188 --> 01:16:57.108 because these olfactory receptors are basically the only receptors that are 01:16:57.108 --> 01:16:59.688 in direct contact with the outside world, 01:16:59.928 --> 01:17:04.888 they're really at the tough end of things, unlike most neurons, 01:17:04.988 --> 01:17:10.868 which are very nicely protected in the neural systems, that they basically suffer damage. 01:17:10.868 --> 01:17:16.408 And so in a month's time, you and I will both have a completely different set 01:17:16.408 --> 01:17:19.668 of olfactory receptors than those that we have now. 01:17:19.968 --> 01:17:23.348 Now they'll be similarly genetically determined, we'll have a sort of similar 01:17:23.348 --> 01:17:27.968 subset of the total, because that's fairly determined, even that may change, 01:17:28.168 --> 01:17:31.728 but somehow the olfactory system has to deal with all of this, 01:17:31.828 --> 01:17:33.668 and in a sense that's kind of like drift, right? 01:17:33.668 --> 01:17:39.988 Because we already know that, for instance, as the olfactory receptors are developing, 01:17:40.268 --> 01:17:41.988 that their tuning is changing. 01:17:42.048 --> 01:17:44.968 They actually become more broadly tuned over time. They start more specific. 01:17:45.128 --> 01:17:48.228 They become more broadly tuned. And so there's a whole drift process actually 01:17:48.228 --> 01:17:49.708 going on in the receptors as well. 01:17:50.588 --> 01:17:54.868 And, okay, that drift process is probably very different from that that we see in the receptors. 01:17:54.988 --> 01:18:00.088 But it's still a challenge that the neuromorphic system might nicely be able 01:18:00.088 --> 01:18:04.668 to solve. Because if you look at a classical engineering solutions to this, 01:18:04.828 --> 01:18:09.328 then it hasn't quite totally solved all of these issues. 01:18:09.428 --> 01:18:14.468 Tell me, what are the different approaches people have taken to solve this problem, technically? 01:18:14.808 --> 01:18:18.048 Of drift? No, and chemical sensing. Let's start with that. Okay. 01:18:18.763 --> 01:18:22.243 Okay, well, yeah, I mean, that's a very broad question, right? 01:18:22.403 --> 01:18:29.023 You can go all the way from developing biosensors through to developing highly 01:18:29.023 --> 01:18:34.263 specific sensors for particular ligands that you're interested in with a problem, 01:18:34.363 --> 01:18:41.683 all the way through to developing sort of electronic noses, using, 01:18:41.903 --> 01:18:44.803 you know, arrays of broadly tuned compounds, compounds 01:18:44.803 --> 01:18:47.863 through to using sort 01:18:47.863 --> 01:18:50.763 of mass spectrometry ideas where you're trying to 01:18:50.763 --> 01:18:55.983 basically measure specific molecular features directly like we were talking 01:18:55.983 --> 01:19:02.023 about originally there might be mass features or through to using gas chromatography 01:19:02.023 --> 01:19:05.563 effects where you're trying to select between different molecules due to how 01:19:05.563 --> 01:19:07.963 much they sorb into different materials or not. 01:19:09.783 --> 01:19:12.723 There's a whole bunch of different strategies uh so 01:19:12.723 --> 01:19:15.923 what's the sensitivity of these systems compared to the biological system 01:19:15.923 --> 01:19:19.423 this is a great question so um well actually 01:19:19.423 --> 01:19:23.783 we did a study right a long time ago which we're together remember that um we 01:19:23.783 --> 01:19:29.783 we we we had this question in mind uh when we first came to this nice artificial 01:19:29.783 --> 01:19:33.743 moth project that we ran that we were wondering well okay so we know that the 01:19:33.743 --> 01:19:38.703 moth is very sensitive overall it's like the world specialist in olfactory detection, 01:19:39.323 --> 01:19:46.363 might we expect that, in fact, its receptors are, you know, orders of magnitude 01:19:46.363 --> 01:19:49.683 more sensitive than, say, for instance, a metal oxide sensor? 01:19:49.863 --> 01:19:54.083 Like a commercial one you just buy. Yeah, which is the most common one that 01:19:54.083 --> 01:19:55.963 tends to be used in these kinds of arrays. 01:19:56.363 --> 01:19:59.523 And indeed, what we found when we took those measurements were that, 01:19:59.643 --> 01:20:03.703 in fact, no, you know, that these sensitivities were kind of more comparable 01:20:03.703 --> 01:20:06.843 than you might expect. So that was kind of a surprise. 01:20:07.083 --> 01:20:11.183 But you also, of course, in doing any comparison like that, you have to take 01:20:11.183 --> 01:20:16.303 into account that we're not necessarily giving that receptor its ideal ligand. 01:20:17.163 --> 01:20:24.503 We may be particularly one extreme end of the affinity for that particular ligand or efficacy. Yeah. 01:20:25.029 --> 01:20:28.129 Parameter you know it may have other ones that it's much more sensitive to 01:20:28.129 --> 01:20:31.249 but it's very difficult to tell unless you measure everything which 01:20:31.249 --> 01:20:34.129 is impossible no but wait but but there's an important point here 01:20:34.129 --> 01:20:37.549 right that sort of intuitively you would say like well you know if i have to 01:20:37.549 --> 01:20:42.049 increase the sensitivity i should be just optimizing at the sensor front end 01:20:42.049 --> 01:20:46.409 of this whole system if i just have high affinities there i'll be just fine 01:20:46.409 --> 01:20:50.449 in my detection but maybe but maybe that's not the right philosophy no you really 01:20:50.449 --> 01:20:52.909 can't do that because then everything binds to everything. 01:20:53.149 --> 01:20:56.589 Right and then you don't have any information anymore so 01:20:56.589 --> 01:20:59.849 already in the story that we saw with the antagonism as 01:20:59.849 --> 01:21:02.649 we talked about before is that it really to get this 01:21:02.649 --> 01:21:05.729 ability to do this over a range and to 01:21:05.729 --> 01:21:08.609 to also be able to get the combination in a 01:21:08.609 --> 01:21:12.069 single olfactory system of sensitivity together 01:21:12.069 --> 01:21:15.109 with high specificity to get those 01:21:15.109 --> 01:21:17.869 two things together and to get them over a 01:21:17.869 --> 01:21:20.909 broad dynamic range you cannot just be sensitive to everything 01:21:20.909 --> 01:21:23.689 right you have to you have to have this 01:21:23.689 --> 01:21:26.729 range of binding so that you can basically you 01:21:26.729 --> 01:21:29.769 have to you have to have it so that you can have a kind of attentional mechanism 01:21:29.769 --> 01:21:35.489 to right to be able to focus in particular but it's not respect that the point 01:21:35.489 --> 01:21:40.769 is maybe if you want to get a highly sensitive olfactory system the real tricks 01:21:40.769 --> 01:21:45.989 sit more at the processing end of things than at the sense of content yeah i 01:21:45.989 --> 01:21:46.749 would I would agree with that. 01:21:46.789 --> 01:21:50.229 I mean, it's certainly clear that there's so many tricks that are played, 01:21:50.289 --> 01:21:56.549 you know, by the biology in terms of getting a sensitive, specific. 01:21:56.869 --> 01:22:04.249 Robust, you know, signal representation out of this that the more I look at 01:22:04.249 --> 01:22:08.629 every level of the system, it's completely optimized to achieving just that. 01:22:08.629 --> 01:22:12.409 But now the amazing thing is that in one of your projects with your collaborators. 01:22:12.769 --> 01:22:18.669 You built an artificial nose where you specifically looked at the impact of 01:22:18.669 --> 01:22:22.189 the mucus layer in processing of olfaction. 01:22:22.289 --> 01:22:27.609 So why did you add that feature to an artificial chemical sensor? 01:22:27.769 --> 01:22:32.109 Yeah, it's a good question. Well, in existing electronic noses up to now, 01:22:32.269 --> 01:22:36.329 really there are two forms of information that have been exploited. 01:22:37.149 --> 01:22:41.249 Uh, the first, uh, mechanism is this, uh, population code. 01:22:41.369 --> 01:22:47.149 So for any particular complex or simple molecule, you've got a particular fingerprint 01:22:47.149 --> 01:22:51.109 of, um, receptor responses. Mm-hmm. 01:22:52.360 --> 01:23:00.800 This is clearly, this is what you would call a spatial code of receptor population 01:23:00.800 --> 01:23:06.820 tuning that give you stimulus-dependent information, enable you to tell what's out there in the world. 01:23:07.060 --> 01:23:11.680 You've got a second mechanism, clearly, which is that each of those receptors 01:23:11.680 --> 01:23:16.400 or sensors is giving you, in itself, specific temporal information. 01:23:16.400 --> 01:23:21.500 So this gives you sort of, en masse, gives you a sort of temporal code, 01:23:21.700 --> 01:23:23.720 temporal population code as you might call it. 01:23:24.560 --> 01:23:29.420 But then there's this very much less studied aspect of olfaction, 01:23:29.560 --> 01:23:35.160 which is that not only are the sensors themselves maybe giving out temporal 01:23:35.160 --> 01:23:37.900 information, depending upon what they're binding with. 01:23:38.940 --> 01:23:44.280 That in fact the stimulus delivery itself may have some nice temporal properties. 01:23:44.280 --> 01:23:47.260 And so we went about building what 01:23:47.260 --> 01:23:51.220 we called an artificial mucosa and this was with collaborators at 01:23:51.220 --> 01:23:57.500 university of warwick including julian gardner who build microsystems so i don't 01:23:57.500 --> 01:24:01.820 build any sensors myself i build these sort of strategies for processing architectural 01:24:01.820 --> 01:24:07.060 ideas for these types of systems and so we specifically you know we it was a 01:24:07.060 --> 01:24:10.580 very nice example of where you see a few neuroscience papers, 01:24:11.400 --> 01:24:16.620 and this basically inspires you to then go and build a technology that you think 01:24:16.620 --> 01:24:18.600 can do better than what we have at the moment. 01:24:18.740 --> 01:24:24.340 And what we ended up building was effectively a microchannel or an artificial 01:24:24.340 --> 01:24:27.860 mucosa where you have a sort of sniff. 01:24:28.080 --> 01:24:32.280 So rather than in many electronic noses, you're sampling for long periods of 01:24:32.280 --> 01:24:33.960 time with very controlled flow rates. 01:24:35.240 --> 01:24:39.180 This is a very different strategy, where you just effectively have a pulse of 01:24:39.180 --> 01:24:42.400 odour that flows through a microchannel. 01:24:42.746 --> 01:24:46.426 And then what you have inside that microchannel, similar to as we talked about 01:24:46.426 --> 01:24:51.806 in the nose, is you have a thin mucosal layer, maybe a few hundred microns thick. 01:24:52.486 --> 01:24:56.006 And this can be of different materials, but you can take ideas, 01:24:56.086 --> 01:24:59.706 for instance, from gas chromatography that use what are called stationary phases. 01:25:00.246 --> 01:25:04.386 And these stationary phases are specific materials you can go and buy off the shelf. 01:25:04.386 --> 01:25:08.306 And they have very beautiful selective properties partitioning 01:25:08.306 --> 01:25:13.886 properties of certain groups of compounds such as hydrophobic or hydrophilic 01:25:13.886 --> 01:25:17.566 compounds would like to be in those stationary phases or like that they would 01:25:17.566 --> 01:25:21.926 like to stay away from them and so what you find is that as you have an odor 01:25:21.926 --> 01:25:27.026 pulse going through a column with these kind of materials you get very beautiful temporal 01:25:27.186 --> 01:25:33.786 profiles in the stationary phases at different points in time that depend in 01:25:33.786 --> 01:25:39.466 very complex ways on these different sorption properties and other properties 01:25:39.466 --> 01:25:45.306 of the molecules that impose an additional time dynamics on the stimulus. 01:25:45.486 --> 01:25:49.046 And particularly when you've got a complex mixture, like a coffee, 01:25:49.206 --> 01:25:51.906 then imagine you've got, what you've really got is 01:25:51.906 --> 01:25:55.866 you've got 400 of these pulses going through simultaneously and you 01:25:55.866 --> 01:25:58.846 get a beautiful sort of spectrum you can think of it like a kind of 01:25:58.846 --> 01:26:01.826 prism system where you take in a complex move and it 01:26:01.826 --> 01:26:04.726 splits it in some way right exactly in a spatial temporal 01:26:04.726 --> 01:26:07.706 code and it imposes this upon the receptor 01:26:07.706 --> 01:26:12.086 population and what you find when you look at the receptor sensor responses 01:26:12.086 --> 01:26:16.766 when you distribute them across this type of physical arrangement is that you 01:26:16.766 --> 01:26:22.466 get absolutely beautiful and stunning very complex extemporal information that's 01:26:22.466 --> 01:26:25.266 exquisitely stimulus dependent, right? 01:26:25.306 --> 01:26:29.006 Because it depends upon exactly where that molecule was at a particular point 01:26:29.006 --> 01:26:30.546 in time. With how much did this improve? 01:26:31.626 --> 01:26:37.086 So basically you translate this whole idea of zoning of the receptor sheet and 01:26:37.086 --> 01:26:42.746 a mucus layer that again controls the binding dynamics of your ligands. 01:26:42.926 --> 01:26:51.546 So you translate that to a technology also basically to test its impact on processing, 01:26:51.686 --> 01:26:55.506 but what was really the improvement in classification that you now observed? 01:26:55.866 --> 01:26:59.506 Yeah, I mean, what we found was very interesting. So if you had, 01:26:59.666 --> 01:27:01.846 it's kind of an interesting result. 01:27:01.906 --> 01:27:05.186 We found in all cases, this obviously improved discrimination. 01:27:06.899 --> 01:27:12.539 Purely because you've got three mechanisms in total, and the correct comparison 01:27:12.539 --> 01:27:16.319 to do is to compare it to just when you have the first two mechanisms without 01:27:16.319 --> 01:27:20.599 this nice, funky, spatiotemporal delivery concept. 01:27:20.879 --> 01:27:24.719 So you do the direct comparison of just saying, well, when you deliver the molecules 01:27:24.719 --> 01:27:29.919 directly onto the senses without any spatiotemporal sorting of any type, 01:27:30.139 --> 01:27:32.459 how does your information compare? 01:27:32.459 --> 01:27:36.939 And so again with Manuel, who very cleverly extended the Fisher information 01:27:36.939 --> 01:27:43.019 formalism to include time so that we can actually now see not only when different 01:27:43.019 --> 01:27:48.079 stimulus dependent features and the noise properties are telling you a particular point in time, 01:27:48.099 --> 01:27:52.139 you can actually accumulate that information over time to tell you within particular 01:27:52.139 --> 01:27:56.999 time windows how much information you've collected about the stimulus and how 01:27:56.999 --> 01:27:59.459 accurate your reconstruction is of it. 01:27:59.459 --> 01:28:04.139 You can actually use this formulism then to compare in a very precise way, 01:28:04.219 --> 01:28:06.139 in a very quantitative way, 01:28:06.299 --> 01:28:11.839 how you would have fared if you tried to discriminate just from the first two 01:28:11.839 --> 01:28:14.759 mechanisms compared to when you also use this additional sorting. 01:28:15.039 --> 01:28:19.239 And what you find, in fact, there's a theorem in our paper that shows that there's 01:28:19.239 --> 01:28:25.999 no situation ever, there's no possibility that you can ever exceed the three 01:28:25.999 --> 01:28:27.399 mechanisms with the first two. 01:28:27.399 --> 01:28:30.779 So there's no situation in which you can ever do better than the three, 01:28:30.859 --> 01:28:33.139 which is kind of common sense, right? 01:28:33.219 --> 01:28:35.859 Because if you're adding more stimulus-dependent information. 01:28:36.279 --> 01:28:40.519 So you can never do worse. So that's already good news. 01:28:41.239 --> 01:28:44.339 It's more like a lower bound. It's like a lower bound. So what's your upper bound? 01:28:45.079 --> 01:28:49.619 So the other very interesting point is how, so then the other question is how 01:28:49.619 --> 01:28:53.619 much better you do totally depends on the complexity of the task. 01:28:53.619 --> 01:28:58.339 So if you set your olfactory system, then a very simple task, 01:28:58.559 --> 01:29:03.659 which is maybe, I don't know, imagine a very, very simple task of you just have 01:29:03.659 --> 01:29:08.239 a single component and you just at one point in time, you push that through 01:29:08.239 --> 01:29:12.979 your microchannel, you get a set of receptor responses and you discriminate that as one odor. 01:29:13.599 --> 01:29:17.819 Then you separately do that for a different odor. This is a very simple discrimination 01:29:17.819 --> 01:29:22.399 task with two classes. So you've got a 50% chance. 01:29:24.319 --> 01:29:28.859 And even the array by itself, depending upon the chemicals you use, 01:29:28.899 --> 01:29:33.219 will trivially separate this, then you'll find that by adding that space to 01:29:33.219 --> 01:29:34.979 the temporal thing, you're not really going to improve it, right? 01:29:35.039 --> 01:29:39.039 Because you already almost certainly probably had 100% success anyway, right? 01:29:39.559 --> 01:29:43.699 So what we found in our study was that when you push it more and more, 01:29:44.739 --> 01:29:50.319 then the effect of this third mechanism makes more and more of a difference. 01:29:50.539 --> 01:29:54.399 So for instance, if you give it a very complex task, which you could imagine 01:29:54.399 --> 01:30:00.499 one of the hardest tasks might be put through the system of, 01:30:01.183 --> 01:30:04.563 a coffee mixture with 500 compounds simultaneously. 01:30:06.083 --> 01:30:11.683 And the task of the chemical receptor array and the subsequent readout is to 01:30:11.683 --> 01:30:16.863 detect the presence or not the presence of one of those compounds in the 500, right? 01:30:16.963 --> 01:30:20.103 Right. When there's a massive overlapping spectra of all of these. 01:30:20.183 --> 01:30:24.983 This is a phenomenally difficult problem because you've got all of these receptors, 01:30:25.023 --> 01:30:30.443 responses are convolved over 500 components in a relatively short period of 01:30:30.443 --> 01:30:32.923 time of responses over just a few seconds and. 01:30:33.823 --> 01:30:38.463 Somehow over this relatively small sensor population you may I think in our 01:30:38.463 --> 01:30:41.183 micro I think we only had 30 sensors right now. 01:30:41.263 --> 01:30:46.463 We only have 30 channels compared to 10 million in there in the in the biological 01:30:46.463 --> 01:30:52.923 system Out of all of this you've got to somehow Detect one of these components. 01:30:53.503 --> 01:30:57.263 This is a very challenging problem and what 01:30:57.263 --> 01:31:00.423 you find when you apply this 01:31:00.423 --> 01:31:03.103 separation at the front end with this third mechanism is that 01:31:03.103 --> 01:31:06.443 you find that this third mechanism becomes 01:31:06.443 --> 01:31:09.283 more and more important so we found 01:31:09.283 --> 01:31:12.403 that at least one or two orders of magnitude improvement in 01:31:12.403 --> 01:31:16.243 the classification and the error reduction in 01:31:16.243 --> 01:31:19.003 the error was two orders of magnitude and we 01:31:19.003 --> 01:31:21.703 i think we only tested at a couple of 01:31:21.703 --> 01:31:24.403 either an easy and a complex task and i 01:31:24.403 --> 01:31:27.503 think there's a lot of interesting ideas 01:31:27.503 --> 01:31:30.863 to push that to harder and harder tasks for instance attentional tasks 01:31:30.863 --> 01:31:35.963 where maybe the target odor is changing over time things like this okay but 01:31:35.963 --> 01:31:41.403 now compared compared to let's say a standard sensor like like a like a cmos 01:31:41.403 --> 01:31:46.323 based sensor so what what's your performance if i if i go classify of our coffee 01:31:46.323 --> 01:31:48.163 with a standard off-the-shelf sensor, 01:31:48.383 --> 01:31:50.983 and now this one with its artificial mucus layer. 01:31:51.857 --> 01:31:56.177 Yeah, I mean, you know, we've got in our first paper just to show that when 01:31:56.177 --> 01:31:59.217 you do it with and without, you get an improved performance. 01:32:00.317 --> 01:32:05.477 And we have this other theoretical study to show that in certain situations 01:32:05.477 --> 01:32:10.597 you do about two orders of magnitude better in terms of the error. 01:32:12.057 --> 01:32:15.137 And that's the sort of level of what we've quantified so far. 01:32:15.357 --> 01:32:19.397 So in terms of practical… So that still has to happen. But how do you fabricate this mucus? 01:32:20.137 --> 01:32:22.997 Uh yeah so in fact it's quite a challenge 01:32:22.997 --> 01:32:26.957 because a lot of these stationary phase materials are quite poisonous so 01:32:26.957 --> 01:32:31.917 they had a big challenge in the lab of how to get these stably inside a microchannel 01:32:31.917 --> 01:32:36.777 array um the warwick people made a very beautiful there effectively that used 01:32:36.777 --> 01:32:41.717 a more advanced micro lithography method of a 3d printer effectively to grow 01:32:41.717 --> 01:32:44.317 grow a microchannel as you can 01:32:44.337 --> 01:32:47.697 imagine we have a great interest in say also growing 01:32:47.697 --> 01:32:50.577 a rat nose structure so maybe of course maybe there are 01:32:50.577 --> 01:32:54.397 well it's almost certain actually that there are uh very 01:32:54.397 --> 01:32:57.697 beautiful aspects of the elaborate structure of 01:32:57.697 --> 01:33:00.757 the nasal turbinates of a rat nose which 01:33:00.757 --> 01:33:06.557 give even more tricks in there for for instance how it controls turbulence and 01:33:06.557 --> 01:33:10.597 laminar flow and all sorts of other stuff that i haven't even talked about right 01:33:10.597 --> 01:33:15.057 that can be stuff in the future perhaps but right now Now you just basically 01:33:15.057 --> 01:33:18.897 grow in a 3D printer a very long microchannel. 01:33:19.557 --> 01:33:25.937 And you introduce this material under high pressure to get it to uniformly pass 01:33:25.937 --> 01:33:28.377 along this microchannel to deposit as a thin layer. 01:33:28.577 --> 01:33:31.877 Have you considered also to use mucus from biological systems? 01:33:31.877 --> 01:33:35.337 We haven't yet, but of course this is a fascinating idea, right? 01:33:35.397 --> 01:33:38.297 Because they already have things like OBPs in there. Exactly right. 01:33:38.437 --> 01:33:41.817 And so on, and other things, and odor-degrading enzymes. times but 01:33:41.817 --> 01:33:45.377 of course they're also not necessarily particularly stable right so these 01:33:45.377 --> 01:33:48.237 are other things to take into account but yeah it's 01:33:48.237 --> 01:33:51.157 a fascinating area of all sorts of different materials that you can potentially 01:33:51.157 --> 01:33:55.497 put in there probably anything you can use pretty much any chemical material 01:33:55.497 --> 01:33:58.917 with a liquid aspect will have some sort of selective properties for groups 01:33:58.917 --> 01:34:03.817 of compounds and they will change the sort of array receptor properties in different 01:34:03.817 --> 01:34:07.677 ways right so there's all sorts of possibilities to have sort of parallel versions 01:34:07.677 --> 01:34:10.357 of these noses right It doesn't have to be one nose. 01:34:10.477 --> 01:34:15.397 You can have arrays of these noses with, I don't know, peanut butter coatings, 01:34:15.417 --> 01:34:20.837 snot coatings, various different stationary phases from GC columns, 01:34:21.017 --> 01:34:23.917 God knows what, right? And they're all giving you different selective answers. 01:34:24.097 --> 01:34:27.757 You can think of it as different windows on the chemical world. Right, exactly. 01:34:28.037 --> 01:34:31.877 Sort of seeing a unique little portion of a chemical world. 01:34:32.137 --> 01:34:37.377 Right. But now, do you believe, I mean, you foresee also the robots of the future sure. 01:34:39.276 --> 01:34:43.076 Largely will be equipped with with artificial noses yeah 01:34:43.076 --> 01:34:46.016 it's a good question there's been a lot 01:34:46.016 --> 01:34:49.036 of efforts actually i mean not just on robots so for at 01:34:49.036 --> 01:34:52.996 least 20 years people have talked about things like will there be smell sensors 01:34:52.996 --> 01:34:55.996 on your mobile phone and uh in fact 01:34:55.996 --> 01:34:59.096 at some point there was a lot of talk about that particularly in japan people were 01:34:59.096 --> 01:35:03.176 very keen to know you know how is my uh breath smelling 01:35:03.176 --> 01:35:06.696 today you know maybe there's a need for having 01:35:06.696 --> 01:35:14.556 these kind of very cheap sensors in a in a in a mobile phone um there's as you 01:35:14.556 --> 01:35:19.656 know very well there's there's uh all sorts of domains for this in terms of 01:35:19.656 --> 01:35:23.416 robot robots for environmental sensing and so on and um, 01:35:24.216 --> 01:35:27.096 uh there are all sorts of aspects in terms of human 01:35:27.096 --> 01:35:30.016 emotion that are obviously important in in 01:35:30.016 --> 01:35:33.336 odors so if you're going to ever going to 01:35:33.336 --> 01:35:36.216 ever make a robot that understands humans probably 01:35:36.216 --> 01:35:40.836 needs to understand odor as well because there's some sort of estimate of something 01:35:40.836 --> 01:35:49.256 like 95 or 99 percent of our um uh olfactory uh consciousness is subconscious 01:35:49.256 --> 01:35:55.796 so it has all of this sort of background biasing on our state of mind at a particular point in time, 01:35:56.376 --> 01:36:00.456 probably a robot would need to know about this right if it needed to understand humans. 01:36:02.156 --> 01:36:05.576 Um and there's all sorts of applications in flavor 01:36:05.576 --> 01:36:08.316 industries and so on so there are 01:36:08.316 --> 01:36:11.136 clearly a lot of applications but i i think 01:36:11.136 --> 01:36:16.916 there it's always going to be a niche okay so to to finish up i mean so you're 01:36:16.916 --> 01:36:21.276 you're you have the one foot in the biology of affection the other one in in 01:36:21.276 --> 01:36:25.696 technology of affection and also you use the technology to understand the biology 01:36:25.696 --> 01:36:30.836 and the biology to advance the technology so it's a very unique position you're in so in 01:36:30.956 --> 01:36:34.136 the study of the brain and in our in our attempts to synthesize 01:36:34.136 --> 01:36:36.876 brains what should be tim's law that we have 01:36:36.876 --> 01:36:40.316 to follow uh well 01:36:40.316 --> 01:36:43.516 it may be a bit obvious but uh to construct is 01:36:43.516 --> 01:36:46.756 to prove so i think um looking at 01:36:46.756 --> 01:36:50.596 the biology uh and making making sense of 01:36:50.596 --> 01:36:53.656 those principles by having them in reality in 01:36:53.656 --> 01:36:56.816 front of you operating in concrete ways 01:36:56.816 --> 01:36:59.736 uh is is 01:36:59.736 --> 01:37:05.136 crucial because uh i know too many models of too many different phenomena that 01:37:05.136 --> 01:37:10.696 uh you never know how robust they are going to be in a realistic setting or 01:37:10.696 --> 01:37:18.776 whatever so that would always be my my advice okay and the another thing So five years from now, 01:37:18.796 --> 01:37:21.516 I'm going to come to Leicester to visit you in your lab, 01:37:21.616 --> 01:37:26.956 and I'm going to confront you with the hypothesis you're going to generate today. 01:37:27.296 --> 01:37:32.336 So the question is really, what's the one hypothesis you really want to commit yourself to today? 01:37:32.676 --> 01:37:36.556 And five years from now, I can come and see if you actually were able to validate 01:37:36.556 --> 01:37:37.696 it and what the outcome was. 01:37:41.612 --> 01:37:48.732 Yeah i i think an under an underrepresented aspect uh for the future is to prove that um, 01:37:49.432 --> 01:37:52.492 attentional processing in olfaction is 01:37:52.492 --> 01:37:55.552 very important so at the moment we don't really 01:37:55.552 --> 01:37:59.632 know much about attentional processing we don't we don't really know about it 01:37:59.632 --> 01:38:04.472 much in biology and we know even less about it in say machines or deploying 01:38:04.472 --> 01:38:10.392 this and so the prediction i would I would hope that in five years I could take 01:38:10.392 --> 01:38:13.372 you to my lab and prove to you, 01:38:13.412 --> 01:38:20.312 demonstrate that I can make olfactory machines that attend to different parts 01:38:20.312 --> 01:38:24.172 of this beautifully complex molecular world that we have. 01:38:24.172 --> 01:38:29.952 And that depending upon operational demands at particular points in time, 01:38:30.092 --> 01:38:38.852 it would be able to give you, say, unique reports or windows on this very complex universe. 01:38:39.132 --> 01:38:45.172 And I would like to be able to prove to you that it's not just making it up, 01:38:45.212 --> 01:38:49.212 but it's looking at different facets of a very complex signal. 01:38:49.212 --> 01:38:56.512 And that hopefully this could be done in a responsive sort of way to a particular task. 01:38:57.072 --> 01:39:00.312 And that doesn't really exist at the moment. Exactly. All right, 01:39:00.332 --> 01:39:02.612 Tim Pearce, thank you very much for this conversation. Thanks a lot. 01:39:04.400 --> 01:39:09.840 Music. 01:39:09.752 --> 01:39:15.232 The CSN Podcast was produced by the Convergent Science Network of Biometrics 01:39:15.232 --> 01:39:22.052 and Biohybrid Systems, a project funded by the European 7th Research Framework Programme. 01:39:23.072 --> 01:39:28.492 For more interviews, recorded lectures or upcoming conferences in the field 01:39:28.492 --> 01:39:34.792 of biometrics and biohybrid systems, go to csnnetwork.eu. 01:39:35.920 --> 01:39:43.600 Music.