How to Build Precision Medicine Solutions that Scale
Chris Gregg PhD speaking at the University of Calgary to the Precision Medicine Program
December 2022: Chris Gregg, Storyline CTO and co-founder speaking at the University of Calgary to clinicians and others about precision medicine and how new technologies and A.I. will allow for low-cost precision care at massive scale.
Transcript: WEBVTT
1
00:00:04.310 --> 00:00:05.630
Room: I'm really excited.
2
00:00:08.100 --> 00:00:16.269
Room: Okay, yeah. Really excited to introduce Dr. Chris Gray. Rox Craig's coming to us from the University of Utah.
3
00:00:29.790 --> 00:00:30.910
Room: It's time
4
00:00:32.790 --> 00:00:40.300
Room: highlight. The things that are, you know, really important to everyone here in Calgary, which is that he is a graduate from the University of Calgary.
5
00:00:40.410 --> 00:00:41.960
Room: So it
6
00:00:44.490 --> 00:00:46.730
Room: Thank you for the support.
7
00:00:46.970 --> 00:00:54.790
Room: Yeah. So one of our our make it good. And he, he's from here went on to Harvard
8
00:00:54.850 --> 00:01:01.099
Room: involved in developing these really important pioneering techniques. Rna seek, and other
9
00:01:01.300 --> 00:01:05.270
Room: tools and and technologies that many of you are are learning about
10
00:01:05.510 --> 00:01:11.629
Room: and also, you know, want to notes a lot of involvement. Lots of entrepreneurial spirits.
11
00:01:11.790 --> 00:01:19.499
Room: chief scientific officer, and in several companies which I think we're going to hear about here and
12
00:01:19.520 --> 00:01:22.639
Room: other scientific advisory roles as well. And
13
00:01:23.130 --> 00:01:30.449
Room: yeah, I always believe that I should get out of the way as quickly as possible. So with that i'll i'll
14
00:01:30.770 --> 00:01:31.690
see there.
15
00:01:32.260 --> 00:01:34.380
Room: Hey, Dave! Thank you very much.
16
00:01:34.550 --> 00:01:36.420
Room: And for those of you who made
17
00:01:36.530 --> 00:01:42.210
Room: time to attend this either virtually or in person. I'm really grateful for the time you made
18
00:01:42.330 --> 00:01:54.110
Room: Dave filled me in on the program here. The the precision medicine program, and I had to say, i'm pretty passionate about it. It sounds extremely unique, and it sounds incredibly important
19
00:01:54.160 --> 00:02:12.220
Room: for the past 20 years or so. I've done basic research, and I've always dreamed of translating a lot of that work into the real world to, you know, improve things out in the real world. But it is a very difficult process. It's a unique skill set, and it's really awesome to see a program that's focused on that problem specifically.
20
00:02:13.180 --> 00:02:29.690
Room: I'm going to talk about my own personal adventures in that in closing that gap from basic research to trying to deliver something that works and operates in the real world by talking about this startup company called Storyline Health.
21
00:02:29.920 --> 00:02:38.979
Room: And please stop me, and i'll try to keep an eye on the screen for online questions as well as in person. Questions
22
00:02:39.040 --> 00:02:43.810
Room: you know, as as they come up. This is a very difficult and challenging and interesting road.
23
00:02:45.280 --> 00:03:02.820
Room: So first of all you know, we've established a mission for storyline, health and storylines. Mission is to understand human behavior and make that knowledge useful for everyone. That's kind of what binds us when we make decisions around the technologies that we're building the problems we're trying to solve and the research we're trying to build.
24
00:03:02.830 --> 00:03:21.069
Room: That's what we think about the most, and it has a lot of broad implications that can be used in a lot of different ecosystems in medicine surgery, and even in places completely that I never even thought of in the legal system and other areas. it's taken us on all kinds of adventures very, very cool.
25
00:03:21.480 --> 00:03:22.730
Room: So.
26
00:03:25.180 --> 00:03:28.409
Room: Oh, how funny! I tried this many times. No, it's not.
27
00:03:29.590 --> 00:03:31.260
Room: It's not
28
00:03:32.470 --> 00:03:33.530
Room: advancing.
29
00:03:34.170 --> 00:03:36.830
Room: Oh, let's see if we can figure out Why.
30
00:03:40.160 --> 00:03:41.400
Room: okay.
31
00:03:42.390 --> 00:03:45.600
Room: we had a problem. But now you know, it's solved the problem.
32
00:03:47.090 --> 00:03:49.990
Room: Okay, that was not the problem I was intending to talk about
33
00:03:50.120 --> 00:03:56.440
Room: we with when starting down this road there were a number of problems that I was particularly interested in.
34
00:03:56.850 --> 00:03:59.689
Room: One is that acute care solutions
35
00:03:59.810 --> 00:04:19.669
Room: come into the doctor and you get, and it's sort of this acute event, and you get treated for something that typically does not resolve your health care problem if it is a chronic illness, right? So some of the biggest illnesses that we're trying to face Alzheimer's disease, which is the most expensive problem in the United States right now. Parkinson's cancer autoimmune diseases, etc.
36
00:04:19.680 --> 00:04:28.609
Room: They involve so many different factors. The immune system, inflammation, genetics, developmental factors, mental health, social support, and more.
37
00:04:28.840 --> 00:04:30.800
Room: that a single acute visit
38
00:04:30.990 --> 00:04:33.219
Room: really isn't enough to resolve the problem.
39
00:04:35.560 --> 00:04:43.019
Room: The other thing that is a really you know, clear problem with difficulty in medicine is that we don't have maps
40
00:04:43.040 --> 00:04:44.850
Room: to navigate health
41
00:04:44.940 --> 00:04:50.320
Room: healthy aging, and then the onset of diseases at different sort of stages of our lives.
42
00:04:50.470 --> 00:04:53.970
Room: and without a map you can't really navigate very rash.
43
00:04:54.090 --> 00:04:57.070
Room: You can't detect adverse events early.
44
00:04:58.010 --> 00:05:03.169
Room: You can't diagnose and patient important subtypes as effectively as you need.
45
00:05:04.210 --> 00:05:10.630
Room: You can't predict progression or treatment responses nearly as effectively or accurately as you need to.
46
00:05:10.920 --> 00:05:15.220
Room: And all of this kind of adds up to an overall
47
00:05:15.300 --> 00:05:26.339
Room: problem where we're locked into reactive care instead of predictive care where we resolve problems, so that before they progress and become chronic illnesses that are very expensive and difficult to manage.
48
00:05:28.140 --> 00:05:34.949
Room: It's not just in the clinical care setting that we have some problems. We also have real fundamental problems in the research ecosystem.
49
00:05:35.640 --> 00:05:51.919
Room: Now, I know i'm going to pick a few fights here. But bear with me because I have a lot of experiences trying to solve problems with some of these tools. Genomics is a wonderful tool, very powerful. It's the backbone of the precision, medicine, ecosystem. But frankly.
50
00:05:52.290 --> 00:05:54.169
Room: it's not very predictive.
51
00:05:54.300 --> 00:05:56.110
Room: and it's not a parry diagnostic
52
00:05:56.590 --> 00:06:09.639
Room: and the genome sequences I have to say that you are born with is the genome sequence that you die with it Doesn't change, if you change your diet, have an infection have a stressful event in your life, or get sick.
53
00:06:10.170 --> 00:06:18.869
Room: Just because I have risk factors for depression or mental illness Doesn't mean that i'm going to get that next week
54
00:06:19.050 --> 00:06:19.840
Room: right?
55
00:06:20.090 --> 00:06:25.889
Room: and in fact, if you look at some of the best polygenic risk scores we have for many of these complex diseases.
56
00:06:26.010 --> 00:06:31.870
Room: Like major depression, they only explained less than 1% of the variance and the risk for that disease.
57
00:06:32.360 --> 00:06:34.430
Room: That's a real butler, right?
58
00:06:36.370 --> 00:06:41.929
Room: The other thing that's very challenging is that some of the diagnostic tools that we're using.
59
00:06:42.090 --> 00:06:43.859
Room: especially in behavioral health
60
00:06:44.120 --> 00:06:53.130
Room: are they were developed through kind of subjective approaches. They're not data-driven in the way that some other tests that we use in medicine are.
61
00:06:53.360 --> 00:07:00.700
Room: And so people have argued that the dsm 5 diagnostic categories. For example, are too broad. They're too simplistic.
62
00:07:00.780 --> 00:07:08.239
Room: and that that's a barrier to research to uncover mechanisms, biological mechanisms that we can target and treat different subtypes.
63
00:07:08.310 --> 00:07:09.070
Perfect
64
00:07:10.320 --> 00:07:30.059
Room: animal research. Now i'm giving another talk on Friday. That's on animal research. So this is going to look like terrible hypocrisy. But animal research at the end of the day. Well, it's very powerful for undercovering basic biological mechanisms and for understanding mammalian biology, verticality, etc.
65
00:07:30.760 --> 00:07:40.770
Room: It often does not translate to humans 95% of the drugs that we get working well in. Pre-clinical trials ultimately fail in clinical trials and don't actually become medicines.
66
00:07:41.630 --> 00:07:43.650
Room: That means that we really need to study
67
00:07:43.780 --> 00:07:44.610
people.
68
00:07:45.820 --> 00:08:00.800
Room: Medical records are the backbone of the information that we have for patients, right? These Ehrs lots of different technologies trying to make use of that more effectively. But ultimately they really weren't ever designed for precision, medicine purposes, data, science.
69
00:08:00.870 --> 00:08:04.780
Room: and they don't capture real world data outside of the clinic.
70
00:08:05.710 --> 00:08:07.699
Room: And then wet lab tests.
71
00:08:08.200 --> 00:08:16.749
Room: Now, these are very exciting tests, these molecular tests that we rely on, and the new ones that are coming. But I often call them second line tests
72
00:08:16.880 --> 00:08:27.259
Room: because they're not massively scalable. These aren't tests that you can take every day or every week or every month, and be monitoring and getting a a picture of your overall health one
73
00:08:27.970 --> 00:08:37.969
Room: the microbial, which is something that you know. I'm very excited about, but it's not likely that we're going to be mailing our poop samples in every week to get regular microbiome checkups.
74
00:08:38.340 --> 00:08:45.420
Room: So the second line test, for example, they are second line. excuse me. These wet lab tests are second line tests.
75
00:08:46.250 --> 00:09:03.759
Room: and you know, as I was thinking about some of the problems that we're trying to solve. there's many, many problems in the United States where I'm. Located right now, but there are all also major opportunities here in the Canadian healthcare system, as many as you many of you know, in this precision medicine program.
76
00:09:03.890 --> 00:09:19.210
Room: health care in Canada rakes as among the most expensive systems in the United States. But we really struggle with the wait times, and as a consequence of these wait times, you know, it is having an impact on patient quality of life.
77
00:09:19.480 --> 00:09:32.550
Room: Acute illness has become chronic illnesses. There's pain and suffering and mental health job loss, economic damage, huge opportunities for smart young people who are creating a precision, medicine technologies to help solve these problems.
78
00:09:32.630 --> 00:09:34.220
Room: improving access.
79
00:09:34.330 --> 00:09:39.299
Room: So all of this really folds into one big problem, I think.
80
00:09:39.660 --> 00:09:54.149
Room: which is that we need massively scalable precision, medicine, research and care solutions, and there are only a few things I think, that can address that need in the world, and and and i'm going to tell you
81
00:09:54.360 --> 00:10:00.420
Room: my opinion. So my opinion is that the human nervous system
82
00:10:00.700 --> 00:10:04.000
Room: offers one of the best solutions that I can see.
83
00:10:04.180 --> 00:10:08.580
Room: If we look across all of the different technologies that seem to be emerging out there in the world.
84
00:10:08.990 --> 00:10:09.990
Room: And
85
00:10:10.280 --> 00:10:18.490
Room: you know, if you kind of step back and think about it. The nervous system of our body is the ultimate precision medicine
86
00:10:18.540 --> 00:10:19.629
Room: system, right?
87
00:10:20.090 --> 00:10:24.530
Room: All of these nerves are innovating different organs and tissues of the cellular level.
88
00:10:24.560 --> 00:10:33.499
Room: They're detecting metabolic changes, endocrine changes, physiological changes, pain and damage movement. All of that information rolls up into the central nervous system.
89
00:10:33.820 --> 00:10:39.710
Room: and it influences cognition, fatigue, energy, motivational drives, homeostasis.
90
00:10:40.160 --> 00:10:42.729
Room: and and much much more right. So
91
00:10:42.820 --> 00:10:47.650
Room: our behavior, which is kind of the manifestation of all of these signals and changes
92
00:10:48.330 --> 00:10:51.019
Room: as the expression of all of this underlying biology.
93
00:10:51.620 --> 00:10:52.310
Room: So
94
00:10:52.490 --> 00:10:55.450
Room: but the problem is
95
00:10:55.510 --> 00:11:00.000
Room: that we're really not making use of that data in an objective
96
00:11:00.100 --> 00:11:02.819
Room: and kind of precision, medicine, type of way.
97
00:11:03.770 --> 00:11:18.450
Room: The way that we take information from patients right now is we've got forms that they might fill out right. How many of the pad. You know, over the past 2 weeks. How many days have you felt depressed or sad? You might say several or most days
98
00:11:18.920 --> 00:11:25.330
Room: we've got medical records. There's a face to face. Exam: right? So that takes a lot of time.
99
00:11:25.740 --> 00:11:34.680
Room: and there's an expert clinician in front of you who asks these questions and is evaluating your health, your behavior, and how you respond.
100
00:11:34.700 --> 00:11:35.360
Room: But
101
00:11:35.520 --> 00:11:40.820
Room: very little of that information actually goes into the record. It's just going into the brain of this expert
102
00:11:41.430 --> 00:11:43.270
Room: who is making a judgment call.
103
00:11:43.450 --> 00:11:47.740
Room: And there's so much information that we do not measure or capture
104
00:11:48.330 --> 00:11:51.209
Room: such a huge opportunity for precision medicine.
105
00:11:53.730 --> 00:12:01.319
Room: The problem. well, let me just say that all decisions in the medical care system start by understanding the patient.
106
00:12:01.570 --> 00:12:10.560
Room: but other symptoms. What are they experiencing? What's the personality, what social support do they have? What access to care that they have, etc., etc.
107
00:12:11.370 --> 00:12:14.040
Room: But there really is no effective.
108
00:12:14.160 --> 00:12:15.750
Room: trusted way
109
00:12:15.850 --> 00:12:31.609
Room: to capture that data and make it objective like a radiologist would do right. They take a scan. They've got extraordinary capabilities now to analyze those images and pull out patterns that are diagnostic and predictive. And we don't have those technologies for understanding patient behavior.
110
00:12:32.920 --> 00:12:41.870
Room: So there's a big opportunity here, but it's a hard problem to solve, and I'm going to tell you a little personal story about why I think it's worth solving.
111
00:12:44.110 --> 00:12:57.090
Room: so so this is me in 2,018, and I'm dressed up like Dr. Evil, and my head is bald, but it's not not because i'm a real aficionado of Dr. Evil characters.
112
00:12:57.260 --> 00:13:07.869
Room: It's because I have lost my hair due to a stage for cancer diagnosis. So it's 2,018, and I was, you know, diagnosed that too many metastatic sites in my body to count.
113
00:13:08.080 --> 00:13:13.429
Room: and my diagnosis was terminal. So you know I would go on palliative care.
114
00:13:13.500 --> 00:13:15.270
Room: And
115
00:13:15.720 --> 00:13:18.719
Room: you know we we we'll get into the some of the details.
116
00:13:18.870 --> 00:13:27.649
Room: My son has, you know, shaped his head for support, so he's mining me, and and there's my wife and my dog also very supportive. So
117
00:13:27.990 --> 00:13:34.280
Room: so we faced this extraordinary problem of what seemed like an incurable and and
118
00:13:35.600 --> 00:13:37.689
Room: you know, impossible situation.
119
00:13:38.130 --> 00:13:39.840
Room: Yeah, but because of
120
00:13:40.090 --> 00:13:50.770
Room: my role in the world as a Phd. In a researcher I had access to all kinds of extraordinary and interesting individuals, and I had published a pay for that year. That had
121
00:13:50.800 --> 00:13:54.770
Room: sort of focused on. Why, elephants don't get cancer, you think? Well.
122
00:13:54.910 --> 00:14:09.059
Room: you know it's a bit of a sideline, but elephants have huge bodies, lots of cells. They, if they got cancer at the rate that we get cancer, all elephants should have cancer, but it turns out they don't they're cancer resistant, and our study was about. Why, that happens.
123
00:14:09.900 --> 00:14:17.019
Room: this this kind of gave me a network of folks who were thinking in new ways in Cancer Field.
124
00:14:17.570 --> 00:14:34.070
Room: and so we put together a cancer and evolution meeting at the Huntsman Cancer Institute, and many of my colleagues flew in just a couple of months. Very, very kind, you know. Sometimes I get teary, I not tonight. So and and the theme of the meeting was.
125
00:14:34.080 --> 00:14:38.070
Room: How could we improve stage 4 cancer outcomes.
126
00:14:38.230 --> 00:14:40.700
Room: using the knowledge that we have today?
127
00:14:41.050 --> 00:14:50.520
Room: Right? So, not generating a new medicine, new drug targets that might translate into a solution in 20 years. But how could we actually improve outcomes for patients today?
128
00:14:51.100 --> 00:14:58.900
Room: And there were a number of big ideas and important lessons that came out of that meeting, and one of the critical insights was
129
00:14:58.970 --> 00:15:01.680
Room: that the way we are treating cancer
130
00:15:02.100 --> 00:15:04.189
Room: currently in standard of care
131
00:15:04.900 --> 00:15:08.159
Room: is arguably one of the worst ways that we could approach the problem.
132
00:15:08.410 --> 00:15:21.920
Room: And how do you? How can you say that the reason that people think that this is a poor strategy which I will go through in a minute is because there are so many lessons that have been learned about attacking pests
133
00:15:21.990 --> 00:15:29.850
Room: in managing pest control in the farming community. And through studying of species, extinction, and evolution.
134
00:15:30.020 --> 00:15:34.629
Room: we have a bunch of really interesting information that's out there in these other fields.
135
00:15:34.870 --> 00:15:36.850
Room: But we haven't brought those lessons
136
00:15:37.150 --> 00:15:38.650
Room: into the cancer world.
137
00:15:38.750 --> 00:15:40.209
Room: So what are those lessons?
138
00:15:40.700 --> 00:15:44.599
Room: A typical cancer patient like me comes into the clinic.
139
00:15:44.670 --> 00:15:47.530
Room: and they've been diagnosed with their
140
00:15:47.710 --> 00:15:51.189
Room: tumor, and they're put on their first line therapy.
141
00:15:51.680 --> 00:16:03.779
Room: And if this is the tumor marker, so their tumor burden goes down. If they get a good response, and maybe they even cross into this Goldie zone fully like zone of any. No evidence of disease.
142
00:16:04.510 --> 00:16:06.069
Room: Then what happens
143
00:16:06.190 --> 00:16:07.649
Room: is,
144
00:16:07.870 --> 00:16:16.389
Room: the the standard of care is to stay on that medicine until the disease returns, and that's called progression.
145
00:16:16.890 --> 00:16:24.619
Room: So in the United States you actually cannot switch to a new medicine. Your insurance company won't pay for it
146
00:16:24.690 --> 00:16:27.440
Room: until you've shown evidence of progression.
147
00:16:27.610 --> 00:16:29.589
Room: and then you're allowed to do a new medicine.
148
00:16:30.060 --> 00:16:36.960
Room: So now you go into the next drug, and maybe you get a good response. The disease goes down, but you wait until it grows back.
149
00:16:37.050 --> 00:16:38.560
Room: and then again.
150
00:16:38.620 --> 00:16:42.219
Room: and then Eventually the oncologist runs out of drugs right?
151
00:16:42.350 --> 00:16:44.469
Room: And depending on the disease you have.
152
00:16:44.520 --> 00:16:48.529
Room: You know, there may be several drugs, or there may be very few options.
153
00:16:49.190 --> 00:16:53.180
Room: and at that stage they've lost control of the disease. So it spreads through your body.
154
00:16:53.210 --> 00:16:55.940
Room: and there's nothing they can do anymore to control that
155
00:16:56.840 --> 00:16:58.680
Room: a different approach
156
00:16:59.390 --> 00:17:04.089
Room: is to switch drugs at the nadir of the response.
157
00:17:04.730 --> 00:17:10.879
Room: And here you're working in the dark right? Because you can't see the disease anymore. If You've got a good response.
158
00:17:11.200 --> 00:17:18.779
Room: but we call this extinction therapy, and the reason we call it extinction therapy is that it's inspired by how species actually go extinct in nature.
159
00:17:18.810 --> 00:17:22.450
Room: When you've got a huge population of animals very diverse.
160
00:17:22.619 --> 00:17:31.690
Room: there's always going to be some subpopulation that can survive the pesticide or the asteroid. You know the meteor strike, or the volcano, or whatever.
161
00:17:31.720 --> 00:17:37.759
Room: because they have some sort of adaptation that allows them to to make it through.
162
00:17:38.140 --> 00:17:40.979
Room: But if the population is small
163
00:17:41.330 --> 00:17:42.750
Room: and sparse.
164
00:17:43.110 --> 00:17:52.260
Room: then that is the only opportunity you have to eradicate that group, and you just have kind of relentless pressures that you apply to the system.
165
00:17:52.390 --> 00:18:00.429
Room: And so this is the idea of extinction therapy. Instead of waiting for progression, we switch it to the deer, and you just keep doing different drugs.
166
00:18:00.550 --> 00:18:03.089
Room: and I switched drugs every 4 months.
167
00:18:05.970 --> 00:18:19.469
Room: So th this is my actual tumor data this is a tumor marker called Ca: 2729. And here I was initially diagnosed with, you know, reasonably high tumor markers, and started down this
168
00:18:19.660 --> 00:18:23.129
Room: idea that came out of the meeting for extinction there.
169
00:18:23.430 --> 00:18:25.950
Room: and by the second drug, Adrian Meison.
170
00:18:26.130 --> 00:18:28.239
Room: I was in the ndp zone.
171
00:18:28.370 --> 00:18:34.150
Room: and I was in need by PET Ct. As well as by tumor markers, and stayed.
172
00:18:34.390 --> 00:18:35.290
Room: though
173
00:18:35.310 --> 00:18:37.180
Room: switching different drugs.
174
00:18:37.220 --> 00:18:46.889
Room: so I never stayed on the same drug until the disease came back. I never went off all of the drugs, which is sometimes the the recommended thing in in standard of care. If you reach me.
175
00:18:49.830 --> 00:18:51.639
Room: In addition to that.
176
00:18:52.460 --> 00:19:00.560
Room: I put together a program that would switch my metabolic States in combination with the drugs. So now we're talking about drugs
177
00:19:00.590 --> 00:19:03.890
Room: plus metabolic and behavioral interventions together.
178
00:19:04.150 --> 00:19:09.470
Room: The metabolic switching program involves a few weeks of being on a low-carb
179
00:19:09.630 --> 00:19:16.220
Room: paleo diet, and that has particular functions for metabolic and microbial and immune repair
180
00:19:16.250 --> 00:19:18.019
Room: a Ketogenic phase.
181
00:19:18.130 --> 00:19:23.839
Room: an extended fasting phase which is a 3 day water fast, and then a low methining diet to recover.
182
00:19:24.030 --> 00:19:28.639
Room: And if you're you're interested in reading what the science behind this are published in a
183
00:19:28.680 --> 00:19:32.279
Room: a special issue of cancer evolution frontiers.
184
00:19:32.760 --> 00:19:41.070
Room: so bringing that whole program together. Now you can couple the drug strikes in different metabolic states.
185
00:19:41.160 --> 00:19:50.589
Room: And so you imagine all these different tumor cells, like a single tumor, has about 2 ability on 2 square. Tumor has billions of cells.
186
00:19:50.610 --> 00:20:00.219
Room: and they're all very diverse. Some of them are amino acid metabolizers, ketone metabolizers, and many of them are in the Warburg effect. And they're like, you know, glycolysis and glucose but analyzers.
187
00:20:00.880 --> 00:20:08.890
Room: and so by putting yourself in different metabolic states, and then coupling the drug, strikes with those you can pick off different populations of tumor cells.
188
00:20:10.580 --> 00:20:12.159
Room: That's the idea.
189
00:20:12.540 --> 00:20:21.450
Room: and here's my tumor marker data again. And here i'm going on a very aggressive treatment of an Astrosol and ibrance.
190
00:20:22.000 --> 00:20:25.179
Room: and then a Cape site of being an oral site toxin.
191
00:20:25.600 --> 00:20:37.060
Room: and then ex the mustang and brazenio. They're all very difficult drugs, and the tumor markers are more or less stable. They're not going down. And then they went off all of my drugs.
192
00:20:37.230 --> 00:20:40.670
Room: It did that program that I just described
193
00:20:40.700 --> 00:20:48.769
Room: and got a 23% reduction in my tumor markers. And so by this metric, the metabolic switching
194
00:20:48.970 --> 00:20:50.530
Room: outperformed
195
00:20:50.910 --> 00:20:52.220
Room: the drug treatments.
196
00:20:52.290 --> 00:20:53.829
Room: at least at this stage.
197
00:20:54.060 --> 00:21:03.249
Room: And and so this was encouraging to me, because it showed for me the power of manipulating metabolism to affect
198
00:21:03.270 --> 00:21:08.800
Room: cancer. And of course, now we think of cancer as a very much a metabolic disease.
199
00:21:09.320 --> 00:21:17.530
Room: And so this makes a lot of sense. And subsequently there's actually been a lot of studies on the benefits of fasting and other things for treating cancer.
200
00:21:19.840 --> 00:21:20.500
Room: So
201
00:21:20.610 --> 00:21:27.089
Room: now we have a picture of a better care pathway potentially for advanced cancer. But i'm just one person.
202
00:21:27.640 --> 00:21:34.470
Room: and I've had success with this and many been able to manage my disease for over 3 years.
203
00:21:34.840 --> 00:21:42.740
Room: And you know, just for reference to medium survival for patients with that disease is 3 years. So it's been any d for that period of time.
204
00:21:43.080 --> 00:21:51.900
Room: and the proportion of patients that ever manage to reach any D with my disease is only 6, so it's very rare in and of itself.
205
00:21:53.210 --> 00:21:57.130
Room: So this makes me encouraged, and I want to get it out there for other folks.
206
00:21:58.140 --> 00:22:04.490
Room: What this means is that the need to think differently from not just the drug, but actually to this care. Algorithm.
207
00:22:04.710 --> 00:22:09.439
Room: where there's a sequence, a combination of different drugs that need to be put together with diet
208
00:22:09.470 --> 00:22:15.829
Room: and metabolic changes. And there's this is very kind of complex program that needs to be delivered.
209
00:22:18.590 --> 00:22:29.670
Room: It's a try to imagine how this will work in the future. It's very easy to run clinical trials where the solution is a single drug. You get the drug and you get the sugar pill, and then we measure the results at the end.
210
00:22:29.970 --> 00:22:33.929
Room: But when the solution is a complex care pathway.
211
00:22:34.510 --> 00:22:35.529
Room: how do you
212
00:22:35.550 --> 00:22:42.609
Room: get that out into the world as a discrete entity that can be run in clinical trials reproducibly across different patients.
213
00:22:42.690 --> 00:22:49.060
Room: And then how do you optimize it in a kind of precision, medicine, ecosystem where you need to modify the diet
214
00:22:49.430 --> 00:22:56.970
Room: or the sequence of the drugs in different combinations for particular patients. Right? So it becomes kind of a personalized intervention.
215
00:22:57.450 --> 00:23:01.020
Room: We need to start by understanding the patient.
216
00:23:02.140 --> 00:23:08.620
Room: and we need new precision medicine platforms and technologies that are focused on that call.
217
00:23:08.650 --> 00:23:10.550
Room: So this comes back to the behavior.
218
00:23:11.210 --> 00:23:21.710
Room: And then, once we put these care pathways together as algorithms, we need to be able to deliver those at massive scale in a way that people can follow them right through their smartphone.
219
00:23:23.370 --> 00:23:36.890
Room: They need to be able to access these and the diagnostic and the monitoring tools through their smartphone Remote can't be commuting regularly into the doctor. It's too much trouble. It's too much of a burden on the patients, and their quality of life.
220
00:23:37.940 --> 00:23:51.289
Room: There's a lot of education that goes into this, so you need to be supporting the patients so that they understand why they need to do these different steps? Why do you need to eat certain fruits during the long assignment stage? Is it just because
221
00:23:51.450 --> 00:24:03.230
Room: apples are good? It's not because apples are good. It's because they have absolutely no methion in them, right? And so there's this key bits of information. They need the right information at the right time to make it way through that.
222
00:24:03.800 --> 00:24:06.299
Room: And then you need to be monitoring the patient
223
00:24:06.430 --> 00:24:19.560
Room: right? So we need new tools to monitor and get ahead of any problems. And all of this, all of this complexity needs to be safe needs to be easy to use, and it needs to be massively scalable and improve outcomes.
224
00:24:19.860 --> 00:24:27.540
Room: And I'm telling you a story from cancer. But you can imagine the same problem for alzheimer's to see Parkinson's disease diabetes.
225
00:24:28.100 --> 00:24:30.459
Room: kidney transplants. Anything right.
226
00:24:30.700 --> 00:24:32.880
Room: There's a huge opportunity to solve this.
227
00:24:35.640 --> 00:24:47.819
Room: So they this is too big of a problem for one goofy guy like me to take on, of course, and and instead, what we decided to do was build a platform to help other people
228
00:24:48.200 --> 00:24:49.560
Room: like you guys
229
00:24:49.660 --> 00:24:55.899
Room: identify these problems and have access to all the tools that you would need to go out into the world and solve them.
230
00:24:56.180 --> 00:24:58.240
Room: Build companies based on them.
231
00:24:58.980 --> 00:25:04.569
Room: build apps based on them. Solve any particular problem that I would be interested in.
232
00:25:04.890 --> 00:25:06.489
Room: so that became the mission.
233
00:25:06.800 --> 00:25:24.289
Room: And I I have a partner in this mission. So there's there's me, and I've got this background in genomics. And we publish this paper using artificial intelligence to decompose complex patterns of behavior. So we had some kind of 10 years of research working on that problem.
234
00:25:25.220 --> 00:25:35.289
Room: and I and I kind of brought this research component into the into the organization, and I met Jeff Barson and Jeff is the CEO of storyline.
235
00:25:35.610 --> 00:25:49.860
Room: and you know together, this this combination of skills that we were able to bring together is very unique. So, as you guys are thinking about the organizations and the companies that you want to build, finding partners that have these extraordinarily
236
00:25:49.870 --> 00:26:06.209
Room: complementary sets of expertise can be such a huge advantage. Right? It's it's fortuitous. It involves kind of a lot of networking and talking to people and helping, you know, getting in a program where they can get you to meet people with these different skills. But making these partnerships is like
237
00:26:06.380 --> 00:26:07.290
Room: critical.
238
00:26:07.390 --> 00:26:22.059
Room: and out of all the technologies and all of the things and all the ideas that you come up with your company. The most valuable things that will be in your company or in your organization will be the unique set of experiences, talent, and understanding
239
00:26:22.110 --> 00:26:24.169
Room: that the founders in the leadership have
240
00:26:24.640 --> 00:26:29.499
Room: so so the people that you bring on board. That's the most valuable resource in your startup.
241
00:26:30.320 --> 00:26:48.200
Room: Jeff had run a series of medical clinics, so he understood medical clinics. He had built a telehealth platform called Tele Doc, which is, you know, pretty successful company has been through a few different different names over the years.
242
00:26:48.210 --> 00:26:53.159
Room: and one of the leading teleheld software, platforms, and then he led a
243
00:26:53.290 --> 00:26:59.139
Room: innovation at the world's largest AI company for Job Hiring called higher
244
00:26:59.480 --> 00:27:02.589
Room: and it so. So if you're going to get a job
245
00:27:02.610 --> 00:27:07.029
Room: driving an uber or something like that. One of the first steps in that process
246
00:27:07.220 --> 00:27:09.040
Room: is a video interview.
247
00:27:09.350 --> 00:27:25.630
Room: and then AI analyzes your facial, vocal and speech patterns, and predicts whether you're going to be a safe and good Uber driver or Delta Airlines employee, or one of these massive entities that needs to hire people effectively and safely.
248
00:27:26.760 --> 00:27:27.700
Room: So
249
00:27:27.820 --> 00:27:33.430
Room: Jeff had seen that AI was being used for analyzing behavior in this ecosystem.
250
00:27:33.490 --> 00:27:42.499
Room: and we could see that we could learn a tremendous amount from that and lift it over into the biomedical research and and care community.
251
00:27:43.720 --> 00:27:54.960
Room: So repurposing some things that we're working in another area and bringing it into this world was really a a big part of our secret so far, so we found it storyline.
252
00:27:55.040 --> 00:28:04.349
Room: and we compare ourselves, you know, to the other technologies. As you're kind of evaluating your startup, and whatever value you're going to bring into the world. It's good to kind of
253
00:28:04.430 --> 00:28:14.659
Room: think about what solutions are out there in the world. We've got questionnaires and wearables and X-rays and Mris, and expert clinicians and fmri's, and all that kind of stuff
254
00:28:14.950 --> 00:28:15.870
Room: and
255
00:28:16.300 --> 00:28:21.509
Room: some of these are, you know, diagnostically very effective, but very expensive
256
00:28:21.580 --> 00:28:22.979
Room: and not very scalable.
257
00:28:23.600 --> 00:28:25.370
Room: Others are very cheap.
258
00:28:25.410 --> 00:28:28.840
Room: but they're not very diagnostic or predictive.
259
00:28:29.380 --> 00:28:31.870
Room: So they're not very information rates like a question.
260
00:28:32.280 --> 00:28:42.920
Room: And what storylines AIM to do is work in this area where it's very cheap and very, very scalable, but it's also very, very valuable and rich and useful and predictive of diagnostic data.
261
00:28:43.820 --> 00:28:47.219
Room: So we work. I I will sometimes say that
262
00:28:47.320 --> 00:28:53.520
Room: it because I come out of this genomics background and model story. Sorry I'm very much on illumina.
263
00:28:53.730 --> 00:28:55.419
Room: and when I was at
264
00:28:55.470 --> 00:29:05.899
Room: Harvard Illumina invested in me and started my career, and you know I you know, I think to Illumina many times. But I learned a lot from them about
265
00:29:06.090 --> 00:29:09.410
Room: how to solve many problems in big data.
266
00:29:09.630 --> 00:29:12.240
Room: And so we modeled storyline very much along them.
267
00:29:14.160 --> 00:29:17.899
Room: this is a screenshot of the interface.
268
00:29:18.070 --> 00:29:20.900
Room: And so if you were
269
00:29:21.000 --> 00:29:27.440
Room: to set up a professional account because you're a researcher, clinician or an entrepreneur.
270
00:29:27.610 --> 00:29:36.299
Room: You log into storyline, and you'd see your dashboard. And this is where you would manage all of the products that you're going to build within the platform using AI,
271
00:29:38.310 --> 00:29:40.790
Room: And this is the
272
00:29:41.420 --> 00:29:44.050
Room: part of the platform. It's called programs.
273
00:29:44.090 --> 00:29:47.580
Room: And essentially what you can do in here is build an app.
274
00:29:47.890 --> 00:29:49.650
Room: and you can build an app
275
00:29:49.800 --> 00:29:53.189
Room: for a care pathway. AI supported care pathway
276
00:29:53.350 --> 00:29:55.399
Room: for whatever problem you want to solve
277
00:29:55.810 --> 00:29:59.150
Room: within minutes. It takes hardly any time at all.
278
00:30:00.600 --> 00:30:05.059
Room: and you can deliver educational content. Video: text
279
00:30:05.120 --> 00:30:14.829
Room: and more. There's all kinds of questions that are would be typical for red cap, other questionnaires and things like that. You can plug in here. And then there's the video question.
280
00:30:14.950 --> 00:30:19.210
Room: and that's what we'll see is where all of the data and the value really comes from.
281
00:30:19.690 --> 00:30:20.740
Room: So
282
00:30:21.370 --> 00:30:30.230
Room: this is, you know, there's there's much to say about the platform. But this is one of the key components, right? So you can design care pathways. Educate folks.
283
00:30:30.240 --> 00:30:42.039
Room: communicate with them at scale. It enables all of the AI tools that i'm going to talk about here in a minute. assess subtype monitor research diagnosed, etc. And then you can. Once you've got that app built
284
00:30:42.740 --> 00:30:46.350
Room: It's yours right. You have built and invented something.
285
00:30:46.780 --> 00:30:53.990
Room: and now you can get it out into the world at massive scale and start to help patients. And this is what I really want people to do with this platform.
286
00:30:55.190 --> 00:31:00.420
Room: When you do that, what happens on the patient's side is they get a note.
287
00:31:00.480 --> 00:31:02.309
Room: and they click on the link.
288
00:31:02.490 --> 00:31:09.010
Room: and they're taken straight into the storyline app. And now they can do these asynchronous interviews over their phone.
289
00:31:09.530 --> 00:31:12.460
Room: So you'll ask a question. How are you feeling today?
290
00:31:12.500 --> 00:31:20.360
Room: The person will respond it videotapes their response that video is moved off their phone and securely stored in the cloud
291
00:31:21.020 --> 00:31:24.630
Room: in a Hipaa Gdpr: Military grade security
292
00:31:24.740 --> 00:31:28.789
Room: data storage solution is designed specifically for this.
293
00:31:29.640 --> 00:31:37.660
Room: I wanted patients to have control over their data, and so patients always have control over their data within the storyline system. They want to delete it.
294
00:31:37.710 --> 00:31:38.899
Room: They can delete it.
295
00:31:38.990 --> 00:31:42.470
Room: And I know that that gives entrepreneurs and innovators kind of a bit of a
296
00:31:43.280 --> 00:31:50.540
Room: art publications. But I can tell you that that is the way the world is going for. Sure. People need to have control over their own data.
297
00:31:52.060 --> 00:31:56.170
Room: once their video has been moved up into the cloud.
298
00:31:56.320 --> 00:31:59.270
Room: Then you have some real superpowers.
299
00:32:00.010 --> 00:32:06.729
Room: We've built a micro services, pipeline of many different AI algorithms.
300
00:32:06.760 --> 00:32:14.640
Room: and all of the algorithms are put together to measure over 30,000 different micro features. Now.
301
00:32:15.080 --> 00:32:19.740
Room: I know the slide says 20,000, but you know, goes up every week. So I gotta update the slide
302
00:32:20.960 --> 00:32:29.870
Room: from the video we get information like objective measures of pupil dilation, right? Which is re regular by sympathetic tone.
303
00:32:30.080 --> 00:32:35.849
Room: You can see, I tracking, and i'll show you some examples of how we use this for neurological assessments.
304
00:32:35.980 --> 00:32:44.460
Room: head movements. You can see blood flow. Patterns of blood flow changes across the face by measuring RGB values across 400 different points on the face.
305
00:32:44.740 --> 00:32:52.930
Room: You can look at respiration, responses, micro expressions, and micro movements across the 40 different muscles in the face.
306
00:32:53.190 --> 00:32:54.169
Room: and then
307
00:32:54.610 --> 00:32:59.490
Room: it's not just the motor components, or even you know, the power of the skin.
308
00:32:59.720 --> 00:33:17.659
Room: But you can analyze speech. And so word choice. All the natural language processing algorithms here have been put in place to go from speech to text, and then you can analyze what people say, the structure of your sentences, personality, trade, speech, patterns, even filler words as pauses.
309
00:33:17.880 --> 00:33:23.999
Room: and how they articulate their responses, the sentiment of their responses, and what they're trying to communicate.
310
00:33:24.520 --> 00:33:30.610
Room: and it's not just what you say, but how you say it! And so there are thousands of audio measures.
311
00:33:30.670 --> 00:33:32.330
Room: you know, harmonic
312
00:33:32.660 --> 00:33:33.960
Room: ratios.
313
00:33:34.010 --> 00:33:37.270
Room: energy in the in the voice
314
00:33:38.420 --> 00:33:42.850
Room: and pitch and tone changes, volume, changes, etc. Pronunciation.
315
00:33:42.920 --> 00:33:54.359
Room: and you can detect a number of different things here. Emotions in the voice, vocal microchammers. There's entire companies built just on the voice for diagnosing different disorders.
316
00:33:54.380 --> 00:33:58.190
Room: and all of that power is built into all this, this ecosystem.
317
00:34:00.280 --> 00:34:01.989
Room: So i'm pretty excited about that
318
00:34:02.230 --> 00:34:13.329
Room: coming from the genomics background where, when we sequence genomes, we could not just print out the 3 billion bases in a big book and give it to people.
319
00:34:13.500 --> 00:34:17.870
Room: We had to create file formats to deal with this, to share it.
320
00:34:17.920 --> 00:34:22.129
Room: to study it, to research, it, to integrate it in other multi-omic studies
321
00:34:22.670 --> 00:34:30.659
Room: and those are called fast queue files right, Sam Files bam files in the genome World. We had to develop analogs of those for the behavior world.
322
00:34:31.110 --> 00:34:35.020
Room: and these are called story, arc, story, time and poem files.
323
00:34:35.120 --> 00:34:52.519
Room: Unlike the genome, there's a dimension of time, right? So you've got these thousands of measures that you might pull out in a particular part of a video. But then all of these have to be built out in time. So these are actually the screw. These poems and screenplay files are are massive data files.
324
00:34:52.600 --> 00:34:58.279
Room: and they can be brought into whatever kind of multi-omics, precision medicine ecosystem you're thinking about building.
325
00:34:58.370 --> 00:35:02.490
Room: If you're collecting genome data microbiome data, immune general data.
326
00:35:03.020 --> 00:35:04.889
Room: You know, whatever
327
00:35:05.480 --> 00:35:12.199
Room: you should be collecting this type of data as well as integrate and integrating it into your subtyping and studies.
328
00:35:14.030 --> 00:35:29.660
Room: Once you have this data, you can do all kinds of things. Research, the data discover subtypes phenotypes for different patients discover symptoms, and you can also train models that become biomarkers
329
00:35:29.700 --> 00:35:33.279
Room: for abnormalities or difficulty. We'll talk about some examples.
330
00:35:33.480 --> 00:35:41.629
Room: and then what storyline enables you to do is to build visualizations of those models that make it very easy for your customer.
331
00:35:41.920 --> 00:35:43.219
Room: or a doctor.
332
00:35:43.250 --> 00:35:47.520
Room: or whomever even a patient to interpret and understand and use the information.
333
00:35:49.840 --> 00:35:57.389
Room: And then, once you built one of these things on the platform, we recognize right away that we needed everybody to be aligned.
334
00:35:57.750 --> 00:36:06.629
Room: So if you build a model on the storyline platform. You build a care pathway, for, you know, liver transplants, or whatever you're imagining.
335
00:36:06.790 --> 00:36:09.759
Room: You can publish that in the storyline library.
336
00:36:10.130 --> 00:36:15.709
Room: and that's like an app store like an apple app store for health care, AI algorithms
337
00:36:15.890 --> 00:36:17.209
Room: and care.
338
00:36:17.400 --> 00:36:21.179
Room: And then immediately that is available to everyone throughout the world
339
00:36:21.210 --> 00:36:22.419
Room: through telehealth.
340
00:36:22.670 --> 00:36:24.049
Room: and they can pay for it.
341
00:36:24.660 --> 00:36:27.179
Room: So people are actually building their start ups.
342
00:36:27.270 --> 00:36:43.319
Room: They don't need to train all the models and build all the software and the permissions and the data, storage solutions and privacy and the safety problems, etc., etc. It's already solved on storyline. You can go straight in, build your care pathway, train your algorithms, publish and have a product you can take out into the market.
343
00:36:46.050 --> 00:36:59.530
Room: Yeah, so that's a really really important question. And
344
00:36:59.780 --> 00:37:06.610
Room: there's there's a number of different answers there. So one is that some metrics are very sensitive to skin tone.
345
00:37:06.820 --> 00:37:18.210
Room: There are challenges with people who have really really black skin and getting enough contrast to measure some of the facial movements and dynamics
346
00:37:18.880 --> 00:37:25.540
Room: folks with lighter skin tones. We're finding that all of that works really really effectively.
347
00:37:25.630 --> 00:37:35.429
Room: Some algorithms and the speech and the text components. If you have accents and some of these other.
348
00:37:35.650 --> 00:37:36.359
you know.
349
00:37:37.040 --> 00:37:39.399
Room: variations in speech patterns
350
00:37:39.440 --> 00:37:42.770
Room: can introduce errors and difficulties.
351
00:37:42.850 --> 00:37:53.269
Room: And so, whatever you're building on the AI platforms, you have to be very cautious of the types of biases that you're introducing, and you need to understand what those biases are.
352
00:37:53.540 --> 00:37:57.089
Room: So what we've done at storyline is to build a bias report.
353
00:37:57.310 --> 00:37:59.480
Room: Let's say you collect data from like
354
00:37:59.590 --> 00:38:02.980
Room: a 1,000 people, and you bring it into your ecosystem
355
00:38:03.040 --> 00:38:11.630
Room: right away. If your model is picking up on, let's say you're diagnosing depression, and you built a model that is diagnostic of depression. Symptoms
356
00:38:13.220 --> 00:38:15.390
Room: The you'll test the model
357
00:38:15.620 --> 00:38:18.070
Room: right? You'll train it, and then you'll test it.
358
00:38:18.230 --> 00:38:24.789
Room: We will see whether there is a bias in your results for particular racial groups or accents.
359
00:38:25.400 --> 00:38:38.400
Room: and that will help you to diagnose problems with within your ecosystem. And then you need to work through that, to be honest, there there is not going to be a single algorithm that solves every problem for every group, 150
360
00:38:41.620 --> 00:38:44.949
Room: super important questions. Whoever put that up. Thank you.
361
00:38:46.740 --> 00:38:51.179
Room: One of the things that I learned from the genome medicine revolution is that
362
00:38:51.230 --> 00:38:53.650
Room: one of the big mistakes they made
363
00:38:53.820 --> 00:38:56.299
Room: is allowing everybody to become silent.
364
00:38:56.860 --> 00:39:01.010
Room: What that means is that every institute or every researcher.
365
00:39:01.370 --> 00:39:18.939
Room: the research is the worst. So they they would build these little sort of empires of data, and not allow anybody else to touch the patients right, and they would have their whole career. And they publish papers based on these patients that only they ever had access to. And
366
00:39:19.240 --> 00:39:28.890
Room: you can imagine, like, how frustrating and difficult that it's very good for that person's career, but very difficult for the whole exercise and problem in general of improving medical care.
367
00:39:28.900 --> 00:39:44.869
Room: in diagnostics. If all genomes have been shared initially, so, there was a way to integrate the data for all genome sequences. As they came off of an aluminum sequencer, we would be a 100 miles ahead of where we are right now.
368
00:39:45.040 --> 00:39:48.549
Room: right data, integration and sharing would have moved us
369
00:39:48.600 --> 00:39:50.100
Room: miles faster.
370
00:39:50.430 --> 00:39:57.660
Room: And so what we've done at storyline is is to correct this problem. It's it's very difficult to correct, but we're trying
371
00:39:58.070 --> 00:40:09.180
Room: where it's one ecosystem right? So there's the research and people are launching their research projects and things like that. The data can be de-identified, and it can be shared into a common platform.
372
00:40:09.320 --> 00:40:11.160
Room: And that's really powerful.
373
00:40:11.310 --> 00:40:20.529
Room: but contributing to that you you really are creating something that can move your own research much, much more quickly, right? Because if you're trying to train an AI model for one.
374
00:40:20.570 --> 00:40:22.010
Room: let's say depression again.
375
00:40:22.770 --> 00:40:25.479
Room: You're not just trying to diagnose depression from healthy.
376
00:40:25.750 --> 00:40:28.899
Room: You're trying to diagnose depression from bipolar
377
00:40:29.280 --> 00:40:41.030
Room: Romania, from sub types of depression, from alcoholism, from you know, all kinds of things so to ever get good models that are effective and accurate. Everybody has to work together.
378
00:40:41.750 --> 00:40:45.149
Room: So the platform has to be designed from the ground up with that in mind.
379
00:40:46.100 --> 00:41:02.150
Room: And then, once these solutions are built, they can be e immediately piped out into the real world right? When I was diagnosed with cancer. It was felt good to write papers. But how how futile right to write a paper and have it published.
380
00:41:02.160 --> 00:41:06.740
Room: and you know hardly anybody reads it right. There's 7,000 papers published every month.
381
00:41:06.870 --> 00:41:08.240
Room: Nobody reads that stuff.
382
00:41:08.520 --> 00:41:12.479
Room: So you you really need to renovate
383
00:41:12.610 --> 00:41:20.329
Room: the whole biomedical research system, so that the research and the discovery is immediately and seamlessly piped into the clinical world.
384
00:41:20.830 --> 00:41:22.569
Room: All of that data just flows.
385
00:41:27.630 --> 00:41:34.180
Room: Any questions. I'm going to get into a few case studies. I want to show show you guys sort of how people are using this.
386
00:41:35.220 --> 00:41:36.109
Room: Yeah.
387
00:41:36.190 --> 00:41:37.589
So to capture
388
00:41:40.210 --> 00:41:54.659
Room: we what we're gonna do is we're gonna be able to integrate wearable data. So let's say you have a problem that you're trying to solve, and part of it has wearable data.
389
00:41:54.990 --> 00:41:58.590
Room: What we can do is the wearable data is
390
00:41:59.180 --> 00:42:17.970
Room: mit ctl, and is going into a data repository storyline through an Api hooks into that repository and pulls your patience data into the same storyline ecosystem. So all of these wearables and other devices can be very easily and seamlessly integrated into the storyline. 150
391
00:42:18.070 --> 00:42:18.919
Room: system
392
00:42:19.050 --> 00:42:22.970
Room: that makes sense. So so it can all be pulled in. Storyline itself
393
00:42:23.090 --> 00:42:25.839
Room: is not focused on wearable type of data.
394
00:42:26.070 --> 00:42:31.640
Room: We're focused specifically on this video captured smartphone interview type of
395
00:42:31.700 --> 00:42:32.490
data.
396
00:42:33.420 --> 00:42:33.979
Yeah.
397
00:42:36.580 --> 00:42:38.790
Room: Any other questions on
398
00:42:39.500 --> 00:42:42.109
Room: If anybody online has a question.
399
00:42:43.200 --> 00:42:45.469
Room: something came up. But i'm not sure.
400
00:42:47.200 --> 00:42:50.529
Room: Oh, okay, okay, okay, good. Yeah, Thank you.
401
00:42:51.390 --> 00:43:00.840
sharaz khan: Yeah, I just you know. on the chat I just put this this program I used last year a nura a I don't know
402
00:43:00.970 --> 00:43:07.179
sharaz khan: this is something similar, but that essentially, you know, on your smartphone you can
403
00:43:07.210 --> 00:43:22.149
sharaz khan: a sort of map out your your your face, and then it it collects, you know. I don't think as as as sophisticated data as a storyline, but it has a good good base. I don't know. If you've heard of that.
404
00:43:23.980 --> 00:43:27.389
Room: I haven't heard of it. Thanks for sharing. I'll I'll check it up.
405
00:43:27.470 --> 00:43:34.599
sharaz khan: sure. Sure. Yeah, No problem. I'll just. I'll just put it on the chat. It's yeah, I see it. I see it up there. Yeah, thank you. Thank you.
406
00:43:36.190 --> 00:43:43.409
Room: You know the AI field is fast moving. I have it. It's it changed by the week. So i'm always grateful for
407
00:43:43.540 --> 00:43:44.919
Room: new information.
408
00:43:47.540 --> 00:43:49.040
Room: yes, I sorry
409
00:43:49.670 --> 00:43:58.920
Room: you're about to move to the next. So, I was wondering then, just about the kind of integrated platform that you were talking about.
410
00:44:00.060 --> 00:44:03.130
Room: I don't know if I got this right? But it seemed like
411
00:44:03.420 --> 00:44:05.300
Room: You're saying how
412
00:44:05.330 --> 00:44:08.189
Room: for example, as a data scientist, I could maybe
413
00:44:08.230 --> 00:44:13.339
Room: look at all the data that has been collected between my models on
414
00:44:13.410 --> 00:44:25.590
Room: and then the files will be in the format of the the story story art storage. So is that accurate that that is accurate. Okay, yeah, if you want to do that, let me know.
415
00:44:25.660 --> 00:44:28.890
Room: Yeah. So so then, my concern or or
416
00:44:28.970 --> 00:44:48.120
Room: the problem that I would be wondering about is, for example, if I wanted to know about depression, then I would have all these features like the spatial expression, etc. But then how would I know the the predictors? Yeah, the labels? Yeah. Yeah. So I haven't got into labeling. But storyline has a whole labeling ecosystem
417
00:44:48.130 --> 00:44:49.750
Room: within its software?
418
00:44:49.880 --> 00:44:52.530
Room: And the data that
419
00:44:53.170 --> 00:44:57.919
Room: You know that within that data sharing ecosystem, we retain the labels.
420
00:44:58.140 --> 00:45:08.409
Room: So there's the the other approach is an unsupervised approach. Right? And and that's actually one of my sort of favorite, but most difficult applications. Yeah.
421
00:45:08.610 --> 00:45:09.940
Room: yeah. And I guess
422
00:45:10.100 --> 00:45:13.810
Room: kind of connected to that would be since you talked about desk, you but
423
00:45:14.010 --> 00:45:23.840
Room: the fax cue files, I I think, would be kind of a bit easier to anonymize just because but with this. You're kind of publishing
424
00:45:24.730 --> 00:45:40.000
Room: So much information about their behavior and their facial expressions, and even their speech patterns. Is there any concern? Or how do you deal with. I is a huge concern for sure, and the we have kind of a de identification process that
425
00:45:40.010 --> 00:45:48.160
Room: it goes through a number of steps. So the first is to scrub. So we have an algorithm that goes through, and all of that speech to text data, right? If they say, you know
426
00:45:48.210 --> 00:45:49.300
Room: my name is
427
00:45:50.090 --> 00:45:57.719
Room: John and my wife so and so is like driving me crazy, right? That's personal information that will all get scrubbed
428
00:45:57.890 --> 00:46:06.049
Room: by the algorithm that goes through the speech to text. So all of that personal information address local, you know, etc. is screwed.
429
00:46:06.490 --> 00:46:11.319
Room: Then there's the facial data, right? And
430
00:46:11.410 --> 00:46:16.129
Room: The facial data is not high enough resolution
431
00:46:16.570 --> 00:46:21.350
Room: the way that we presented in story our and story time to reconstruct a person's face
432
00:46:21.460 --> 00:46:22.740
Room: accurately.
433
00:46:22.860 --> 00:46:36.889
Room: It's it's good, but it's not designed for that specific problem. And the the added thing that we can do to even further anonymize that information is just dimension reduction.
434
00:46:37.680 --> 00:46:38.450
Room: So
435
00:46:39.140 --> 00:46:42.159
Room: then you're you know, inheriting kind of a
436
00:46:42.600 --> 00:46:44.950
Room: reduced dimension feature set?
437
00:46:46.140 --> 00:46:48.650
Room: Yeah, yeah, it's a great question.
438
00:46:48.800 --> 00:46:52.749
Room: Yeah. Oh, oh, sure.
439
00:46:53.030 --> 00:46:54.310
I see them the screen. That
440
00:46:54.350 --> 00:46:56.270
Room: Oh, oh, good, good!
441
00:46:58.240 --> 00:46:59.260
It's fine, Thank you.
442
00:46:59.950 --> 00:47:02.680
It seems to
443
00:47:02.830 --> 00:47:04.690
the social.
444
00:47:05.130 --> 00:47:05.899
Let me see.
445
00:47:06.620 --> 00:47:07.240
Got it.
446
00:47:07.600 --> 00:47:11.770
Room: Yeah, yeah. What a great question. So
447
00:47:12.240 --> 00:47:13.359
Room: you know we
448
00:47:13.480 --> 00:47:19.669
Room: The definition of behavior in this case is anything you can video record on a video of a person.
449
00:47:20.200 --> 00:47:25.529
Room: So so you're you're absolutely right. It's a bit of a narrow term in this case.
450
00:47:26.210 --> 00:47:32.349
Room: because your app might record. You might ask people to take a video of their surgical wound.
451
00:47:32.610 --> 00:47:39.290
Room: Right. And that's the data that you capture in your storyline app. And you're trying to monitor inflammation or something like that.
452
00:47:39.790 --> 00:47:50.509
Room: we have people that approach this because they want to look at skin health right like bags into the eyes. So much people are sleeping acne, and that kind of stuff, and that's not behavior. That's
453
00:47:51.330 --> 00:47:52.859
Room: you know, skin.
454
00:47:53.230 --> 00:47:57.919
Room: So so it you you you're right. The the term behaviors but
455
00:47:58.040 --> 00:48:00.939
Room: narrow, but it's a component big component of this.
456
00:48:03.550 --> 00:48:05.500
Room: I hope that answers the question.
457
00:48:05.560 --> 00:48:10.309
Room: Anything in video basically is is of use on the story like platform.
458
00:48:11.690 --> 00:48:15.579
Room: The other questions i'm curious how stored I might support
459
00:48:15.940 --> 00:48:33.390
Room: engaged feedback with patients and care partners along with clinicians and researchers. Yeah, really, really important thing. So you can use storyline literally to sir. have these kind of asynchronous conversations with your patients or
460
00:48:33.450 --> 00:48:52.009
Room: study participants where they can leave you a video message that you watch or staff at your clinic watches. You can obviously just send them a little quick questionnaire Like, Are you doing okay? Are you finding that this is working? Okay, Are you happy with my care. It's, you know, whatever the questionnaires could be.
461
00:48:52.300 --> 00:48:57.150
Room: But but you can also do this through video and asynchronous video messaging.
462
00:48:57.170 --> 00:48:57.770
Yeah.
463
00:48:58.580 --> 00:49:06.790
Room: And then the other question is, how do you sort verbal variables, word, choice, set and structure, etc., for patients
464
00:49:06.940 --> 00:49:25.930
Room: for whom English is not their first language. Yeah, so that's a really really good point. And this comes into the problem of bias. For folks who who have a accent because English is not their first language, you know, on the data science side of things. You need to be able to
465
00:49:25.940 --> 00:49:31.340
Room: think about that and solve that problem within your ecosystem. And what we AIM to do is give you
466
00:49:31.480 --> 00:49:35.209
Room: all of the data that you would need to do that.
467
00:49:35.970 --> 00:49:41.849
Room: Many of the AI algorithms that exist in the world were pretty well for Spanish or for English speakers.
468
00:49:41.980 --> 00:49:44.119
Room: Spanish is getting pretty good.
469
00:49:44.160 --> 00:49:59.599
Room: and there's a number of different platforms that are coming with really good speech to text for many, many other languages. So we're getting better and better. The problem is, the intersection between right where you've got a thick accent. And and you're speaking language is not your native town.
470
00:49:59.650 --> 00:50:13.040
Room: So it's a good question. Another way to answer that question is that when you're trying to build an AI model to diagnose something. You want a lot of redundancy and features. You don't want everything to rely on one feature that might be contaminated by accent.
471
00:50:13.140 --> 00:50:23.149
Room: So you've got other measures of you know, blinking rate patterns. For example, if you, if you look at folks with treatment, resistant depression. Their faces are flat.
472
00:50:23.290 --> 00:50:24.489
Room: and they played
473
00:50:25.130 --> 00:50:31.039
Room: very slowly, so you could. You not only get information out of the voice, but you get a lot of
474
00:50:31.110 --> 00:50:33.020
Room: other information that's redundant.
475
00:50:34.450 --> 00:50:35.649
Room: I hope that's helpful.
476
00:50:36.560 --> 00:50:40.740
Room: These biases are so. There's 2 ways to look at the biases. One is like
477
00:50:41.210 --> 00:50:42.930
Room: scary problem, right?
478
00:50:42.960 --> 00:50:59.510
Room: But the other is huge opportunity. Because if you're out there collecting data and you can identify these different sub populations, people and build algorithms that are useful. That's personalized medicine for particular subgroups. So it's an offer. Think of it as an opportunity to go out there and get the data and solve the problem.
479
00:51:00.440 --> 00:51:01.149
Room: Yeah.
480
00:51:01.710 --> 00:51:08.330
Room: so this might be the misunderstanding like that. AI can go. But
481
00:51:08.660 --> 00:51:12.310
Room: how did you Did you meet with experts on
482
00:51:12.690 --> 00:51:14.590
Room: conditions to build
483
00:51:14.700 --> 00:51:18.799
Room: those feature profiles? Well, what a good question! So
484
00:51:19.610 --> 00:51:26.179
Room: yes, the way I I I by far, not even close to being an expert
485
00:51:26.280 --> 00:51:31.539
Room: in any of these different domains. Right? We're talking about. We're going to talk about a lot of different stuff cancer
486
00:51:32.230 --> 00:51:38.199
Room: depression, psychedelic treatments, you know we're we're doing so many of homelessness and addiction.
487
00:51:40.100 --> 00:51:43.049
Room: So so the way it has to work is
488
00:51:43.360 --> 00:51:45.719
Room: storyline builds the platform.
489
00:51:46.020 --> 00:51:50.430
Room: and then we partner with people who are experts in a particular domain.
490
00:51:50.870 --> 00:51:52.290
Room: and they go out
491
00:51:52.370 --> 00:51:56.930
Room: and figure out how to solve, how to apply the platform in their area right?
492
00:51:57.000 --> 00:51:59.839
Room: Because they understand the problems they understand.
493
00:51:59.980 --> 00:52:06.040
Room: You know the sub types, and and what sorts of features and measures and assessments might actually be useful
494
00:52:06.360 --> 00:52:11.180
Room: in in many cases. You can take an assessment that's already being used like animal naming.
495
00:52:11.400 --> 00:52:15.160
Room: which is, you know, kind of a cognitive test. It's been used for a long, long time.
496
00:52:15.480 --> 00:52:17.790
Room: and now just delivered as An AI
497
00:52:17.880 --> 00:52:29.070
Room: kind of thing. And now you're and now you get all this new information as somebody tries to name animals over the course of a minute and and the phenotypic patterns are quite striking and interesting.
498
00:52:33.220 --> 00:52:37.849
Room: Okay, so. And this is going to speak a bit to your your point.
499
00:52:37.980 --> 00:52:43.539
Room: So this is I'm going to just tell a couple 3 examples. One example is neurology.
500
00:52:43.580 --> 00:52:52.890
Room: So the way neurology typically works right now is you've got a person that comes into the clinic they presented with symptoms that bother them enough in their life that they now come in to see a physician
501
00:52:52.940 --> 00:52:55.379
Room: and that neurologist does it work up
502
00:52:55.780 --> 00:53:04.630
Room: typically takes an hour or more and and goes through. And and they're assessing different things and making expert judgments around phenotypes, abnormalities, etc.
503
00:53:05.120 --> 00:53:19.240
Room: So the problem that Fanny, who's an outstanding neurologist, a researcher at Mount Sinai in New York she has taken on using storyline is that neurological disorders are incredibly diverse.
504
00:53:19.570 --> 00:53:24.810
Room: and they're frankly, frequently missed or misdiagnosed in the primary care setting.
505
00:53:25.330 --> 00:53:32.899
Room: You know, primary care. Doctors are not trained to detect and diagnose the over 400 different neurological disorders there are in the world.
506
00:53:34.250 --> 00:53:47.239
Room: But these disorders really benefit from early detection. And so we want to be able to make, you know assessments of neurological disorders more scalable, readily accessible and and available for people to get diagnosed early.
507
00:53:47.570 --> 00:53:54.449
Room: So Fanny seeks to build a neurological burden index which, where the objective is just to say.
508
00:53:54.490 --> 00:53:55.569
Room: are you kind of
509
00:53:55.620 --> 00:54:01.750
Room: showing some symptoms that look a little abnormal compared to the population of people that is in your demographic.
510
00:54:02.070 --> 00:54:05.239
Room: and then, if so, maybe you know, this is an alarm that
511
00:54:05.910 --> 00:54:09.800
Room: would would would drive you to go into the clinic to see care.
512
00:54:11.170 --> 00:54:20.389
Room: there's the supervised problem where we've got over 400 known neurological disorders, and we'd like to have an objective smartphone based assessment that says you've got this versus that.
513
00:54:21.370 --> 00:54:34.130
Room: And then there's the one that I you know. I'm. A researchers of one I love the most, which is the unsupervised approach where you collect massive amounts of data, and you discover new neurological disorders that nobody even had names for
514
00:54:36.390 --> 00:54:40.970
Room: right. And so Fanny's approach. Then is she set up on storyline.
515
00:54:41.050 --> 00:54:54.979
Room: And she's built a 15 min integrative, neurological, psychiatric and clinical psychology assessment. And then she delivers that to people's smartphones, captures the data because they do the assessment. I'll show you how it works.
516
00:54:55.280 --> 00:55:00.459
Room: And then from the data, the idea is to discover these important phenotypes
517
00:55:00.520 --> 00:55:09.260
Room: train diagnostic models based on the labels for different different phenotypes and subtypes, sometimes quite rare. Neurological disorders
518
00:55:09.750 --> 00:55:11.659
Room: validate those models
519
00:55:11.940 --> 00:55:16.030
Room: and then make them available for everybody to use worldwide through telehealth.
520
00:55:16.640 --> 00:55:29.069
Room: and it's so easy to enroll people. They can just take a picture of the QR. Code on their phone, and that will take you straight into download the storyline assessment and you could do a neurological system on your phone right now. it.
521
00:55:29.090 --> 00:55:32.010
Room: And yeah, it looks like somebody's gonna go for it. So great.
522
00:55:36.110 --> 00:55:44.669
Room: Okay. So now, here's what it's like. So the neurological assessment what? What I think is so cool about this. It's asynchronous.
523
00:55:44.710 --> 00:55:52.100
Room: People are watching the clinician. Tell them what to do, and they're just following along right. You can never do this with a questionnaire.
524
00:55:52.670 --> 00:55:53.379
Room: Oh, hold on!
525
00:55:59.640 --> 00:56:01.429
Room: Oh, no! What's up.
526
00:56:04.520 --> 00:56:06.799
Room: Oh, I know why Hold on.
527
00:56:10.070 --> 00:56:10.629
Hang on!
528
00:56:16.590 --> 00:56:18.150
Room: The video is not in the
529
00:56:19.700 --> 00:56:20.299
here.
530
00:56:25.980 --> 00:56:27.830
Room: can. Can folks on
531
00:56:27.990 --> 00:56:30.460
Room: slack or excuse me on
532
00:56:30.630 --> 00:56:32.089
Room: zoom see the
533
00:56:32.350 --> 00:56:33.310
Room: No.
534
00:56:35.840 --> 00:56:38.260
Room: So maybe I just need to share.
535
00:56:39.000 --> 00:56:40.459
I think this will work
536
00:56:40.650 --> 00:56:41.330
to.
537
00:56:43.050 --> 00:56:43.839
Okay.
538
00:56:45.680 --> 00:56:46.620
Yeah, this is exactly.
539
00:56:47.900 --> 00:56:49.039
And I have to.
540
00:56:49.430 --> 00:56:50.100
if You'
541
00:56:53.330 --> 00:56:53.930
on
542
00:56:57.530 --> 00:56:58.129
by.
543
00:57:05.210 --> 00:57:06.370
Room: Okay.
544
00:57:13.000 --> 00:57:18.339
Room: So so that would go on, you know, for 50 min, or whatever you feel like You're having this face to face
545
00:57:18.400 --> 00:57:20.810
Room: assessment and
546
00:57:22.720 --> 00:57:33.060
Room: And then what we can do with AI models is, we can do very precise tracking of all of the motor patterns and movements in the face. Right? So this is me. I'm looking up.
547
00:57:33.260 --> 00:57:51.670
Room: You're right looking left down like you would in a typical neurological assessment and then micro-expressions and changes movements and things like that in the face are being measured. And across all these different points in the face that data is being extracted in Xyz coordinates, and then architect it into those story story time, trials.
548
00:57:52.070 --> 00:57:54.189
Room: It's not just the
549
00:57:55.470 --> 00:57:57.759
Room: face that we can analyze.
550
00:57:58.110 --> 00:58:08.999
Room: So we can do essentially these full kind of clinical interviews, but capture a lot of different data. So here's me and it's measuring over 400 different points across my face, 101.
551
00:58:09.450 --> 00:58:11.950
Room: So all micro expressions and movements.
552
00:58:12.170 --> 00:58:18.709
Room: It's also capturing the RGB values. So we're going to get information about skin power, skin, health
553
00:58:18.920 --> 00:58:22.849
Room: and potentially for fusion patterns across the face.
554
00:58:23.610 --> 00:58:35.820
Room: Motor tests, you know. So in neurological assessments, so often have these types of little motor assessments, all of this becomes measurable and quantitative instead of just a kind of judgment.
555
00:58:36.240 --> 00:58:44.410
Room: and you know i'm just doing it at home. So if I I don't have to come in from the country, or whatever to get access to the character Major center.
556
00:58:46.090 --> 00:58:49.540
Room: And that's good for early diagnosis. Right? That's that's the goal there.
557
00:58:52.480 --> 00:59:10.919
Room: So once you have this type of data, you can do dimension reduction and project that data into multi dimensional space. And here, you know, even though the data is captured in 2D on the video, we can use a models to estimate the D projection of the data.
558
00:59:10.930 --> 00:59:15.929
Room: And so you're seeing somebody's micro expressions and movements. They're being projected on across all these spots.
559
00:59:16.000 --> 00:59:31.910
Room: And we're just plotting the data as we move through the video frames, and each frame has, we just a Pca. On the data to come, produce it, and then you can follow the changes to micro expressions and move through this dimension reduction
560
00:59:33.540 --> 00:59:37.819
Room: very, very cool. So a neurological assessment becomes highly, highly quantitative.
561
00:59:41.740 --> 01:00:01.309
Room: like, I said, one of my favorite PET projects is not supervised learning, but actually unsupervised learning, because this is where you get the opportunity for discovery. And so some of the data scientists in the in the group have really been focused on approving, unsupervised learning algorithms. And Here is an example where they made synthetic data
562
01:00:01.380 --> 01:00:09.159
Room: where there's no. So if each.is a patient you can imagine here, and you've captured thousands of measures for each patient. So it's high dimensional data.
563
01:00:09.260 --> 01:00:10.200
Room: And then
564
01:00:10.320 --> 01:00:21.799
Room: here, there's really no subgroups, and then we gradually create subgroups. So we've got a ground truth right. And now there are real subgroups coming out in the synthetic data.
565
01:00:21.920 --> 01:00:33.089
Room: And what Jared on the team has done is he's tested some of the different unsupervised learning algorithms that exist in the world against his own storyline cluster test
566
01:00:33.110 --> 01:00:44.540
Room: and what this shows this is the in-group proportion test from Rob Tip Shirani at Stanford, and his test, you know, as you move from not real clusters to very lots of really coherent clusters. One
567
01:00:44.760 --> 01:00:48.779
Room: it sort of gradually improves, but it's not very sensitive.
568
01:00:51.000 --> 01:00:52.070
Room: don't tell Rob
569
01:00:52.120 --> 01:01:09.490
Room: so, and then, but but Jared's test is beautiful. It really scales nicely, as true biological subgroups. Emerge in the data. The score comes up beautifully with with the coherence of that data. And so he has a nice algorithm I think, for finding bona fide subgroups
570
01:01:09.500 --> 01:01:18.310
Room: from this high dimensional data. So if you collect lots of data from different patients, you can sub-type and discover some types in new ways. That's that
571
01:01:18.490 --> 01:01:19.720
Room: cool in my
572
01:01:19.810 --> 01:01:20.880
Room: in my world.
573
01:01:21.020 --> 01:01:32.229
Room: and and you know I always come back to my roots in genetics where we used to. We still think about liability thresholds where you've got a certain number of genetic mutations in your genome
574
01:01:32.640 --> 01:01:35.079
Room: tips you into this clinically relevant
575
01:01:35.180 --> 01:01:54.049
Room: tail of the distribution. Right? So these liability thresholds, and we can start to think about behavior in the same way, where you gather many, many different measures in a certain number of abnormal kind of behavioral features and neurological symptoms, and it tips you into this kind of clinically relevant realm.
576
01:01:56.090 --> 01:01:57.060
Room: Okay.
577
01:01:58.050 --> 01:02:00.589
Room: i'm over time.
578
01:02:00.830 --> 01:02:05.170
Room: Is it? Okay? Yeah, okay, Thank you all of you, for you know.
579
01:02:05.730 --> 01:02:07.820
Room: not going to hockey again. Thank you.
580
01:02:09.050 --> 01:02:10.560
Room: Okay,
581
01:02:11.300 --> 01:02:13.340
Room: This is a project that is.
582
01:02:13.520 --> 01:02:16.999
Room: you know. It's one of those big missions right? And
583
01:02:17.190 --> 01:02:24.389
Room: in in Canada. We're pretty good at taking care of of people right? We're, you know. I'm really proud of that. And and
584
01:02:24.420 --> 01:02:27.229
Room: the home country, but the United States
585
01:02:27.530 --> 01:02:34.990
Room: it's a little different. So in in the United States, you know, 2 and a half 1 million people are incarcerated.
586
01:02:35.730 --> 01:02:44.049
Room: Many people are homeless. They do not have in many cases the same social infrastructure to to help people
587
01:02:44.690 --> 01:02:45.669
Room: and
588
01:02:45.810 --> 01:03:02.029
Room: so there's a huge need in in the world to help folks. And really it's going to be a technology solution that does this. And i'll talk about. You know the details here. So this is a partnership between the National Institute for Jail operations in the United States
589
01:03:02.080 --> 01:03:14.950
Room: last mile, which is an organization that helps folks coming out of prison get jobs and transition and reintegrate into the world and seek haver and seek even helps folks like veterans.
590
01:03:14.960 --> 01:03:22.509
Room: and people living on the street get off the street, get homes, get stable, and if they're suffering from addiction, and things like that recover.
591
01:03:23.620 --> 01:03:40.449
Room: so this is a huge, huge social medical problem in in the Us. You know, for those of you who have been some of the cities down there there are huge cities of tents and people living on the street, and things like that. They have more mentally ill people in the jails than in the hospitals
592
01:03:40.590 --> 01:03:51.270
Room: erez agmoni. So they're essentially using the incarceration system to take care of the mentally ill down there. 64% of inmates have clinical mental illness symptoms, 150
593
01:03:51.330 --> 01:03:58.019
Room: 2.5 million incarcerated people, and suicide is the leading cause of death in these jails, in prisons.
594
01:03:58.350 --> 01:04:01.750
Room: and then 500,000 people are homeless.
595
01:04:01.940 --> 01:04:06.190
Room: and you know, like I said before, one of the challenges that.
596
01:04:06.420 --> 01:04:12.109
Room: Well, there's a number of challenges here. One of the challenges is that the folks that work in the jails.
597
01:04:12.280 --> 01:04:15.890
Room: They're not psychiatrists. They're not expert
598
01:04:15.950 --> 01:04:22.300
Room: clinicians. They are getting paid $18 an hour as prison guards, and they have to make
599
01:04:22.430 --> 01:04:41.040
Room: very stressful decisions every day that they're at work right? They're dealing with a very, very high stress environment. There's a lot of turnover in the staff. So even though you get somebody who becomes an expert and becomes really good in the position. Somebody just gets arrested, and they know what to deal with them, how to house them how to take care of them, so that they're safe.
600
01:04:41.810 --> 01:04:42.980
Room: and they burn out.
601
01:04:43.400 --> 01:04:44.899
Room: And then there's a new person that comes in.
602
01:04:45.930 --> 01:05:04.970
Room: So we really need a better solutions to help manage decision making in these ecosystems. And one of the challenges. And one of the opportunities here is that because so many mentally ill people are in this ecosystem. We can start to understand mental illness more objectively and find different subtypes.
603
01:05:05.550 --> 01:05:13.020
Room: And so this is an example of a company that has been built on the storyline platform called Rubicon AI,
604
01:05:13.230 --> 01:05:20.180
Room: and that's their mission is to help the the jail, and the prison systems make better decisions.
605
01:05:20.210 --> 01:05:24.189
Room: and they partnered with the National Institute jail operations to do that.
606
01:05:24.760 --> 01:05:34.980
Room: And basically what they're focused on right now is they they built a 15 min intake assessment. So police pick up somebody off of the street.
607
01:05:35.760 --> 01:05:39.399
Room: They drop them off at the desk at the jail, and they drive away.
608
01:05:39.990 --> 01:05:44.409
Room: The person sitting there behind the desk has no information about this individual.
609
01:05:44.460 --> 01:05:51.210
Room: They they don't know what substances they might be addicted to. They don't know what mental health issues they may be struggling with.
610
01:05:51.370 --> 01:05:58.550
Room: They have, but at the same time they need to bring them into a very stressful environment and make sure that the whole thing doesn't explode.
611
01:05:59.320 --> 01:06:02.809
Room: So their cat they have built a little intake assessment.
612
01:06:03.200 --> 01:06:07.839
Room: and then their first mission is to solve suicide and suicide risks
613
01:06:08.110 --> 01:06:18.970
Room: mit Ctl. And so within the first 72 h there's a reasonably high suicide risk among folks that get arrested, and they're trying to train an AI model to predict and identify people at high risk. One
614
01:06:19.040 --> 01:06:21.210
Room: and those people will get special character.
615
01:06:21.540 --> 01:06:26.190
Room: They need to validate that model, and then they need to deploy it at massive scale
616
01:06:26.660 --> 01:06:28.899
Room: for use by jail and prison staff.
617
01:06:29.480 --> 01:06:40.299
Room: So so they have got a workable solution in place. This is what their interface looks like. Storyline, sitting on the back end, capturing, analyzing the data, and then it's presented into the front end
618
01:06:40.330 --> 01:06:46.470
Room: to help to build suicide, scores, suicide, risk scores that are easy for the prison staff to interpret.
619
01:06:46.690 --> 01:06:50.060
Room: and these are active in jails right now in in Utah.
620
01:06:53.420 --> 01:06:57.890
Room: and and so i'm just gonna finish. You know my talk talking about
621
01:06:57.960 --> 01:06:59.729
Room: cancer, which is where we started?
622
01:07:02.290 --> 01:07:03.560
Room: One of my
623
01:07:03.590 --> 01:07:06.589
Room: dreams is to take this technology
624
01:07:06.760 --> 01:07:23.300
Room: and use it to support cancer patients and approve outcomes. And this is a collaboration now with the Cancer Institute, the Moffat Cancer Center, which is got a 750 million dollar injection of cash from Florida State government, so they will be one of the largest, if not the largest, cancer center in the world.
625
01:07:23.610 --> 01:07:28.030
Room: Arizona State University in the Huntsman Mental Health Center.
626
01:07:29.580 --> 01:07:41.690
Room: 1 One of the things that surprised me is I became a patient and learn more and more about the disease is that it's not our genetics that are the primary drivers and the risk for getting cancer. We think
627
01:07:42.150 --> 01:07:51.289
Room: genetic risk. Factors are actually the genetic forms of cancer pretty where we from many record one record, 2 pretty rare causes of cancer actually
628
01:07:51.760 --> 01:07:58.900
Room: overall it's actually behavioral factors. We think that play one of the biggest parts smoking
629
01:07:59.300 --> 01:08:00.259
Room: alcohol.
630
01:08:01.010 --> 01:08:11.390
Room: right, bad lifestyle, bad diet, social support, nutrition, stress, treatment, compliance, exercise, all of these components add up to affect the
631
01:08:11.520 --> 01:08:14.940
Room: risk for cancer outside later in life.
632
01:08:15.750 --> 01:08:23.579
Room: and and like I said, You know, cancer is now thought to be a kind of immuno metabolic disease. And so a lot of these environmental behavioral factors can impair
633
01:08:23.640 --> 01:08:26.439
Room: metabolism and immune pathways.
634
01:08:29.529 --> 01:08:30.840
Room: And and and
635
01:08:31.100 --> 01:08:38.229
Room: this means that this is kind of a complex problem to solve. But behavior is the problem to kind of get in and help.
636
01:08:38.590 --> 01:08:44.239
Room: and it's a place where nobody's focused. Most of the cancer research efforts are focused on the tumor
637
01:08:44.300 --> 01:08:57.459
Room: right sequencing tumors, epigenome, genome understanding, immune cell infiltration into the tumors. This is where most of the research is focused. There's a huge opportunity, I think to focus on the patient and patient behavior.
638
01:08:57.960 --> 01:09:14.680
Room: and so one of the things I've done is build this class cancer, patient master class online building your community of patients, or and providing them with the tools to understand themselves and their disease. One of the things that I found as a cancer patient is just
639
01:09:15.000 --> 01:09:19.010
Room: having an understanding that what I was doing was making things better
640
01:09:19.180 --> 01:09:21.250
Room: was incredibly motivating.
641
01:09:21.279 --> 01:09:28.069
Room: even though it's very hard to do a lot of these lifestyle changes and behavioral interventions. If you can see the results.
642
01:09:28.200 --> 01:09:31.939
Room: it's incredibly motivating, and it gives you something to work towards
643
01:09:32.120 --> 01:09:44.170
Room: the AI tools that storyline provides can help people to see the improvements. Fatigue is improved. Mental health is pretty improved. Skin power and coloration is improved.
644
01:09:44.470 --> 01:09:45.599
Room: muscle, tone.
645
01:09:45.670 --> 01:09:54.840
Room: etc., etc. So we can build these scores that help to support the patients and help to drive them towards better outcomes and better actions
646
01:09:56.160 --> 01:10:07.659
Room: at You know I I describe the extinction therapy approach, and God bless them! At the Moffat Cancer center, where they think a lot about these types of strategies.
647
01:10:07.670 --> 01:10:16.619
Room: There. they have got a donor who's gonna support a pilot trial of the extinction therapy regiment for metastatic breast cancer starting in the spring.
648
01:10:17.090 --> 01:10:27.279
Room: And so that program of drug switching and metabolic switching will be put together into an algorithm that will be delivered through storyline, and we will see if it works for other patients
649
01:10:27.640 --> 01:10:34.259
Room: and the the metabolic switching paradigm will also be tested in the trial at the hunt to be cancer in the New Year.
650
01:10:34.510 --> 01:10:43.289
Room: So that's very exciting for me, and it's a very exciting opportunity to see what types of patients do. Well, through this type of program.
651
01:10:43.650 --> 01:10:47.349
Room: what types of patients struggle with particular components
652
01:10:47.560 --> 01:10:57.940
Room: Can I diagnose and see which patients are going to struggle with particular parts early, and then provide them with special support to help them get through the whole pipeline and program
653
01:10:58.430 --> 01:11:00.500
Room: man. That'd be amazing right.
654
01:11:00.720 --> 01:11:15.530
Room: And then, if you can provide that support at scale through a smartphone ecosystem, so it's not face to face. Care providers. It can all be provided, you know, at massive mass of scale. That's a a dream. Right then you really can move the needle for a lot of people.
655
01:11:16.240 --> 01:11:17.300
Room: So
656
01:11:17.620 --> 01:11:20.959
Room: what I've tried to do today is tell you this story
657
01:11:21.170 --> 01:11:22.059
Room: of
658
01:11:22.320 --> 01:11:32.389
Room: we're building a solution for a problem that we think we've identified in the precision, research, and care world which is around a human behavior analysis.
659
01:11:32.450 --> 01:11:42.020
Room: And we feel like there's an opportunity to make that more objective and data driven and precise, and then integrated in a lot of different aspects of medicine and medical care.
660
01:11:42.510 --> 01:11:52.809
Room: And of course, the goal, as I described is really to build the platform so that other smart people can use the tools really to move very, very quickly to solve problems that they understand people.
661
01:11:53.320 --> 01:12:11.389
Room: So you know, a behavior data is incredibly rich, useful, and powerful and easy easy to capture. So you can capture that massive scale on a smartphone. This will be a front line diagnostic, right? I talked about the second line molecular lab tests very important.
662
01:12:11.500 --> 01:12:19.290
Room: but getting something out on the front lines that is massively scalable and powerful, I think, is a real opportunity.
663
01:12:19.460 --> 01:12:25.270
Room: Decisions always start from patient symptoms behavior, mental health expression, and what they kind of look like.
664
01:12:25.740 --> 01:12:29.710
Room: So there's a there's an opportunity to make that measurable and objective.
665
01:12:29.750 --> 01:12:31.769
Room: It is a predictor of outcomes.
666
01:12:32.000 --> 01:12:41.640
Room: and it's now, you know, I think, ready for prime time in in biomedical research in a lot of different areas, many of which I haven't even thought of. So
667
01:12:41.680 --> 01:12:47.649
Room: we're aiming to support this conceptual shift from just thinking about drugs and drug development
668
01:12:47.770 --> 01:12:52.159
Room: to building these more effective care algorithms and care.
669
01:12:54.010 --> 01:12:55.150
Room: And you know, I
670
01:12:55.700 --> 01:13:03.820
Room: you know, I I don't know if it's limitless. But there are really hopefully useful applications that come out of all of this work.
671
01:13:04.510 --> 01:13:09.129
Room: And if you have any questions about, what was it like to build this.
672
01:13:09.680 --> 01:13:12.120
Room: What were the problems that you encountered?
673
01:13:12.440 --> 01:13:22.030
Room: Date security, blah blah blah! Just email me and ask me, You guys are embarking on all these cool adventures to build and solve precision medicine problems.
674
01:13:22.150 --> 01:13:27.910
Room: I'd be delighted, you know, to share my experiences and war wounds. It's very hard.
675
01:13:28.070 --> 01:13:39.489
Room: very, very, very, very hard. It's like. Imagine how hard you think it is, and then multiply that by 100 it's so hard and working in the Canadian health care system has a lot of wonderful benefits.
676
01:13:39.580 --> 01:13:42.319
Room: a lot of amazing opportunities, but it's
677
01:13:42.860 --> 01:13:44.620
Room: very hard
678
01:13:44.670 --> 01:13:53.869
Room: to innovate within this system. And so you guys have to go in really tough, you know. My My brother is a surgeon here in Alberta.
679
01:13:53.890 --> 01:13:58.109
Room: and he, over the course of his career, faced many
680
01:13:58.280 --> 01:14:00.560
Room: challenges and frustrations with
681
01:14:00.680 --> 01:14:05.099
Room: wait times and difficulties, providing the care that he wanted to provide.
682
01:14:05.230 --> 01:14:19.940
Room: But he's finding ways to build private clinics, to open up entrepreneurial opportunities that increase access to care for folks in Calgary, Alberta. So you know it is going to is going to change. Just keep pushing, pushing.
683
01:14:21.510 --> 01:14:22.200
I'm done
684
01:14:22.330 --> 01:14:24.010
Room: thank you.
685
01:14:30.470 --> 01:14:32.769
Room: Sure. Sure
686
01:14:34.690 --> 01:14:36.159
a good old question.
687
01:14:38.720 --> 01:14:43.839
Room: Oh, yeah, so the starting cost is the best cost of all. It's free.
688
01:14:45.480 --> 01:15:04.420
Room: And then, as you scale up you know I think it's $99 a month, or something like that, and then 399 depending on the number of accounts, and the amount of data that you want to process. And and then, you know, if you want to grow some massive entity we're capturing enormous amounts of data. Then.
689
01:15:04.430 --> 01:15:12.629
Room: you know we have these sort of enterprise, scale solutions, or partnerships or opportunities to Bootstrap. You know everybody wants
690
01:15:12.760 --> 01:15:19.020
Room: you guys to succeed right like we we want whoever works with the platform to win. So
691
01:15:19.620 --> 01:15:20.639
Room: hopefully that's helpful.
692
01:15:20.990 --> 01:15:24.379
Room: Like lots of flexibility there.
693
01:15:24.680 --> 01:15:31.229
Room: Can this platform be beneficial and used by researchers who are building research products apps.
694
01:15:31.260 --> 01:15:37.129
Room: or is more so for business. I actually perceive it more as a researcher based
695
01:15:37.270 --> 01:15:40.610
Room: tool. I think of it very much like genome sequencing
696
01:15:41.030 --> 01:15:48.450
Room: where you know right now, everybody's sequencing staff on an aluminum sequencer. And
697
01:15:48.600 --> 01:15:51.310
Room: you're capturing all this wonderful molecular data.
698
01:15:51.420 --> 01:15:54.920
Room: Why not capture rich symptom and behavioral data
699
01:15:55.090 --> 01:15:57.879
Room: that you can intersect with that kind of multi-omics work.
700
01:15:57.900 --> 01:16:03.449
Room: So so you know, please consider it for your big research projects, consortia, all that kind of stuff.
701
01:16:03.790 --> 01:16:08.170
Room: and it's compared to most of the things that you're doing in the lab.
702
01:16:08.370 --> 01:16:14.180
Room: I run a lab so I know how much it costs. This is really cheap. It's really really cheap.
703
01:16:15.900 --> 01:16:20.440
Room: you know. I laugh because we sneeze, and we spend like a $1,000 an hour in the lab.
704
01:16:20.540 --> 01:16:23.249
Room: and the $1,000 to get you really far in storyline.
705
01:16:26.270 --> 01:16:28.279
Room: Oh, if you'd like to.
706
01:16:28.300 --> 01:16:35.239
Room: Oh, oh, good, thanks, yeah. Cost so on the website if you go to the website you can. You can sign up for free and get a
707
01:16:35.390 --> 01:16:38.619
Room: get an account right away. Any of the other costs
708
01:16:39.130 --> 01:16:48.839
Room: is the cost dependent on the services that storyline would provide. sort of to be honest right now. It's mostly a subscription based model.
709
01:16:49.060 --> 01:16:54.239
Room: so not not so much like turning features on and features off.
710
01:16:54.330 --> 01:16:56.589
Room: mostly like, Get a subscription
711
01:16:56.950 --> 01:16:58.480
Room: and you have the keys to the kingdom.
712
01:16:58.590 --> 01:16:59.190
Yeah.
713
01:17:03.690 --> 01:17:04.429
Room: yeah.
714
01:17:05.750 --> 01:17:17.670
Room: to capture more of the yeah, yeah, yeah, all of that's there. So you know, nobody is doing this on the platform, but the algorithms have been built
715
01:17:17.960 --> 01:17:20.369
Room: so so cool like
716
01:17:20.930 --> 01:17:22.199
Room: I was wonder about
717
01:17:22.470 --> 01:17:34.030
Room: people doing yoga and things like that, like if you can pick up on interesting movements. But of course there's also neurological assistance like you're evaluating Parkinson's or tremors and walking patterns, etc.
718
01:17:34.120 --> 01:17:35.099
Room: there's
719
01:17:35.550 --> 01:17:39.909
Room: It's all built. So if you want to do that, we can collect the data, and
720
01:17:40.030 --> 01:17:40.610
you know.
721
01:17:46.950 --> 01:17:49.290
Room: yeah, yeah.
722
01:17:53.670 --> 01:17:57.910
And now, how does that?
723
01:17:57.980 --> 01:18:00.260
And how does that help?
724
01:18:03.470 --> 01:18:09.799
Room: interesting? So so the imagine Fanny situation where she so
725
01:18:10.270 --> 01:18:11.090
Room: there's
726
01:18:11.400 --> 01:18:29.070
Room: imagine the future right? So Fanny situation today is, somebody comes into her clinic, and she doesn't necessarily know what's wrong with them. If it's Alzheimer's disease, it will take an hour and a half to diagnose that there'll be a series of exclusionary tests that need to be done to work through. You know
727
01:18:29.100 --> 01:18:38.369
Room: I forget. Is it vitamin d deficiencies, I forget. But there's you know there's all sorts of things you need to rule out to get down to. likely Alzheimer's diagnosis
728
01:18:38.390 --> 01:18:40.089
Room: Okay, takes her an hour.
729
01:18:40.360 --> 01:18:41.340
Room: No.
730
01:18:41.550 --> 01:18:53.389
Room: if the person could just do that assessment in 15 min, we measure 30,000 different features, and we can get a model that's 99% accurate, 95 and 90% accurate for that diagnosis.
731
01:18:53.500 --> 01:18:55.380
Room: When the patient comes into the clinic.
732
01:18:55.430 --> 01:19:06.830
Room: Fanny's way ahead of time, right? And so she doesn't need to spend necessarily an hour and a half on that problem anymore. She might move straight to second line tests blood test for amyloid
733
01:19:07.780 --> 01:19:12.779
Room: order scans. You know what whatever she she decides is the next course of action.
734
01:19:13.690 --> 01:19:16.669
Room: There's the time-saving on her part
735
01:19:17.660 --> 01:19:18.590
Room: Now
736
01:19:19.430 --> 01:19:20.230
Room: the
737
01:19:20.740 --> 01:19:25.529
Room: so this is the clash of the Canadian versus the Us. Health care system
738
01:19:25.780 --> 01:19:29.980
Room: in the Canadian health care system. You want improved efficiency
739
01:19:30.340 --> 01:19:31.450
Room: because.
740
01:19:31.940 --> 01:19:40.890
Room: having many, many more patients come to you, just clogs the system and causes a lot of problems. So you want to improve the efficiency, and you want to get more patients through effectively and safely
741
01:19:41.360 --> 01:19:43.479
Room: in the American health care system.
742
01:19:43.780 --> 01:19:47.929
Room: You're a business. You want more patients to come to. You, you know.
743
01:19:48.070 --> 01:19:56.249
Room: worried that they have neurological health problems. So if you can get early diagnostics out there on the market that help to capture
744
01:19:56.600 --> 01:19:58.980
Alzheimer's.
745
01:20:00.030 --> 01:20:09.649
Room: You're pulling them into your ecosystem where you can run $3,000 MRI Scans, etc. Etc. You know what I mean. So yeah, so that you know
746
01:20:10.130 --> 01:20:12.020
Room: 2 different worlds, right? Yeah.
747
01:20:14.370 --> 01:20:26.089
Room: it's a completely different right? Yeah, yeah. So you can have these clear approved tests that you get out in the world. And the idea is just to help to inform and get early diagnosis for all kinds of things.
748
01:20:26.310 --> 01:20:26.929
Yeah.
749
01:20:27.410 --> 01:20:27.969
yeah.
750
01:20:29.400 --> 01:20:31.529
How are you guys?
751
01:20:31.730 --> 01:20:34.599
I get your your your diagnosis.
752
01:20:34.760 --> 01:20:35.820
There's the only thing
753
01:20:36.300 --> 01:20:38.509
there is any differences like,
754
01:20:39.510 --> 01:20:40.549
how are you? Guys?
755
01:20:42.890 --> 01:20:54.130
Room: Yeah. Yeah. So the way that process works it's kind of the same in any kind of test that one's developing in medicine. It's always starts with an expert
756
01:20:54.220 --> 01:21:07.369
Room: clinician, and typically you, if you can, you get multiple expert clinicians. So somebody comes in and they've got a diagnosis of depression, and you might get 3, 4, 5 psychiatrists that say.
757
01:21:07.660 --> 01:21:14.640
Room: I agree. You know this person is not this. It's not that they have depression, so that person gets accurately labeled.
758
01:21:15.320 --> 01:21:20.359
Room: and this is why it's so expensive and challenging to build these algorithms.
759
01:21:20.620 --> 01:21:25.730
Room: Now, you've got a ground truth data set of folks that have real.
760
01:21:26.430 --> 01:21:28.429
Room: Everybody agrees on depression.
761
01:21:29.400 --> 01:21:32.699
Room: And your first job is to
762
01:21:32.940 --> 01:21:36.370
Room: build an algorithm that matches those clinical experts
763
01:21:36.520 --> 01:21:44.470
Room: and gets a certain accuracy and generalization error that's acceptable for for that. Does that make sense? So that's where it starts
764
01:21:44.620 --> 01:21:45.559
Room: Now
765
01:21:45.670 --> 01:21:53.560
Room: The reason i'm excited about unsupervised learning is that let's say depression is a fairly broad diagnosis. You throw a whole bunch of people into that bucket
766
01:21:53.740 --> 01:22:05.610
Room: mit ctl. And and then you apply an unsupervised learning algorithm that clusters them out into subgroups that becomes an exciting to research paradigm right where you've got these new subgroups, and you follow them over time, and maybe one group. One
767
01:22:06.160 --> 01:22:14.270
Room: responds very well to a particular dose of the particular Ssri and the other group is treatment resistant to everything until you give them psilocybin.
768
01:22:14.870 --> 01:22:16.349
Room: and then you know
769
01:22:16.730 --> 01:22:18.540
Room: that's the solution for them
770
01:22:18.560 --> 01:22:21.900
Room: or it's, you know. Hopefully not, but ect whatever.
771
01:22:22.010 --> 01:22:22.610
Room: Yeah.
772
01:22:31.470 --> 01:22:37.809
Room: is the overall anonymized data within storyline available to play with
773
01:22:38.090 --> 01:22:46.719
Room: and overlay, to identify insights and opportunities. So, Krista. Great question. Yes, yes, I can do that.
774
01:22:48.100 --> 01:22:53.290
Room: I understand. You know. I'm a scientist. You want to play with it right? See what it's good for.
775
01:22:53.840 --> 01:22:59.630
Room: Is it possible to use data generated by others in the program? Yeah. So that's the solution for the Api thing? Where
776
01:23:00.810 --> 01:23:08.129
Room: there there are now these Api ecosystems. And if you're gathering data on an apple watch in your study, or you're gathering
777
01:23:08.240 --> 01:23:11.879
Room: fitbit data, or you know, rings and things like that.
778
01:23:13.050 --> 01:23:16.740
Room: you can pull that into the ecosystem through through Api calls.
779
01:23:17.300 --> 01:23:17.910
Yeah.
780
01:23:19.560 --> 01:23:22.409
Room: Awesome questions. You guys are
781
01:23:22.690 --> 01:23:23.530
Room: You're on it
782
01:23:23.610 --> 01:23:24.610
Room: pretty cool.
783
01:23:25.770 --> 01:23:28.440
Room: We have just a few minutes like, oh, go ahead. I'll check.
784
01:23:29.640 --> 01:23:30.760
Yeah.
785
01:23:33.310 --> 01:23:34.380
Yeah.
786
01:23:37.290 --> 01:23:37.990
yes.
787
01:23:38.950 --> 01:23:41.550
Room: I can hear you. It's okay, and I
788
01:23:42.080 --> 01:23:42.660
right?
789
01:23:43.430 --> 01:23:46.220
So about that. That's Why, i'm supervised learning
790
01:23:46.670 --> 01:23:48.099
how you
791
01:23:48.610 --> 01:23:49.530
because you're added it.
792
01:23:50.960 --> 01:23:51.780
So
793
01:23:51.890 --> 01:23:52.809
I want to
794
01:23:53.300 --> 01:23:55.180
kind of was curious about what kind of
795
01:23:55.650 --> 01:23:56.990
features who are just a
796
01:23:57.730 --> 01:23:58.389
paper.
797
01:23:58.910 --> 01:24:01.039
Room: No, there is not a paper.
798
01:24:02.670 --> 01:24:08.720
Room: I'm: the bottleneck on writing that paper. so yeah, we have not
799
01:24:08.970 --> 01:24:15.579
Room: published the algorithm that that does that yet. I want Jared to test it in a few different ways before it's like ready for.
800
01:24:16.660 --> 01:24:17.449
But
801
01:24:17.530 --> 01:24:18.970
Room: yeah, and so you
802
01:24:22.010 --> 01:24:22.639
about it.
803
01:24:26.970 --> 01:24:30.389
No.
804
01:24:39.180 --> 01:24:40.009
I really wanna
805
01:24:45.000 --> 01:24:45.590
one.
806
01:24:53.080 --> 01:24:54.559
so it's a lot of
807
01:24:55.960 --> 01:24:59.130
the next.
808
01:25:02.190 --> 01:25:03.150
It should be pretty
809
01:25:03.350 --> 01:25:03.920
very excited.
810
01:25:04.470 --> 01:25:05.030
Okay.
811
01:25:05.200 --> 01:25:06.750
Room: some great question.
812
01:25:10.350 --> 01:25:15.140
Room: Can I talk one question on you?
813
01:25:16.740 --> 01:25:18.030
Room: Can you hear me now?
814
01:25:20.230 --> 01:25:21.639
Room: Okay, okay.
815
01:25:22.200 --> 01:25:23.130
Room: All right for that.
816
01:25:23.930 --> 01:25:27.180
Okay, I'll just ask one question. I
817
01:25:34.640 --> 01:25:36.790
kind of
818
01:25:41.080 --> 01:25:44.270
You've described this care.
819
01:25:44.490 --> 01:25:53.009
Room: And so I I wonder if you have a vision of like, you know, using this as an application to allow, you know individuals to get in there and
820
01:25:53.120 --> 01:25:58.909
Room: customize their own journey. I their own plan.
821
01:25:59.060 --> 01:26:00.130
Room: Yeah.
822
01:26:00.230 --> 01:26:13.460
Room: is this
823
01:26:13.770 --> 01:26:15.629
no changes or whatever?
824
01:26:16.620 --> 01:26:17.580
So
825
01:26:18.670 --> 01:26:26.480
this is huge.
826
01:26:40.470 --> 01:26:41.790
2 guys
827
01:26:42.150 --> 01:26:42.990
no
828
01:26:43.720 --> 01:26:49.120
Room: showing up to the
829
01:26:49.630 --> 01:26:50.570
Room: So
830
01:26:50.590 --> 01:26:52.350
Room: it's because they
831
01:26:52.740 --> 01:27:05.689
Room: And so
832
01:27:06.820 --> 01:27:08.199
Room: I think we
833
01:27:08.230 --> 01:27:09.160
Room: really
834
01:27:18.780 --> 01:27:19.809
Room: sorry.
835
01:27:23.760 --> 01:27:29.479
Room: Well, we're at 7 30. So.
836
01:27:29.600 --> 01:27:32.960
Room: guys, thank you very much for your time.
837
01:27:38.910 --> 01:27:41.969
Room: It yeah, thanks to everybody online and sorry I,
838
01:27:42.330 --> 01:27:46.989
Room: you know, came in, and I will reach out if you have any questions about anything.
839
01:28:07.800 --> 01:28:11.049
Room: Thanks so much for this opportunity.
840
01:28:14.360 --> 01:28:16.269
I signed up for the.
