Chris Gregg PhD, Storyline’s CTO discussing why behavioral data is the richest, most predictive data there is, and why it's better than current tools like genome, epigenome, omics, and lab tests.

Transcript:

Welcome to the storyline, where our mission is to understand human behavior and make that knowledge useful for everyone.

Today, we're going to talk about why that's so important for researchers and physician scientists. And this really comes down to the fact that behavior is the only data that can make massively scalable precision medicine possible.

If we're talking about improving the care for millions of people, there really aren't that many choices. When we think about genomics, which is an area where that is a major area of focus in the precision medicine world. One of the problems is that the genome sequence fundamentally does not change.

If you have an infection, you get sick, you change your diet or whatever is happening to you in your life. Your genome sequence doesn't change, so it is not a metric of changes to your health or status over time.

The genome sequence that you're born with is the same sequence that you die with epigenome data, where changes to chromatin marks on the genome sequence change in response to our diet or stress or illnesses are important, but they're not accessible.

You can't go in with a needle and biopsy someone's brain or someone's kidney or liver to measure changes to the epigenome over time. Weekly, monthly for millions of people, just not possible.

The microbiome is an exciting new area that does have important implications for our health, but it's difficult to access and it's expensive to profile.

You can't be profiling the microbiome of millions and millions of people every week or every month where they're trying to harvest poop and put it into the mail in these tubes and send it off to companies to get profiled and sequenced.

And it's only part of the picture anyway. So none of these platforms that we have currently solve the unmet need for enabling precision medicine for millions and millions of people.

Wet lab tests fundamentally really are not meant for that. They're not massively scalable for remote health assessments, for such large numbers of folks. They're really second line diagnostic tests. For as long as medicine has been around, we've known that patient behaviors, feelings, expressions, cognitive changes and symptoms are really the richest and most useful and predictive information in the medical world.

It's really the primary source of information that all providers use for every care decision. And of course, this makes sense because the human nervous system over 500 million years of evolution has evolved to innervate all the different organs down at the cellular level and detect changes to metabolism, inflammation, endocrine signals.

And then all of these pieces of information are rolling up through the peripheral nervous system into the central nervous system to affect mood, cognition, decision making processes, homeostasis, et cetera. And so behavior really is the ultimate, evolution's ultimate solution to precision medicine and modern monitoring, all of the different things that are going on in the body. And it is, you know, the richest, most useful and predictive data.

We think the problem is that behavior analysis currently in the research ecosystem is vastly oversimplified and subjective.

Typically, we're talking about patient reported outcomes or simplified questionnaires, and there really has been no objective, precise and trusted way to understand what seems to be infinitely variable and flexible, this very, very complex data that comes out of analyzing human behavior. But we have been able to change that, and we've built a platform that makes behavioral analysis rich, precise, objective and data driven, and it all operates through people's smartphones.

And that's why it's so incredibly scalable. 70% of people on the planet have access to a smartphone, and so you can reach them and gather data and collect data and deliver care. So storyline is a smartphone based platform where people perform interviews in response to different questions or tests, and then their responses are video recorded, moved into the cloud where they're safely stored, and then we activate a microservices pipeline of artificial intelligence algorithms that extract thousands and thousands of measures.

Right now, we've got a pipeline that measures over 20,000 different micro features.

From people's responses to different questions. And in the video data, we can capture many, many, many important and interesting things like pupil dilation, eye tracking, head movement, changes to blood, perfusion patterns across the face, respiration patterns, micro-expressions, movement patterns to different parts of the face. That can be, for example, very useful in neurological tests where you're trying to map damage to cranial nerves, things like that.

We analyze speech patterns, so there's thousands and thousands of measures that are extrapolated around what people say their word choice, how they articulate their responses, sentence structure, repetitive patterns, personality traits, the sentiment that they express around what they're feeling. And it's not just what they say, but how they say it.

So we also capture nearly 10,000 different measures of vocal patterns to look at changes to micro tremors, changes to emotion and expression in the voice, pitch tone, pronunciation syllables, speech rate, etc., etc., et cetera. So this creates a massive, massive data space for discovery and objective analysis of patient behavior symptoms, cognitive patterns, neurological effects, et cetera so we used to have a patient come into the clinic and we'd be able to take a history.

Much of that information would actually be lost. It was just an oral kind of interview and discussion. There would be medical records, it would be a physical exam. And these sorts of things, of course. But there's massive amounts of information that we've never been able to get at objectively and precisely.

And now we can. Things like attention, expression, expression of pain, pain, responses, discomfort, stress, trauma, cognitive functioning, motor patterns, skin health, sympathetic tone. All of these things are suddenly now accessible for measure and analysis. And so storyline, behavioral AI is this objective, sensitive, quantitative and precise platform for measuring the space that has been so subjective and difficult to use in the past. And now there's kind of this potential for a new field that we can call behavior omics, where we're leaping off of the ideas of the microbiome and the epigenome and the genome.

We've got now these behavioral homes where we can use storyline, Deep I phenotyping of patient behavior and expression and symptoms all through a smartphone at massive scale, giving everybody access. This really is the only platform for massively scalable precision medicine and research. It's the only way we're going to do it.

And now we have a framework to understand behavior, just like genetics and genomics, where we had the central dogma, biology of DNA makes RNA makes protein, and that provides us these mechanistic models of biology and health and disease.

What storyline deep phenotyping enables us to do is have a kind of unified theory of behavior where we start off with these micro features, where we measure Tens of thousands of these over time

. Micro features can roll up into modules of behavioral sequences and types, subtypes of expressions by patients, and these build up into greater behavioral phenotypes. In that layered structure, that hierarchical structure provides us a mechanistic way of understanding brain function, behavior, health and disease and linking that to biologically valid underlying mechanisms.

And this gives us we can, again, kind of learn from the genetics field and build a better approach to understanding behavioral pathology, where, you know, for example, in the genetics field, we would look at mutation burden and have this liability threshold where you tip into a disease state and have a high risk for Disease based on the number of mutations in the genome.

And we can do something very analogous now in behavior where we can measure thousands and thousands of micro features and look at how many of these features seem to be affected and abnormal. And this builds a liability threshold for abnormal behavior that's indicative of an underlying health problem mental illness, neurological disorder, cognitive disorder, et cetera.

And so these new behavioral thresholds, liability thresholds, become very, very useful and something that we never had access to before.

Storyline is operating now in a completely new space for biomedical research, where we've had things that were cheap but not very diagnostic, for example, like patient reported outcomes, which are notoriously subjective and inaccurate but very cheap and scalable.

And now storyline is cheap and scalable but objective, data driven and precise.

And so this can be used essentially to replace that, to understand behavior and symptomology. It's a whole new area for data capture and value creation. As a researcher and a scientist, that's really exciting for me because it's a new opportunity to make discoveries, both conceptual discoveries and technical discoveries, build new software, make new advances, and create new fields.

From a health care perspective. It's very important because now we have a whole new precision medicine platform that will improve patient subtype thing, building biomarkers of different illnesses that would predictive models for drug responses, predicting outcomes, monitoring patients at massive scale remotely through their smartphone to provide better care and ideally get ahead of problems so that you can steer the patient ahead rather than reacting to problems after they've already manifested and then trying to correct correct the damage.

So in summary, this is why storyline behavioral eye data really is the only data that you absolutely must have if you're doing a clinical study or you're doing other types of research.

You need a deep understanding of the behavior and symptomology of the patients that are going into your study.

Whether you're doing genomics, you're doing microbiome, some sort of epigenome test, immune test, you know, inflammatory markers, etc., et cetera. Having deep behavioral, objective behavioral and symptom data that can be ultimately scaled is really essential.

And now you have the opportunity to get access to that.

Thank you.

Christopher Gregg PhD is Associate Professor Neurobiology & Human Genetics University of Utah, Co-founder and CSO at Storyline Health