1 00:00:00,000 --> 00:00:07,800 Welcome to the Cannabis Data Science Meetup Group. 2 00:00:07,800 --> 00:00:09,000 Happy to have you all here. 3 00:00:09,000 --> 00:00:11,560 So long story short, my name is Keegan. 4 00:00:11,560 --> 00:00:17,120 I started a company, Cantlydx, spent a little bit of time in the cannabis space, preaching 5 00:00:17,120 --> 00:00:18,120 data. 6 00:00:18,120 --> 00:00:20,360 My background is in statistics. 7 00:00:20,360 --> 00:00:24,400 And so I was thinking, well, this is something that I know a bit about. 8 00:00:24,400 --> 00:00:33,280 So why not get together with some other knowledgeable people, data scientists, cannabis operators, 9 00:00:33,280 --> 00:00:39,280 regulators, you name it, and see if we can use statistics to help advance the cannabis 10 00:00:39,280 --> 00:00:40,320 industry. 11 00:00:40,320 --> 00:00:42,920 So that's where I'm coming from. 12 00:00:42,920 --> 00:00:47,920 This is basically a roundtable for us all to get together, talk about what's on our 13 00:00:47,920 --> 00:00:54,320 minds, what's pressing, what we want to accomplish, how we think data science can help us. 14 00:00:54,320 --> 00:00:57,360 Any pressing research questions we may have. 15 00:00:57,360 --> 00:01:02,680 Then I'm always happy to share my research and ideas with you. 16 00:01:02,680 --> 00:01:09,320 Always open to your feedback, because your feedback is how improvement happens. 17 00:01:09,320 --> 00:01:13,680 And then if you have any cool ideas or research endeavors, and I'm always happy to hear about 18 00:01:13,680 --> 00:01:17,600 those too, just so we can keep bouncing ideas off of each other. 19 00:01:17,600 --> 00:01:19,080 So that's enough about me. 20 00:01:19,080 --> 00:01:24,240 I'll share with you some cool statistics here today, cannabis related. 21 00:01:24,240 --> 00:01:27,880 Before we get to that, just want to give everybody a chance just to say, hey, a bunch of new 22 00:01:27,880 --> 00:01:29,320 faces today. 23 00:01:29,320 --> 00:01:35,720 So you don't have to, but if you want, spend anywhere from 30 seconds to a couple minutes 24 00:01:35,720 --> 00:01:42,680 talking about maybe your background or what you hope to get out of applying data science 25 00:01:42,680 --> 00:01:43,840 to cannabis. 26 00:01:43,840 --> 00:01:49,240 So Nick, good to meet you. 27 00:01:49,240 --> 00:01:50,240 What are you interested in? 28 00:01:50,240 --> 00:01:54,520 What do you hope to learn here in 2023? 29 00:01:54,520 --> 00:01:59,880 Well, I was involved in cannabis growing for about four years. 30 00:01:59,880 --> 00:02:08,920 And basically I'm interested, how the business will go further in 2023, because we all know 31 00:02:08,920 --> 00:02:18,920 what is now happening and people are struggling and how data analytics can help people achieve 32 00:02:18,920 --> 00:02:25,240 their goals, I guess. 33 00:02:25,240 --> 00:02:26,240 I love it. 34 00:02:26,240 --> 00:02:29,440 And I love your optimism, right? 35 00:02:29,440 --> 00:02:35,120 And then just mute me while I talk, just there was a little bit of feedback there. 36 00:02:35,120 --> 00:02:39,240 So long story short is I love your optimism because yes, you pointed out there's a problem 37 00:02:39,240 --> 00:02:42,280 at hand, but there's ways to fix it. 38 00:02:42,280 --> 00:02:43,280 And how do we start? 39 00:02:43,280 --> 00:02:45,880 Let's just start by any solution. 40 00:02:45,880 --> 00:02:48,520 Let's just start thinking of ideas. 41 00:02:48,520 --> 00:02:52,960 Cannabis doesn't have to be the best idea in the world, but it can be a starting point. 42 00:02:52,960 --> 00:02:56,400 So as you said, there's some companies that are struggling. 43 00:02:56,400 --> 00:03:02,320 We looked at the survival rate of cannabis companies and it varies by the type of business, 44 00:03:02,320 --> 00:03:09,160 say if you're a retailer, a processor, a cultivator, but on average, cannabis companies weren't 45 00:03:09,160 --> 00:03:17,760 having a lifespan much longer than a year on average, which is not an encouraging sight. 46 00:03:17,760 --> 00:03:23,000 It's not comparable to say the restaurant industry, but people are out there trying 47 00:03:23,000 --> 00:03:25,640 to start long-term companies. 48 00:03:25,640 --> 00:03:27,560 They want to have successful cultivations. 49 00:03:27,560 --> 00:03:35,680 They want to have successful retail establishments, successful processing. 50 00:03:35,680 --> 00:03:36,680 So exactly. 51 00:03:36,680 --> 00:03:39,920 So how can we alleviate some of the burdens? 52 00:03:39,920 --> 00:03:42,360 Well, maybe statistics can help. 53 00:03:42,360 --> 00:03:46,920 So we'll talk about how we can get back to the basics today. 54 00:03:46,920 --> 00:03:48,200 Real cool things coming up. 55 00:03:48,200 --> 00:03:54,480 So just once again, give everybody a chance to say, hey, Noah, good to meet you. 56 00:03:54,480 --> 00:03:57,400 We'd love to hear about your area that you're interested in. 57 00:03:57,400 --> 00:03:58,400 Yeah. 58 00:03:58,400 --> 00:03:59,400 Hi. 59 00:03:59,400 --> 00:04:00,400 Good to meet everyone. 60 00:04:00,400 --> 00:04:01,400 Yeah. 61 00:04:01,400 --> 00:04:04,720 I just found this meetup a few days ago. 62 00:04:04,720 --> 00:04:06,040 I thought it'd be really interesting. 63 00:04:06,040 --> 00:04:12,800 I just graduated from a degree in computer science with a data science emphasis. 64 00:04:12,800 --> 00:04:14,840 Still was kind of going around looking for work. 65 00:04:14,840 --> 00:04:17,720 I just relocated to Portland area. 66 00:04:17,720 --> 00:04:18,720 And I don't know. 67 00:04:18,720 --> 00:04:21,120 I'm just I'm really passionate about cannabis. 68 00:04:21,120 --> 00:04:24,520 And I've spent some time growing my own cannabis. 69 00:04:24,520 --> 00:04:30,560 And I'm really interested in some of the applications of using reinforcement learning to improve 70 00:04:30,560 --> 00:04:31,560 growth and that kind of thing. 71 00:04:31,560 --> 00:04:38,080 And I'm also interested in using large language models to kind of work out some of the effects 72 00:04:38,080 --> 00:04:40,400 of cannabis on different people. 73 00:04:40,400 --> 00:04:44,820 I think there's some really interesting application there and some holes that I think can see 74 00:04:44,820 --> 00:04:49,240 consumers could benefit from kind of understanding other people's experience with like certain 75 00:04:49,240 --> 00:04:50,720 strains and that kind of thing. 76 00:04:50,720 --> 00:04:55,640 But yeah, I'm mainly just here to just kind of hear what I didn't even know there was 77 00:04:55,640 --> 00:04:58,640 like a field of analytics in cannabis. 78 00:04:58,640 --> 00:05:05,440 So it's exciting to see that there are people working here and I'm just here to learn. 79 00:05:05,440 --> 00:05:10,200 So I'll give you a little taste of everything. 80 00:05:10,200 --> 00:05:14,400 So of course, I've got a lot of enthusiasm for this area. 81 00:05:14,400 --> 00:05:19,920 So I think there's a lot of value to be added by applying data science to cannabis. 82 00:05:19,920 --> 00:05:26,120 And so I started can let it just to show the other side. 83 00:05:26,120 --> 00:05:30,720 Data science is essentially a bleeding edge field. 84 00:05:30,720 --> 00:05:31,720 So it's really new. 85 00:05:31,720 --> 00:05:37,280 And so it can be difficult to convince people of the value. 86 00:05:37,280 --> 00:05:46,160 That's part of the one of the difficulties in a lot of what analytics companies do is 87 00:05:46,160 --> 00:05:52,640 talk to people and try to say try to essentially convince them that the analytics are in fact 88 00:05:52,640 --> 00:05:53,640 valuable. 89 00:05:53,640 --> 00:05:58,980 I guess don't want to to, you know, miss sell anything. 90 00:05:58,980 --> 00:06:01,080 But I think they are there. 91 00:06:01,080 --> 00:06:05,200 I think it's just asymmetric information. 92 00:06:05,200 --> 00:06:09,720 So to any who there is is a little bit of an uphill battle. 93 00:06:09,720 --> 00:06:12,760 But in my opinion, there's a lot of value to be added. 94 00:06:12,760 --> 00:06:17,720 The subjects you're talking about are exactly what people are interested in. 95 00:06:17,720 --> 00:06:26,000 Of course, on the retail side, especially since the medical industry is with how do 96 00:06:26,000 --> 00:06:33,520 you say they're they can't wait to dip their toes in the to dip their toes in. 97 00:06:33,520 --> 00:06:35,560 And of course, why? 98 00:06:35,560 --> 00:06:39,600 Because they're interested in potential medical effects. 99 00:06:39,600 --> 00:06:42,940 This is a lot of the talk on the town on the retail side. 100 00:06:42,940 --> 00:06:47,640 Another thing you raised is, well, what about the the home growers? 101 00:06:47,640 --> 00:06:52,800 And this was something that was neglected for a long time. 102 00:06:52,800 --> 00:06:58,020 The reason being in a lot of states, for example, Washington State, Washington State is still 103 00:06:58,020 --> 00:07:05,680 hesitant to permit home grow as Massachusetts has because they're worried about diversion. 104 00:07:05,680 --> 00:07:11,160 I don't know how legitimate of a claim that is, but I was thinking that analytics could 105 00:07:11,160 --> 00:07:15,400 be something that home growers could find quite useful. 106 00:07:15,400 --> 00:07:23,080 And I've encouraged meetup members in the past that this is something that I think people 107 00:07:23,080 --> 00:07:24,080 have dabbled in. 108 00:07:24,080 --> 00:07:27,600 But once again, you can always have a fresh take on it. 109 00:07:27,600 --> 00:07:30,280 Especially home grow analytics. 110 00:07:30,280 --> 00:07:33,640 And I don't know if it's just the material that I've searched for. 111 00:07:33,640 --> 00:07:40,240 But I see a lot of people doing, you know, hydroponic setups. 112 00:07:40,240 --> 00:07:46,760 And I wouldn't be surprised if at home hydroponic farming could be a thing in the future. 113 00:07:46,760 --> 00:07:51,240 But once again, bleeding edge, but you're well positioned. 114 00:07:51,240 --> 00:07:53,560 So I'm kind of rambling. 115 00:07:53,560 --> 00:07:57,720 If you have any more thoughts, feel free to chime in. 116 00:07:57,720 --> 00:08:01,840 And then as I continue to think about all your cool ideas, Noah. 117 00:08:01,840 --> 00:08:04,280 Michelle, welcome to the group. 118 00:08:04,280 --> 00:08:05,280 Happy to have you. 119 00:08:05,280 --> 00:08:08,160 We'd love to hear about what you're particularly interested in. 120 00:08:08,160 --> 00:08:11,480 That way we can make sure to touch on your interests. 121 00:08:11,480 --> 00:08:12,480 Sure. 122 00:08:12,480 --> 00:08:13,480 Hi. 123 00:08:13,480 --> 00:08:17,040 Keeping my camera turned off because I tend to get up and move around a lot. 124 00:08:17,040 --> 00:08:22,840 But like Noah, I kind of just stumbled upon the meetup a few days ago. 125 00:08:22,840 --> 00:08:29,520 I'm a math teacher who's actually, I'm not a math teacher anymore, but I'm transitioning 126 00:08:29,520 --> 00:08:32,120 into data analysis. 127 00:08:32,120 --> 00:08:37,580 And as I've been looking for projects to do and things like that, one of the things on 128 00:08:37,580 --> 00:08:40,720 my list was the cannabis industry. 129 00:08:40,720 --> 00:08:46,160 And I have a gajillion questions from not only the business aspect, but also how it's 130 00:08:46,160 --> 00:08:51,480 affecting society, you know, benefiting, I think probably benefiting most. 131 00:08:51,480 --> 00:08:57,280 But I would like to see the data on that for, you know, just other, you know, social ailments 132 00:08:57,280 --> 00:09:00,760 and mental struggles and that sort of thing. 133 00:09:00,760 --> 00:09:01,760 Depression. 134 00:09:01,760 --> 00:09:03,880 I deal with depression. 135 00:09:03,880 --> 00:09:08,760 But most of all, cannabis has been a part of my life for as long as I can remember. 136 00:09:08,760 --> 00:09:11,320 I grew up in the 70s with a hippie dad. 137 00:09:11,320 --> 00:09:14,880 He grew it on the roof of our apartment in New York. 138 00:09:14,880 --> 00:09:18,960 And yeah, so I just thought it was a really interesting way that I could maybe develop 139 00:09:18,960 --> 00:09:23,160 a project for my portfolio as well. 140 00:09:23,160 --> 00:09:24,160 I love it, Michelle. 141 00:09:24,160 --> 00:09:27,560 And you attended on the perfect day. 142 00:09:27,560 --> 00:09:31,240 Everything that you raised are essentially exactly what we're touching on. 143 00:09:31,240 --> 00:09:33,160 So first off, math. 144 00:09:33,160 --> 00:09:40,660 Coincidentally, we actually have a fun math or statistical exercise today. 145 00:09:40,660 --> 00:09:46,760 Going back to the 70s, we're actually going to hit on some real cool history today, which 146 00:09:46,760 --> 00:09:50,840 is tangentially related to growing on rooftops. 147 00:09:50,840 --> 00:09:55,240 And we'll get more into that next week. 148 00:09:55,240 --> 00:10:01,720 And then the final thing is, and I'll get to this more, relates to the insight of the 149 00:10:01,720 --> 00:10:02,720 day. 150 00:10:02,720 --> 00:10:05,200 And I'll sort of tease that. 151 00:10:05,200 --> 00:10:11,480 There's sort of a big takeaway at the end of today. 152 00:10:11,480 --> 00:10:12,600 Statistical insight. 153 00:10:12,600 --> 00:10:14,360 So I'll tease that. 154 00:10:14,360 --> 00:10:15,400 But that's all coming up. 155 00:10:15,400 --> 00:10:17,400 So you attended on the perfect day. 156 00:10:17,400 --> 00:10:20,680 Now, Isaac, happy to see you today. 157 00:10:20,680 --> 00:10:23,380 I love that you made it to the morning. 158 00:10:23,380 --> 00:10:26,360 So hopefully this works out well for you. 159 00:10:26,360 --> 00:10:31,840 We're doing, and the final thing is we're going to be working on a lot of projects. 160 00:10:31,840 --> 00:10:36,540 So this is essentially the, we're going into the third year of the Cannabis Data Science 161 00:10:36,540 --> 00:10:37,840 Meetup Group. 162 00:10:37,840 --> 00:10:40,320 The first year was a lot of data wrangling. 163 00:10:40,320 --> 00:10:44,880 So we were just trying to find what are all the cannabis data sources? 164 00:10:44,880 --> 00:10:46,480 What are all the data points? 165 00:10:46,480 --> 00:10:48,480 How can we get them? 166 00:10:48,480 --> 00:10:50,880 That was a lot of what the first year was. 167 00:10:50,880 --> 00:10:57,080 The second year, we laid out all of the statistical tools that we'll be using. 168 00:10:57,080 --> 00:10:59,680 So we started to lay out model after model. 169 00:10:59,680 --> 00:11:04,080 We have about 30 solid models that we can pull from. 170 00:11:04,080 --> 00:11:08,440 And then we began to see how we could apply those in the cannabis space. 171 00:11:08,440 --> 00:11:12,000 Now this year, we've got the data. 172 00:11:12,000 --> 00:11:19,320 We've got our statistical tools, and now it's time to finally embark on interesting projects. 173 00:11:19,320 --> 00:11:23,000 And so that's why it's all hands on deck. 174 00:11:23,000 --> 00:11:30,520 So if any of you, Michelle, Noah, Nag, Isaac, are looking for projects, feel free to get 175 00:11:30,520 --> 00:11:38,160 in touch because I'm actually starting to work with this group, cannabis data, and we're 176 00:11:38,160 --> 00:11:41,520 just wrangling data in statistics. 177 00:11:41,520 --> 00:11:44,760 And there's a lot of issues, open source project. 178 00:11:44,760 --> 00:11:49,120 So feel free to see if there's any way that that could shoot your fancy and you could 179 00:11:49,120 --> 00:11:52,960 get some value out of that. 180 00:11:52,960 --> 00:12:03,120 So anywho, before I continue rambling on, Isaac, any cool projects that you're interested 181 00:12:03,120 --> 00:12:05,680 in tackling in this coming year? 182 00:12:05,680 --> 00:12:15,960 Well, one thing that I'm interested in is lab fraud. 183 00:12:15,960 --> 00:12:23,040 So last year, well, end of last year, I tried to use Benford to do some detection. 184 00:12:23,040 --> 00:12:32,560 And also, last week, I plotted out the microbiome results from Washington and I noticed a strange 185 00:12:32,560 --> 00:12:39,120 decrease of detections around the state regulatory limit. 186 00:12:39,120 --> 00:12:46,300 So that seems pretty fishy to me and I perhaps dig into that a little bit more. 187 00:12:46,300 --> 00:12:51,240 And also, I found a pretty good test. 188 00:12:51,240 --> 00:12:54,520 It's called McCray density test. 189 00:12:54,520 --> 00:13:08,200 And it's like, well, so one method for detection or for just in general, measurement around 190 00:13:08,200 --> 00:13:13,360 one intervention is called discontinuity design. 191 00:13:13,360 --> 00:13:20,280 And it's that there is a, you are basically working around a ball of data around one single 192 00:13:20,280 --> 00:13:23,440 point and you're trying to see if the means are different. 193 00:13:23,440 --> 00:13:29,280 And this method called McCray density test allows you to do that on a histogram, say 194 00:13:29,280 --> 00:13:30,840 of a normal distribution. 195 00:13:30,840 --> 00:13:38,640 And you can use that to measure if there is a gap of difference of whatever mean or standard 196 00:13:38,640 --> 00:13:45,380 deviation of a normal distribution to detect if it's really off. 197 00:13:45,380 --> 00:13:48,500 So that seems pretty promising method. 198 00:13:48,500 --> 00:13:51,360 So yeah. 199 00:13:51,360 --> 00:13:55,920 I love it, Isaac, and I love that you bring this topic up today because once again, it 200 00:13:55,920 --> 00:14:03,960 is the exact topic at hand, lab results and statistical distributions. 201 00:14:03,960 --> 00:14:08,000 And if you've got time, I'll be attending. 202 00:14:08,000 --> 00:14:15,120 There is a quality control meeting that's open to the public in Washington state tomorrow. 203 00:14:15,120 --> 00:14:19,320 And these meetings, there's often a lot of people in the industry there, right? 204 00:14:19,320 --> 00:14:24,560 There'll be representatives from the labs and the licensees, but very few times do you 205 00:14:24,560 --> 00:14:26,040 see people from the public. 206 00:14:26,040 --> 00:14:30,120 So if you were able to attend, that would be a whole new perspective. 207 00:14:30,120 --> 00:14:35,840 And as you mentioned, and this was sort of the topic last week, there's all the, of course, 208 00:14:35,840 --> 00:14:39,840 there's a lot of talk about, you know, THC and CBD. 209 00:14:39,840 --> 00:14:42,840 There's a lot of natural variation there too. 210 00:14:42,840 --> 00:14:49,840 And like you said, of course we do want to identify any fraud. 211 00:14:49,840 --> 00:14:53,320 However do we only want to look at those analytes? 212 00:14:53,320 --> 00:14:58,720 And as we were pointing out last week, you know, moisture and water activity were looking 213 00:14:58,720 --> 00:15:00,480 a little odd. 214 00:15:00,480 --> 00:15:03,840 And I kind of messed up my Benford's analysis, I realized. 215 00:15:03,840 --> 00:15:06,680 So I'm happy that you repeated it. 216 00:15:06,680 --> 00:15:10,320 And so the moisture and water activity could use another look. 217 00:15:10,320 --> 00:15:16,360 And then as you said, the microbes and the mycotoxins, especially the mycotoxins could 218 00:15:16,360 --> 00:15:19,600 potentially use another look. 219 00:15:19,600 --> 00:15:22,920 Because they may have changed the method in Washington state. 220 00:15:22,920 --> 00:15:27,440 I need to follow it a bit more closely, but as you, I'm sure you know, there's multiple 221 00:15:27,440 --> 00:15:32,160 ways that you can screen for mycotoxins. 222 00:15:32,160 --> 00:15:39,680 From my understanding, you can use what's called plating, which I think takes maybe 223 00:15:39,680 --> 00:15:47,840 24 to 48 hours and maybe isn't as precise as one may hope. 224 00:15:47,840 --> 00:15:54,000 And then I think people may be moving more towards testing for mycotoxins on a mass spec, 225 00:15:54,000 --> 00:16:03,000 which is faster because I think you can run maybe a mass spec run maybe in under an hour. 226 00:16:03,000 --> 00:16:07,540 But it's more expensive because you have to have a mass spec. 227 00:16:07,540 --> 00:16:11,320 You have to have somebody that can run a mass spec and then you can have somebody that can 228 00:16:11,320 --> 00:16:13,320 analyze the results. 229 00:16:13,320 --> 00:16:15,960 So it's a bit more complicated. 230 00:16:15,960 --> 00:16:20,160 So it's interesting that you noted that. 231 00:16:20,160 --> 00:16:26,560 And so it's worth investigating a bit more because the labs and the licensees, they have 232 00:16:26,560 --> 00:16:33,120 their priorities about what they want to talk about, which will probably be lab testing 233 00:16:33,120 --> 00:16:34,600 is expensive. 234 00:16:34,600 --> 00:16:38,240 However, consumers have their concerns. 235 00:16:38,240 --> 00:16:44,960 And so if we see, say, mycotoxins is being measured oddly, that's a good point to bring 236 00:16:44,960 --> 00:16:45,960 up. 237 00:16:45,960 --> 00:16:48,680 So I like it. 238 00:16:48,680 --> 00:16:58,440 So just to go ahead and dive in to the topic at hand. 239 00:16:58,440 --> 00:17:00,120 Well I'll go ahead and show you. 240 00:17:00,120 --> 00:17:02,520 I got in a hold of this book. 241 00:17:02,520 --> 00:17:04,920 I think you can get it for fairly cheap. 242 00:17:04,920 --> 00:17:10,080 This is Sense Amelia Tips compiled by Tom Alexander. 243 00:17:10,080 --> 00:17:14,480 I got a used copy, I think, for under 10 bucks or so. 244 00:17:14,480 --> 00:17:16,680 So you may be able to find a copy. 245 00:17:16,680 --> 00:17:20,400 This is a surprisingly good find. 246 00:17:20,400 --> 00:17:26,460 So I was expecting just your typical news journal. 247 00:17:26,460 --> 00:17:29,720 But there's actually a good bit of substance to this. 248 00:17:29,720 --> 00:17:32,840 This was published in 1988. 249 00:17:32,840 --> 00:17:39,160 So that's, I don't know when in 1988, but that's going on 35 years ago. 250 00:17:39,160 --> 00:17:42,060 And that was when it was compiled. 251 00:17:42,060 --> 00:17:44,720 So those could have been written before then. 252 00:17:44,720 --> 00:17:47,640 So this could be over 35 years old. 253 00:17:47,640 --> 00:17:51,080 And so let's see if there's any nuggets of wisdom in there. 254 00:17:51,080 --> 00:17:54,000 Sorry, could you repeat the title again? 255 00:17:54,000 --> 00:17:56,040 I couldn't see the title clearly. 256 00:17:56,040 --> 00:17:57,040 Sure. 257 00:17:57,040 --> 00:18:09,160 I'll make all these links available, but it's SINSEMILATIP. 258 00:18:09,160 --> 00:18:17,120 And so Sense Amelia is, I think it may even be a Spanish name. 259 00:18:17,120 --> 00:18:23,200 That's when they, originally, when people cultivated cannabis, you would just have one 260 00:18:23,200 --> 00:18:28,680 field, where you would just have female and male plants. 261 00:18:28,680 --> 00:18:36,120 Well, the male plants would pollinate the female plants, and you'd get flowers filled 262 00:18:36,120 --> 00:18:37,520 with seeds. 263 00:18:37,520 --> 00:18:44,840 And so that's why the whole sublime song, pick out the seeds from stands. 264 00:18:44,840 --> 00:18:51,760 So people used to have to pick out the seeds because they just grew it all together. 265 00:18:51,760 --> 00:19:02,120 And then in the 70s or so, definitely by the 80s, people would realize if you are solely 266 00:19:02,120 --> 00:19:09,600 interested in the flower bud, then if you pick out the male flowers, then it's more 267 00:19:09,600 --> 00:19:10,600 efficient. 268 00:19:10,600 --> 00:19:19,600 You're using water and nutrients and light and space only for the productive female plants, 269 00:19:19,600 --> 00:19:22,960 and they don't get pollinated by the males. 270 00:19:22,960 --> 00:19:31,040 So cannabis is an odd plant, so there still may be an odd seed or two, but there's fewer 271 00:19:31,040 --> 00:19:32,040 seeds. 272 00:19:32,040 --> 00:19:39,000 So that's Sense Amelia, a technique that became more and more popular. 273 00:19:39,000 --> 00:19:43,760 So anywho, that's the name of the article. 274 00:19:43,760 --> 00:19:51,200 And this journal was put together by this interesting character, Tom Alexander, who 275 00:19:51,200 --> 00:19:54,240 was growing cannabis in Oregon. 276 00:19:54,240 --> 00:20:05,200 And I think they confiscated his crop, and I think he was facing some criminal penalties. 277 00:20:05,200 --> 00:20:10,840 I think the case ended up getting dropped for one reason or the other, I do believe. 278 00:20:10,840 --> 00:20:14,160 But then after that, he just went into publishing. 279 00:20:14,160 --> 00:20:20,680 And so he just talked with a bunch of cultivators, growers, and compiled a lot of techniques. 280 00:20:20,680 --> 00:20:23,800 So it's a real interesting look into the past here. 281 00:20:23,800 --> 00:20:25,040 But enough of me rambling on. 282 00:20:25,040 --> 00:20:32,720 Let me actually show you what's valuable about this. 283 00:20:32,720 --> 00:20:38,800 Okay, too cool. 284 00:20:38,800 --> 00:20:42,840 So this is a question that's come up multiple times. 285 00:20:42,840 --> 00:20:46,560 And Isaac, you may have brought this up at one point. 286 00:20:46,560 --> 00:20:57,580 And I think it's pretty well known, but I wanted to go back to the source and really 287 00:20:57,580 --> 00:21:02,520 try to figure out some of the early research that's been done on this. 288 00:21:02,520 --> 00:21:10,400 So okay, so we see a lot of variation in THC and CBD. 289 00:21:10,400 --> 00:21:16,200 And so the question is, well, is that because of environmental factors? 290 00:21:16,200 --> 00:21:19,080 And so that would be, are some cultivators better than others? 291 00:21:19,080 --> 00:21:21,160 Do some people have brighter lights? 292 00:21:21,160 --> 00:21:24,160 Do some people have more nutrients? 293 00:21:24,160 --> 00:21:28,040 Do some people have less disease? 294 00:21:28,040 --> 00:21:29,320 So on and so forth. 295 00:21:29,320 --> 00:21:32,360 So those would be environmental factors. 296 00:21:32,360 --> 00:21:36,980 And then, of course, people are interested in about breeding. 297 00:21:36,980 --> 00:21:40,120 And so here we get to the math. 298 00:21:40,120 --> 00:21:47,760 So I'll blast you with some of the math and statistics, and then we'll get to the takeaways. 299 00:21:47,760 --> 00:21:54,320 So sigma, this is a typo. 300 00:21:54,320 --> 00:21:57,520 This should be a sigma squared. 301 00:21:57,520 --> 00:22:02,840 This is a measure of variance. 302 00:22:02,840 --> 00:22:06,760 So it's a measure of central tendency. 303 00:22:06,760 --> 00:22:09,760 So we know about the mean. 304 00:22:09,760 --> 00:22:15,280 So if you've got a bunch of observations, you can calculate the average. 305 00:22:15,280 --> 00:22:21,720 But then you can also calculate how much do each of these vary from each other. 306 00:22:21,720 --> 00:22:23,960 So that's your variance. 307 00:22:23,960 --> 00:22:32,800 And so if we're interested in about, say, THC, well, presumably, there's some variance 308 00:22:32,800 --> 00:22:35,880 in THC that comes from genetics. 309 00:22:35,880 --> 00:22:46,040 So if you have a good seed, maybe that leads a certain amount to your variance in THC, 310 00:22:46,040 --> 00:22:49,340 as well as the environment. 311 00:22:49,340 --> 00:22:57,560 So if you have dim lights, one would expect your THC maybe a little lower. 312 00:22:57,560 --> 00:23:06,960 Likewise, if you've got poor quality soil, poor nutrients, poor airflow, odd temperatures, 313 00:23:06,960 --> 00:23:13,380 all of these things are your environment and could lead to different THC concentrations. 314 00:23:13,380 --> 00:23:23,160 And as we've seen, we've observed people selecting different breeds of cannabis to go into quite 315 00:23:23,160 --> 00:23:24,720 different directions. 316 00:23:24,720 --> 00:23:28,160 You see people breeding for high THC. 317 00:23:28,160 --> 00:23:34,400 And then you also see people breeding for low THC, people trying to grow the CBD hemp, 318 00:23:34,400 --> 00:23:38,920 who are trying to stay under the 0.3% THC limit. 319 00:23:38,920 --> 00:23:44,840 OK, so you've got genetic variance and environmental variance. 320 00:23:44,840 --> 00:23:54,760 And so we're going to define heterotability of a trait as the amount of total variance. 321 00:23:54,760 --> 00:24:04,520 So the total variance we see is the genetic variance plus the environmental variance. 322 00:24:04,520 --> 00:24:13,840 So we want to know what percentage of all the variance that we see comes from genetics. 323 00:24:13,840 --> 00:24:19,040 Is it a small amount or is it a large amount? 324 00:24:19,040 --> 00:24:22,280 This is really the crux. 325 00:24:22,280 --> 00:24:30,200 Because if genetics doesn't matter, then you just want to get your environment as good 326 00:24:30,200 --> 00:24:31,680 as possible. 327 00:24:31,680 --> 00:24:40,120 Likewise, if genetic variance, if that represents a large proportion of heritability, then you'd 328 00:24:40,120 --> 00:24:46,720 be more interested in focusing on breeding and less focused on environment. 329 00:24:46,720 --> 00:24:52,400 So can we disentangle these two? 330 00:24:52,400 --> 00:25:00,760 Also, just pointing this out, this was written by Chief 7 Turtles. 331 00:25:00,760 --> 00:25:08,840 And the turtle is a symbol for cantlytics because we're sort of the turtle over the 332 00:25:08,840 --> 00:25:09,840 hare. 333 00:25:09,840 --> 00:25:12,920 And so I just thought this was just a funny coincidence. 334 00:25:12,920 --> 00:25:16,040 Anywho, what are some of the key points? 335 00:25:16,040 --> 00:25:18,560 And then I'll get to the actual data. 336 00:25:18,560 --> 00:25:25,520 So some populations have more genetic variation than others. 337 00:25:25,520 --> 00:25:34,960 Likewise, certain traits are more susceptible to genetic variation than others. 338 00:25:34,960 --> 00:25:39,680 Of course, environmental variation depends on your conditions. 339 00:25:39,680 --> 00:25:47,280 The example given in Sinsemilia tips was your environmental variation will be different 340 00:25:47,280 --> 00:25:51,160 outdoors than it will be in a greenhouse. 341 00:25:51,160 --> 00:25:57,280 I need to think a little bit harder on that. 342 00:25:57,280 --> 00:26:03,480 But the conditions matter, and also the trait matters as well. 343 00:26:03,480 --> 00:26:09,040 So those two are pretty straightforward and logical. 344 00:26:09,040 --> 00:26:10,400 Those make sense. 345 00:26:10,400 --> 00:26:14,240 Okay, well, why am I droning on about this? 346 00:26:14,240 --> 00:26:21,640 Well, heritability is actually a difficult topic to study. 347 00:26:21,640 --> 00:26:25,560 And it's normally infeasible. 348 00:26:25,560 --> 00:26:33,960 What's interesting about cannabis is it's often grown with clones. 349 00:26:33,960 --> 00:26:37,640 So people will clone cannabis. 350 00:26:37,640 --> 00:26:40,120 It's pretty typical. 351 00:26:40,120 --> 00:26:48,160 Well, when you make a clone, there's going to be no genetic variation. 352 00:26:48,160 --> 00:26:58,400 All you'll have in the clone is environmental variation. 353 00:26:58,400 --> 00:27:12,060 So the idea is you can see the total variation in a crop of plants, and you can subtract 354 00:27:12,060 --> 00:27:17,120 away the variation from the clones. 355 00:27:17,120 --> 00:27:22,800 And what you're left with is the genetic variation. 356 00:27:22,800 --> 00:27:30,160 And then once you're left with the genetic variation and you know the environmental variation, 357 00:27:30,160 --> 00:27:33,560 you can calculate the heritability. 358 00:27:33,560 --> 00:27:42,100 So the long story short is heritability can range from zero, something doesn't inherit, 359 00:27:42,100 --> 00:27:43,360 to one. 360 00:27:43,360 --> 00:27:47,000 It's going to... 361 00:27:47,000 --> 00:27:52,540 The genetics is going to explain 100% of the variability in that trait. 362 00:27:52,540 --> 00:28:01,600 So we're interested in is the heritability of THC, is that closer to zero or is it closer 363 00:28:01,600 --> 00:28:07,080 to one? 364 00:28:07,080 --> 00:28:09,580 So that is sort of the question at hand. 365 00:28:09,580 --> 00:28:14,840 And if this is getting a little abstract, then don't worry because we're going to get 366 00:28:14,840 --> 00:28:16,480 our hands on the data. 367 00:28:16,480 --> 00:28:25,920 And I should have committed this to the repository beforehand, but I'll commit this right afterwards. 368 00:28:25,920 --> 00:28:37,920 So in this book, Since Amelia Tips, there's a table where they have two different plots, 369 00:28:37,920 --> 00:28:48,440 plot one and plot two, and in each plot, they have some plants that were grown from seed 370 00:28:48,440 --> 00:28:57,240 and some plants that were grown from clone, and they were measuring the THC concentration 371 00:28:57,240 --> 00:28:59,520 of these plants. 372 00:28:59,520 --> 00:29:06,280 And so this is, I was saying last week that I think cultivators have been keeping track 373 00:29:06,280 --> 00:29:09,320 of THC concentration for a long time. 374 00:29:09,320 --> 00:29:16,800 Well, it turns out they've been keeping track of THC concentration perhaps for more than 375 00:29:16,800 --> 00:29:21,800 35 years, which is wild to think about. 376 00:29:21,800 --> 00:29:23,800 So let's actually... 377 00:29:23,800 --> 00:29:24,800 Oops. 378 00:29:24,800 --> 00:29:25,800 Oops. 379 00:29:25,800 --> 00:29:30,200 Let's not give everything away. 380 00:29:30,200 --> 00:29:40,240 So first things first, just going to read this data in, and just to show you, this is 381 00:29:40,240 --> 00:29:42,360 all the data, right? 382 00:29:42,360 --> 00:29:52,280 We were just looking at in Excel, just going to print it out to the console, and then we'll 383 00:29:52,280 --> 00:29:58,280 essentially repeat the analysis done in Since Amelia Tips. 384 00:29:58,280 --> 00:30:12,080 I wasn't sure about if I could just take a picture of their table or not. 385 00:30:12,080 --> 00:30:19,400 I decided not to, but maybe it would have been a helpful benchmark. 386 00:30:19,400 --> 00:30:29,320 So basically the idea, well, an idea of the scientific process is reproducibility. 387 00:30:29,320 --> 00:30:33,640 And for statistics, this becomes really interesting. 388 00:30:33,640 --> 00:30:42,960 It's actually non-trivial to repeat someone's statistics, and it's often a good exercise. 389 00:30:42,960 --> 00:30:48,560 So if somebody publishes a paper, to go back through, get a hold of the data, and try to 390 00:30:48,560 --> 00:30:56,200 follow the steps they took to reproduce the same statistics, because you'll have to follow 391 00:30:56,200 --> 00:30:59,560 through all the assumptions, and it can be difficult. 392 00:30:59,560 --> 00:31:01,640 So anywho, we've got this data here. 393 00:31:01,640 --> 00:31:03,240 Well, let's start looking at it. 394 00:31:03,240 --> 00:31:05,680 So we've got seeds and clones. 395 00:31:05,680 --> 00:31:13,840 So if we just look at the seeds, so these were cannabis grown in two different plots 396 00:31:13,840 --> 00:31:15,420 by seed. 397 00:31:15,420 --> 00:31:28,160 And so as you can see, back in 1988, you've got an average THC of around 8%. 398 00:31:28,160 --> 00:31:35,680 So the average THC has more than doubled in the past 35 years. 399 00:31:35,680 --> 00:31:38,640 So that's an observation. 400 00:31:38,640 --> 00:31:43,080 Next we can look at the distribution of clones. 401 00:31:43,080 --> 00:31:47,160 And so these are clones in two different plots. 402 00:31:47,160 --> 00:31:57,120 And this would be environmental variance between these two different plots. 403 00:31:57,120 --> 00:32:03,200 There's no genetic variation between clones. 404 00:32:03,200 --> 00:32:12,280 So any variation in the distribution, so any variation in variation, will be explainable 405 00:32:12,280 --> 00:32:14,600 from environment. 406 00:32:14,600 --> 00:32:26,520 So if you did the same study that Chief Seven Turtles conducted, then you would see, oh, 407 00:32:26,520 --> 00:32:35,120 plot one is the better plot for growing clones than plot two. 408 00:32:35,120 --> 00:32:37,440 And that's purely environmental. 409 00:32:37,440 --> 00:32:39,360 Cool. 410 00:32:39,360 --> 00:32:50,080 So now if you plot the seeds versus the clones, it becomes quite interesting. 411 00:32:50,080 --> 00:32:58,120 So here you see the distribution of plants grown from seeds, distribution grown by clones. 412 00:32:58,120 --> 00:33:08,520 What's readily apparent, the variance is much smaller for plants grown from clone than from 413 00:33:08,520 --> 00:33:12,120 plants grown from seed. 414 00:33:12,120 --> 00:33:18,480 The mean is lower, but that may just be the clone that was selected. 415 00:33:18,480 --> 00:33:26,720 So for example, if you just selected a random plant from this distribution as your clone, 416 00:33:26,720 --> 00:33:33,720 maybe they selected a plant from a seed from this end of the distribution. 417 00:33:33,720 --> 00:33:35,520 Cool. 418 00:33:35,520 --> 00:33:45,080 So now we can actually calculate heritability the same way that Tom Alexander did. 419 00:33:45,080 --> 00:33:49,520 So the way they did, they actually use standard deviation. 420 00:33:49,520 --> 00:33:55,920 And so I'll show you this with standard deviation, and then I'll go back to variance. 421 00:33:55,920 --> 00:34:06,320 But basically, if you look at the, they also calculated variance grouped by plot. 422 00:34:06,320 --> 00:34:13,720 And so this is what's so interesting about reproducing statistics, because all they have 423 00:34:13,720 --> 00:34:19,320 is a table with the means and variances in their calculations. 424 00:34:19,320 --> 00:34:27,320 And so it takes reproducing it to realize that, OK, when they calculated the standard 425 00:34:27,320 --> 00:34:35,320 deviation, they did this by grouping it by plot, and then they took the average. 426 00:34:35,320 --> 00:34:45,720 So basically, they took the standard deviation of these two plots, which is odd that it's 427 00:34:45,720 --> 00:34:48,320 the same, but it is. 428 00:34:48,320 --> 00:34:55,320 It could be just the nature of numbers. 429 00:34:55,320 --> 00:34:59,320 Anywho, you've got the standard deviation of clones. 430 00:34:59,320 --> 00:35:04,320 And as you can see, the standard deviation of seeds is higher. 431 00:35:04,320 --> 00:35:15,320 So basically, the argument is all of the variation from clones is environmental variation. 432 00:35:15,320 --> 00:35:25,320 So that's how you get the environmental variation of 0.79% THC. 433 00:35:25,320 --> 00:35:30,320 And then they say, OK, well, what's the phenotypic variation? 434 00:35:30,320 --> 00:35:36,320 And so the phenotypic variation, well, that's environment plus genetics. 435 00:35:36,320 --> 00:35:43,320 And so they say, oh, we'll just get that by taking the standard deviation of the seeds. 436 00:35:43,320 --> 00:35:50,320 And we'll call that our phenotypic variation, because that's going to be all of the environment 437 00:35:50,320 --> 00:35:59,320 variation that the seeds experience plus the genetic variation of the seeds. 438 00:35:59,320 --> 00:36:00,320 Cool. 439 00:36:00,320 --> 00:36:07,320 And so then they say, oh, well, we can just calculate the genetic variation by subtracting 440 00:36:07,320 --> 00:36:13,320 the environmental variation from the phenotypic variation. 441 00:36:13,320 --> 00:36:14,320 Cool. 442 00:36:14,320 --> 00:36:22,320 So now they're saying the genetic variation is around 1.27% THC. 443 00:36:22,320 --> 00:36:26,320 And so you can now get a measure of heritability. 444 00:36:26,320 --> 00:36:28,320 Drum roll. 445 00:36:28,320 --> 00:36:37,320 So you're interested in, on a scale of 0 to 1, how heritable is THC? 446 00:36:37,320 --> 00:36:42,320 Dun-dun-dun-dun. 447 00:36:42,320 --> 00:36:50,320 Chief, seven turtles calculated the heritability of cannabis, or THC in cannabis, 448 00:36:50,320 --> 00:36:57,320 to be approximately 0.62, or 62%, which is quite high. 449 00:36:57,320 --> 00:37:06,320 So this would indicate that THC is, I guess you could argue what strongly is, 450 00:37:06,320 --> 00:37:10,320 but in my opinion, is strongly heritable. 451 00:37:10,320 --> 00:37:22,320 And so this would explain, okay, people figured out in 1988 that you can control 452 00:37:22,320 --> 00:37:27,320 variation in THC through genetics. 453 00:37:27,320 --> 00:37:33,320 So, for example, going back to our chart here, 454 00:37:33,320 --> 00:37:45,320 this now explains how people have been able to grow both CBD low THC cannabis 455 00:37:45,320 --> 00:37:52,320 and high THC cannabis because it's strongly heritable. 456 00:37:52,320 --> 00:37:59,320 So, for example, this clone would actually be a good selection if you wanted to move 457 00:37:59,320 --> 00:38:02,320 in the low THC direction. 458 00:38:02,320 --> 00:38:11,320 And so the idea would be you'd grow a bunch of seeds and you would pick one from the low end 459 00:38:11,320 --> 00:38:17,320 of the distribution and use that one for breeding and just repeat the process over 460 00:38:17,320 --> 00:38:24,320 and over again and try to push this distribution towards zero. 461 00:38:24,320 --> 00:38:33,320 Conversely, if you're interested in high THC, you would pick, you would breed the plants 462 00:38:33,320 --> 00:38:42,320 that tested at high concentrations and just keep doing selective breeding. 463 00:38:42,320 --> 00:38:53,320 Quick note, technically the formula for heritability is variance. 464 00:38:53,320 --> 00:39:04,320 And for I don't know why, for whatever reason, they chose to use standard deviation. 465 00:39:04,320 --> 00:39:10,320 I don't know if this is just because statistics has kind of evolved or people's understanding 466 00:39:10,320 --> 00:39:13,320 of statistics have kind of evolved in the past 35 years. 467 00:39:13,320 --> 00:39:19,320 I mean, statistics as a whole is a relatively new study in the grand scheme of things, 468 00:39:19,320 --> 00:39:29,320 rigorously, maybe a past 150 years or so, or not rigorously, that's maybe when it began 469 00:39:29,320 --> 00:39:36,320 and then maybe more rigorously, I don't know where, I'll quit conjecturing. 470 00:39:36,320 --> 00:39:46,320 The long story short is if we did say use variance, then we would get a higher measure 471 00:39:46,320 --> 00:39:47,320 of heritability. 472 00:39:47,320 --> 00:39:56,320 So if we used variance, we would measure heritability at almost 85%, which seems too high to me. 473 00:39:56,320 --> 00:40:02,320 But I also want to point out a couple shortcomings real quick. 474 00:40:02,320 --> 00:40:06,320 One, this is a statistic. 475 00:40:06,320 --> 00:40:17,320 So our estimation of heritability will become better as sample size increases. 476 00:40:17,320 --> 00:40:28,320 So here we have a sample size of 20 plants grown by seed and 20 plants grown by clone. 477 00:40:28,320 --> 00:40:33,320 So we could potentially use a bigger sample size. 478 00:40:33,320 --> 00:40:41,320 Also, what the author points out, measurement matters. 479 00:40:41,320 --> 00:40:49,320 And they say that you can view measurement error as an environmental variation, essentially, 480 00:40:49,320 --> 00:40:59,320 but they stress that you really want to measure as accurately and as consistently as possible. 481 00:40:59,320 --> 00:41:07,320 And then also, if you really do want to try to disentangle environmental variation from genetic variation, 482 00:41:07,320 --> 00:41:13,320 then you want to try to have as uniform an environment as possible. 483 00:41:13,320 --> 00:41:27,320 This was just a short, quick exercise just demonstrating how you can measure heritability of traits in cannabis. 484 00:41:27,320 --> 00:41:32,320 And remember, this will vary by population. 485 00:41:32,320 --> 00:41:37,320 So that means it may vary by strain population. 486 00:41:37,320 --> 00:41:42,320 I mean, it's not out of the realm. You'd have to do some investigation. 487 00:41:42,320 --> 00:41:49,320 But potentially, different strains may have different rates of heritability. 488 00:41:49,320 --> 00:41:54,320 Different traits may have different rates of heritability. 489 00:41:54,320 --> 00:42:01,320 So, for example, CBD may have a different rate of heritability than THC. 490 00:42:01,320 --> 00:42:04,320 And there's other factors people select for. 491 00:42:04,320 --> 00:42:16,320 So right out of the gate, they mentioned people are selecting for yield, for height, how many days it takes the flower. 492 00:42:16,320 --> 00:42:19,320 There's many factors that people are selecting for. 493 00:42:19,320 --> 00:42:24,320 So you'll have to look at heritability for each one. 494 00:42:24,320 --> 00:42:37,320 But this lays the framework for how you can apply statistics to be a better breeder. 495 00:42:37,320 --> 00:42:43,320 So this could apply to home growers or large scale cultivating. 496 00:42:43,320 --> 00:42:51,320 So I've been rambling long enough. I'll let you all ask any questions. 497 00:42:51,320 --> 00:42:56,320 And a teaser for next week. This is what I was going to do this week. 498 00:42:56,320 --> 00:42:59,320 But once again, lots of material. 499 00:42:59,320 --> 00:43:04,320 And so we wanted to take this quick tangent on heritability. 500 00:43:04,320 --> 00:43:16,320 For next week, I'll see if you can't find the author of this quote or the person who this can be attributed. 501 00:43:16,320 --> 00:43:27,320 But in relation to statistics, the idea is you don't want to try too hard. 502 00:43:27,320 --> 00:43:31,320 If you do, you may get spurious causation. 503 00:43:31,320 --> 00:43:40,320 And so in fact, obviously, I started a company, Candelitics. 504 00:43:40,320 --> 00:43:45,320 So we're interested in the cannabis space. 505 00:43:45,320 --> 00:43:53,320 As you're researching cannabis, it would be easy and you would have to be careful not to let your biases take over. 506 00:43:53,320 --> 00:44:05,320 So, for example, if you really, really did want to see a medicinal effect from CBD, that may bias your research. 507 00:44:05,320 --> 00:44:15,320 And so if there is just a readily apparent effect from CBD, then there may be something there that's worth looking at more. 508 00:44:15,320 --> 00:44:25,320 But the idea is if there's not just a real readily available signal or noise, you may want to be cautious. 509 00:44:25,320 --> 00:44:27,320 It doesn't mean don't proceed. 510 00:44:27,320 --> 00:44:39,320 You just may want to be cautious about biases, spurious causation, overfitting your model. 511 00:44:39,320 --> 00:44:49,320 And so, for example, basically, the lesson from today, how could we apply that lesson today? 512 00:44:49,320 --> 00:44:56,320 Well, if you really wanted to disentangle the environmental variation, 513 00:44:56,320 --> 00:45:07,320 there may be something going on between plot one and plot two, but it may be a stretch. 514 00:45:07,320 --> 00:45:18,320 However, it does seem readily apparent that there's differences in variance between seeds and clones. 515 00:45:18,320 --> 00:45:30,320 You've got a small sample size, 40 total observations, and already you can start to see a difference between the two. 516 00:45:30,320 --> 00:45:34,320 So that may warrant further investigation. 517 00:45:34,320 --> 00:45:40,320 So that's just sort of a word of caution as you're using statistics. 518 00:45:40,320 --> 00:45:50,320 And so if you're interested in this, then try to find the author behind this, because we'll continue to, or I'll at least continue to talk about this. 519 00:45:50,320 --> 00:46:06,320 And if you're not interested, then please let me know what sparks your interest as we continue to search for the genetic histories behind some of these cannabis treatments. 520 00:46:06,320 --> 00:46:15,320 Anywho, that's what I had to share with you all today, my latest journey. 521 00:46:15,320 --> 00:46:21,320 So my main project is compiling cannabis data from Washington state. 522 00:46:21,320 --> 00:46:37,320 And so, Michelle, this may be of direct relevance to you because you could start to do real interesting statistics as far as population level statistics goes there. 523 00:46:37,320 --> 00:46:40,320 So lots of lots more cool data coming down the pipelines. 524 00:46:40,320 --> 00:46:46,320 But real quick, does anyone have any thoughts, comments, questions from the material on heritability? 525 00:46:46,320 --> 00:47:07,320 Yeah, I can go first. I mean, I'm just wondering if there's anything we can do with more modern data because that and also the environmental variable, I mean, itself could be different. 526 00:47:07,320 --> 00:47:19,320 So I'm just thinking how the experiment can be improved with modern data to get a better measurement. 527 00:47:19,320 --> 00:47:27,320 Maybe we can even get a constant that's applied to most cannabis plants as one number, say 0.7. 528 00:47:27,320 --> 00:47:32,320 And then that number as an anchor point will allow us to do more analysis. 529 00:47:32,320 --> 00:47:38,320 But yeah, I agree 100 percent, Isaac. 530 00:47:38,320 --> 00:47:46,320 In fact, I would encourage you to and I'm going to be setting up a lot more on heritability now that we've figured this out. 531 00:47:46,320 --> 00:47:50,320 I think there are more sophisticated ways to estimate heritability. 532 00:47:50,320 --> 00:48:01,320 I think what you're talking about is more of a regression based approach where you'll want to control for more factors and, of course, increase your sample size. 533 00:48:01,320 --> 00:48:06,320 So I think the regression based approach, maybe the modern day approach. 534 00:48:06,320 --> 00:48:11,320 Remember, this was this was research that was done 35 years ago. 535 00:48:11,320 --> 00:48:15,320 So I think there is more sophisticated readily. 536 00:48:15,320 --> 00:48:21,320 And I mean, for all we know, they may have been calculating these means and variances with pen and paper. 537 00:48:21,320 --> 00:48:24,320 So we can do a lot better job today. 538 00:48:24,320 --> 00:48:37,320 So, for example, just loading the data into Python, right, I'm able to to do all these groupings and calculate means and variances and standard deviations really quickly. 539 00:48:37,320 --> 00:48:43,320 So it was really quick for me to to figure out, oh, they use standard deviation and not variance. 540 00:48:43,320 --> 00:48:47,320 So we can work quicker, work smarter. 541 00:48:47,320 --> 00:48:51,320 I think it's just just a starting point. 542 00:48:51,320 --> 00:49:09,320 And then the other part you raised about the environment, I think that does need to be taken into consideration because the way they went about identifying the environmental variation and then subtracting it from the variation of the seeds, 543 00:49:09,320 --> 00:49:12,320 it seems a little ad hoc. 544 00:49:12,320 --> 00:49:20,320 So this may be an ad hoc method for for calculating or estimating heritability. 545 00:49:20,320 --> 00:49:31,320 So this may be a starting point just to just get an idea of, OK, it is heritability closer to zero or maybe closer to one. 546 00:49:31,320 --> 00:49:37,320 So this got us an idea that, OK, there may be a significant amount of heritability there. 547 00:49:37,320 --> 00:49:42,320 And so, as you said, I think a more sophisticated analysis is needed. 548 00:49:42,320 --> 00:49:46,320 People are going to want to see a lot more controls of environment. 549 00:49:46,320 --> 00:49:53,320 That's a big criticism that people give to cannabis research. 550 00:49:53,320 --> 00:49:56,320 It's inevitable. We've talked about this before. 551 00:49:56,320 --> 00:50:00,320 You can't control for every factor under the sun. 552 00:50:00,320 --> 00:50:08,320 Right. If you controlled for everything and you did it in Massachusetts, someone will say, well, why didn't you do that same study in Washington state? 553 00:50:08,320 --> 00:50:12,320 You know, the air pressure is different or so, you know, so on and so forth. 554 00:50:12,320 --> 00:50:18,320 So there's always going to be something. But as you said, just the more things you control for the better. 555 00:50:18,320 --> 00:50:22,320 That gets us to sort of the lesson of the day. 556 00:50:22,320 --> 00:50:28,320 The more things you control for should get you an accurate, more accurate and consistent measure. 557 00:50:28,320 --> 00:50:38,320 But hopefully it's still relatively apparent, even if you say make a mistake or two or don't control for everything. 558 00:50:38,320 --> 00:50:41,320 I love it, Isaac. And so I'll continue thinking about this. 559 00:50:41,320 --> 00:50:51,320 The main thing is just working on the Washington state data was going to almost had a hold of some Oregon data that's still in the coming down the pipeline. 560 00:50:51,320 --> 00:50:55,320 And that may come for you as well as Massachusetts and Florida data. 561 00:50:55,320 --> 00:50:58,320 So hopefully those things come. 562 00:50:58,320 --> 00:51:03,320 OK, what about now? Awesome. Awesome. I can hear you. Love it. 563 00:51:03,320 --> 00:51:18,320 I'm talking about from a grover's perspective about these levels. So basically what I know that I was growing the same train outdoor greenhouse and indoor. 564 00:51:18,320 --> 00:51:23,320 And it was giving very different levels of THC. 565 00:51:23,320 --> 00:51:33,320 So the environment variance is very important in growing. Also, the seeds were growing faster than the clones. 566 00:51:33,320 --> 00:51:41,320 I know why they develop faster. They just have looks like they have the better genetics than the clones. 567 00:51:41,320 --> 00:51:46,320 And that's what I observed. OK, so there's a lot going on here. 568 00:51:46,320 --> 00:51:52,320 One thing I want to hit on that I don't know if a lot of people mention, but this has come across my research. 569 00:51:52,320 --> 00:51:59,320 From my understanding, seeds have tap roots, whereas clones do not. 570 00:51:59,320 --> 00:52:05,320 And to me, that makes it seem like that would have a big impact. 571 00:52:05,320 --> 00:52:09,320 It seems to me like the tap root would be important. 572 00:52:09,320 --> 00:52:13,320 So that maybe that may explain why it would grow faster. 573 00:52:13,320 --> 00:52:22,320 Technically, the clones are genetically identical to where they were cut from. 574 00:52:22,320 --> 00:52:32,320 So if they did grow differently, one could maybe chalk that up to maybe structural. 575 00:52:32,320 --> 00:52:37,320 They don't have a tap root. Also, there may be environment there. 576 00:52:37,320 --> 00:52:44,320 Also, it depends on a mother plant where you cut the clones. 577 00:52:44,320 --> 00:52:57,320 If you cut more than 30 percent of the plant, they say, and I also noticed that the clones can become hemophrodites. 578 00:52:57,320 --> 00:53:08,320 So they can just change. But also, we didn't talk about the nutrients that you put into the soil. 579 00:53:08,320 --> 00:53:15,320 So that's also very important for the plant to develop THC. 580 00:53:15,320 --> 00:53:28,320 Also, timing when you put certain things, is it like phosphorus or calcium or things like that? 581 00:53:28,320 --> 00:53:30,320 That's also very important. 582 00:53:30,320 --> 00:53:35,320 A hundred percent. This is the tricky part. 583 00:53:35,320 --> 00:53:40,320 Because there's variation in THC and obviously there's some genetics. 584 00:53:40,320 --> 00:53:47,320 That's how people have been able to breed over time. But even if, say, the genetics is 60 percent, 585 00:53:47,320 --> 00:53:56,320 that's still 40 percent or give or take more or less of variation that's explained from your environment. 586 00:53:56,320 --> 00:54:03,320 And that will be things like whether it was indoor or outdoor, because those are two different environments. 587 00:54:03,320 --> 00:54:10,320 How much nutrients the plant gets, that's an environmental factor. 588 00:54:10,320 --> 00:54:14,320 So those will all affect the environmental variation. 589 00:54:14,320 --> 00:54:20,320 And this matters when you're cultivating, right, because you want to grow top notch flower. 590 00:54:20,320 --> 00:54:33,320 The idea is, and this may be just a simple fact, but essentially, it may be obvious, 591 00:54:33,320 --> 00:54:43,320 you can't just select by clone to get better and better cannabis, because it's going to be genetically identical. 592 00:54:43,320 --> 00:54:50,320 Basically, the point was being in order to basically increase the THC levels over time, 593 00:54:50,320 --> 00:54:55,320 you would have to approach that genetically. 594 00:54:55,320 --> 00:55:03,320 It wouldn't be possible to just select clones that test higher and higher percentages. 595 00:55:03,320 --> 00:55:13,320 And then this is sort of the interesting thing is the seed may have genetics to produce high THC, 596 00:55:13,320 --> 00:55:21,320 but say, because of the environment, it may just have a lot of environmental variation. 597 00:55:21,320 --> 00:55:25,320 So it doesn't look like a high THC plant. 598 00:55:25,320 --> 00:55:32,320 And this is my argument about how California may have such prized genetics, 599 00:55:32,320 --> 00:55:37,320 because I think they're growing outside by seed. 600 00:55:37,320 --> 00:55:44,320 So these seeds in the plants, they may not grow into the best plants ever, right? 601 00:55:44,320 --> 00:55:51,320 They may be getting eaten by bugs and they're subject to all the elements. 602 00:55:51,320 --> 00:55:56,320 But people are able to then select the better and better variety. 603 00:55:56,320 --> 00:56:05,320 So if you then say, yeah, you took that clone from those plants and you went and grew it in a warehouse, 604 00:56:05,320 --> 00:56:09,320 you may be able to get higher THC on average. 605 00:56:09,320 --> 00:56:18,320 But over time, the people outdoor selecting and selecting and selecting will be able to increase the average. 606 00:56:18,320 --> 00:56:20,320 And then Noah, you had a question? 607 00:56:20,320 --> 00:56:26,320 Yeah, it was kind of related to that. Have there been any studies yet on like epigenetic factors? 608 00:56:26,320 --> 00:56:32,320 I'm curious if that's something that is even in effect here or I don't know. 609 00:56:32,320 --> 00:56:36,320 So this is bleeding edge. 610 00:56:36,320 --> 00:56:39,320 I can only report back hearsay. 611 00:56:39,320 --> 00:56:49,320 So as you may have known, LinkedIn is a good place for cannabis, people in the cannabis industry to socialize. 612 00:56:49,320 --> 00:57:00,320 And I've actually heard talk on the town that, yes, potentially cannabis may be especially susceptible to epigenetic change. 613 00:57:00,320 --> 00:57:04,320 But this is this is really new frontier. 614 00:57:04,320 --> 00:57:10,320 And I don't know how much I want to stray into it because I'm a super novice. 615 00:57:10,320 --> 00:57:13,320 I wouldn't even consider myself a biologist or a chemist. 616 00:57:13,320 --> 00:57:22,320 So actual biologists and chemists would probably frown upon upon me. So I kind of want to stay as humble as possible. 617 00:57:22,320 --> 00:57:26,320 But Michelle, you had a thought or question? 618 00:57:26,320 --> 00:57:34,320 Yeah, I was just wondering whether or not given that so many states are legalizing cannabis now, whether or not commercial growers, 619 00:57:34,320 --> 00:57:41,320 I know that they're tracking their THC levels and their CBD levels and levels. 620 00:57:41,320 --> 00:57:46,320 Are they required to report that data so that there's some sort of database that can be accessed? 621 00:57:46,320 --> 00:57:52,320 Yes. We have that data for you, the entire population in Washington state. 622 00:57:52,320 --> 00:58:00,320 And so that's what's real cool. We may even be able to parse out if things came from clone or seed. 623 00:58:00,320 --> 00:58:07,320 I don't know how well that's recorded in the traceability system, but we do have plant data. 624 00:58:07,320 --> 00:58:14,320 And all they were keeping track of were THC levels. And we have those. 625 00:58:14,320 --> 00:58:22,320 So we've got THC levels, CBD levels, and we may even have the propagation. 626 00:58:22,320 --> 00:58:25,320 We just may not have the genetic lineage. 627 00:58:25,320 --> 00:58:33,320 So there is maybe work we can do with data that we have at this time in Washington state. 628 00:58:33,320 --> 00:58:38,320 So let's keep talking about this because I think there is work we can do, Michelle. 629 00:58:38,320 --> 00:58:41,320 No, no, what further thoughts? 630 00:58:41,320 --> 00:58:49,320 Yeah, just following up on what Michelle said, I've heard anecdotal information about some of those THC testing, 631 00:58:49,320 --> 00:58:53,320 at least the ones reported on the label, and that they can be dubious. 632 00:58:53,320 --> 00:59:01,320 How do we know that these numbers that are being reported by the growers or the cultivars are accurate 633 00:59:01,320 --> 00:59:07,320 and aren't just kind of being, you know, torturing the numbers to kind of get high THC or low THC or whatever? 634 00:59:07,320 --> 00:59:13,320 Perfectly reasonable question. And the short answer is that's measurement error. 635 00:59:13,320 --> 00:59:19,320 So this is specifically, you know, even way back, I keep pointing this out, 636 00:59:19,320 --> 00:59:26,320 but it just shocks my mind how these topics were on mind 35 years ago. 637 00:59:26,320 --> 00:59:32,320 And exactly, if you have inaccurate measures of lab results, 638 00:59:32,320 --> 00:59:38,320 you're not going to be able to parse out the genetic variation as well. 639 00:59:38,320 --> 00:59:49,320 So disentangling environmental variation and genetic variation requires accurate and consistent testing. 640 00:59:49,320 --> 00:59:56,320 So that's why it is important for Isaac. I love that you're working on this. 641 00:59:56,320 --> 01:00:02,320 And that's why, you know, it is important for labs to measure accurately and consistently. 642 01:00:02,320 --> 01:00:05,320 It helps the breeders in the long run. 643 01:00:05,320 --> 01:00:10,320 Unfortunately, we may just have to anticipate that there is some measurement error. 644 01:00:10,320 --> 01:00:15,320 What I would say is, so for example, Isaac or anyone else studying this, 645 01:00:15,320 --> 01:00:19,320 that can be a condition that you try to control for. 646 01:00:19,320 --> 01:00:25,320 So say you were going to replicate their study at scale. 647 01:00:25,320 --> 01:00:30,320 Well, maybe one of the conditions you could control for would be laboratory. 648 01:00:30,320 --> 01:00:35,320 So maybe you would just grow all these plants from seed and clone, 649 01:00:35,320 --> 01:00:39,320 and then you'd send them to three or more different laboratories. 650 01:00:39,320 --> 01:00:43,320 And then you would just use that as one of your conditions 651 01:00:43,320 --> 01:00:49,320 and try to disentangle any measurement error coming from labs. 652 01:00:49,320 --> 01:00:53,320 Try to disentangle that from your genetic variation. 653 01:00:53,320 --> 01:01:00,320 So from a cultivator point of view, the measurement error is surmountable. 654 01:01:00,320 --> 01:01:05,320 I think it's not great, but I think you can overcome it. 655 01:01:05,320 --> 01:01:10,320 And you can still do, I think you can still do heritability studies. 656 01:01:10,320 --> 01:01:15,320 It just throws a monkey wrench into things, but you can just account for that. 657 01:01:15,320 --> 01:01:19,320 But Isaac, the question at hand? 658 01:01:19,320 --> 01:01:23,320 Yeah, just to chime in my two cents on the question. 659 01:01:23,320 --> 01:01:28,320 I mean, there are many sources of this uncertainty in measurement. 660 01:01:28,320 --> 01:01:34,320 On one hand, it could definitely, it's the most standard spread of measurement. 661 01:01:34,320 --> 01:01:38,320 You measure something exactly the same thing, and you'll get a spread. 662 01:01:38,320 --> 01:01:44,320 That's one type. And another type is that there might be lab issues 663 01:01:44,320 --> 01:01:51,320 that are just their control samples or their spikes was called is having issues. 664 01:01:51,320 --> 01:01:54,320 So it's like their quality control issues. 665 01:01:54,320 --> 01:02:00,320 And that might skew their test results, maybe 10% plus or minus, 666 01:02:00,320 --> 01:02:05,320 and also their habits of actually measuring it. 667 01:02:05,320 --> 01:02:11,320 So there is just natural variation. 668 01:02:11,320 --> 01:02:15,320 And second is variation based on their reference standards. 669 01:02:15,320 --> 01:02:21,320 And the third one, it could be based on their method for calculating the area 670 01:02:21,320 --> 01:02:25,320 or the amount of say THC. 671 01:02:25,320 --> 01:02:31,320 And it's only the fourth one is that they might knowingly fudge the number. 672 01:02:31,320 --> 01:02:38,320 So there are a lot of sources and the variations can be attributed to a combination. 673 01:02:38,320 --> 01:02:44,320 But in terms of how to get a data set that you can trust, 674 01:02:44,320 --> 01:02:47,320 it will take a small leap of faith. 675 01:02:47,320 --> 01:02:52,320 That is, you just look at the data of, for example, the Washington data set. 676 01:02:52,320 --> 01:02:57,320 It consists of data from 12 labs, I believe. 677 01:02:57,320 --> 01:03:03,320 And just by looking at the distribution of their THC measurements 678 01:03:03,320 --> 01:03:09,320 and also their micro measurements, some labs does have a better or more normal, 679 01:03:09,320 --> 01:03:13,320 like naturally occurring distribution than other ones. 680 01:03:13,320 --> 01:03:19,320 So if it's a concern for you that you want to go for reliable data, 681 01:03:19,320 --> 01:03:24,320 then perhaps you can filter for the labs that have the best looking distributions. 682 01:03:24,320 --> 01:03:30,320 Good points. And this is the beauty in the art of statistics. 683 01:03:30,320 --> 01:03:39,320 And why stress the, don't read into things too hard because there is a lot of noise. 684 01:03:39,320 --> 01:03:42,320 So you don't want to lean too much into the noise. 685 01:03:42,320 --> 01:03:46,320 But I love that this is the way you're thinking, Noah, 686 01:03:46,320 --> 01:03:49,320 because this is something that's often overlooked. 687 01:03:49,320 --> 01:03:55,320 And a good statistician like yourself will hammer this home. 688 01:03:55,320 --> 01:03:57,320 What about measurement error? 689 01:03:57,320 --> 01:04:01,320 Because things like this compound, right? 690 01:04:01,320 --> 01:04:07,320 Because your statistical analysis assumes maybe some variation. 691 01:04:07,320 --> 01:04:13,320 But then if all of a sudden there's other variation that you're not taking into consideration, 692 01:04:13,320 --> 01:04:20,320 like measurement error or other imperfections in the data, missing data, imputations, 693 01:04:20,320 --> 01:04:23,320 it just adds to the uncertainty. 694 01:04:23,320 --> 01:04:28,320 And I'll just leave it there that I don't know if you all know, 695 01:04:28,320 --> 01:04:34,320 but from my understanding, statistics is the study of uncertainty. 696 01:04:34,320 --> 01:04:36,320 And so that's what we're doing. 697 01:04:36,320 --> 01:04:39,320 This is an uncertain topic. 698 01:04:39,320 --> 01:04:44,320 And we're just trying to study it and see if we can't glean any knowledge out of it. 699 01:04:44,320 --> 01:04:46,320 Let's continue on this. 700 01:04:46,320 --> 01:04:49,320 I may want to go ahead and conclude this here. 701 01:04:49,320 --> 01:04:52,320 I just want to be respectful of everybody's time. 702 01:04:52,320 --> 01:04:55,320 And we've covered a lot of ground today. 703 01:04:55,320 --> 01:05:04,320 And for next week, I'll continue teasing some real fun things out of cannabis history 704 01:05:04,320 --> 01:05:10,320 and try to tie them back to the modern day with modern data sets, 705 01:05:10,320 --> 01:05:13,320 for example, this data set out of Washington State. 706 01:05:13,320 --> 01:05:15,320 So lots of cool things coming. 707 01:05:15,320 --> 01:05:20,320 And then also, if any of you have projects that you want to embark on, 708 01:05:20,320 --> 01:05:24,320 feel free to get in touch and let's keep the conversation going throughout the week. 709 01:05:24,320 --> 01:05:30,320 Because as I said, we're at the cutting edge, the bleeding edge of the cannabis area of research. 710 01:05:30,320 --> 01:05:32,320 And there's a lot to uncover. 711 01:05:32,320 --> 01:05:35,320 So I think there's exciting things to come. 712 01:05:35,320 --> 01:05:36,320 Thank you all for coming. 713 01:05:36,320 --> 01:05:41,320 Thank you for bringing your eyes, your ears, your brilliant minds. 714 01:05:41,320 --> 01:05:45,320 Your attention helps advance cannabis research, cannabis science. 715 01:05:45,320 --> 01:05:51,320 And so I think if we can help move things forward, even if it's only one molecule at a time, 716 01:05:51,320 --> 01:06:03,320 then let's keep at it.