Predicting and modulating personal immunity and health

Transcript

• 00:00I'm gonna,
• 00:01kick it off partly because,
• 00:03I wanna sort of motivate
• 00:04some of the themes,
• 00:06and and directions of the
• 00:07CSCI that Nancy briefly mentioned.
• 00:09So, so I am a
• 00:11a faculty in immuno biology
• 00:13also with a secondary appointment,
• 00:15in biomedical engineering. And, I'm
• 00:17also an investigator of the
• 00:19new Chan Zuckerberg Biohub New
• 00:20York.
• 00:22So,
• 00:23first, I wanna talk a
• 00:24bit about our work, and
• 00:27then using that as a
• 00:28as a as a basis
• 00:29to motivate
• 00:30the directions and themes, at
• 00:32the CSCI.
• 00:33So one of the things
• 00:34we study in in in
• 00:35my lab, it's variation,
• 00:37in the human immune system.
• 00:38And it it's really a
• 00:40hallmark in the sense that
• 00:41you see that in different
• 00:42contexts.
• 00:43If in in for example,
• 00:45in terms of response to
• 00:46immunotherapy,
• 00:47you typically see a subset
• 00:48of patients respond,
• 00:50and you see that in
• 00:51vaccination and infection responses. For
• 00:53example, here, I'm showing,
• 00:55the is there a point
• 00:56zero?
• 00:57Of
• 01:00the mouse? Okay. Yeah. So
• 01:02so, yeah, you see extensive
• 01:04variation because of age, but
• 01:06also even within young individuals,
• 01:07you see these hundreds of
• 01:08folds of differences in antibody
• 01:10response
• 01:10to vaccination.
• 01:12And you also see,
• 01:14in autoimmunity
• 01:15in terms of both if
• 01:17somebody has a higher or
• 01:18lower risk for developing the
• 01:19disease and also in terms
• 01:20of having disease activities over
• 01:22time, you see very distinct
• 01:23patterns and trajectories within individuals.
• 01:26And, actually, many of you
• 01:27are wondering about questions about
• 01:29ourselves also. Right? But, actually,
• 01:30no general tools exist today
• 01:32to answer questions like, did
• 01:33I respond well to my
• 01:35last vaccine? Am I gonna
• 01:36develop autoimmunity in the future?
• 01:38What about allergy?
• 01:39And can my future health
• 01:40trajectory be predicted based on
• 01:41the status of of my
• 01:42immune system? And how well
• 01:44will I will I be
• 01:45responding to an intervention? Right?
• 01:47So these are questions that
• 01:48we actually don't have generalized
• 01:50tools right now to answer.
• 01:52So
• 01:53and so why do we
• 01:54have such variable responses? And
• 01:56and various factors, genetics and
• 01:58environmental exposure history,
• 02:00together shape,
• 02:01what we call a personal
• 02:03immune state within individuals.
• 02:05And these kinds of states
• 02:06can be actually quite temporally
• 02:08stable. So we often through
• 02:09years of looking at humans
• 02:11and the immune system, we
• 02:12actually often see a situation
• 02:14like here,
• 02:16where you see that two
• 02:17two individuals would have distinct
• 02:19set point states.
• 02:21And a key question is
• 02:22when you see something like
• 02:23this, when you look at,
• 02:24individuals in a population, it's
• 02:25whether these could result,
• 02:27right, in differences in some
• 02:29of the outcomes I mentioned.
• 02:30Auto d c autoimmune disease
• 02:32risk,
• 02:33disease activities, and how they
• 02:34respond, and so on.
• 02:36So, a number of years
• 02:37ago,
• 02:38that maybe more than a
• 02:39decade now, my lab published
• 02:40a study
• 02:41showing that, independent,
• 02:43we measure sort of the
• 02:44immune system fairly broadly at
• 02:46the time and show that
• 02:47independent of factors like age,
• 02:49sex,
• 02:50and preexisting immunity to influenza,
• 02:53the response to the influenza
• 02:54vaccine
• 02:55can be predicted based on
• 02:56the baseline state of the
• 02:58individual.
• 02:59And later on, we've used
• 03:01single cell technologies to sort
• 03:03of unravel the sort of
• 03:04the cellular basis of those
• 03:05signatures
• 03:06and also show using some
• 03:07of the data, collected by
• 03:09Hib c one
• 03:10that this could actually be
• 03:12be be generalized to other
• 03:13other cohorts and seasons,
• 03:15and even extended to, a
• 03:17vaccine that's very different from
• 03:18flu, in this case, yellow
• 03:19fever, which is a live,
• 03:20attenuated infection.
• 03:22And and further on, we
• 03:23show that this kind of,
• 03:25baseline set points,
• 03:26is relevant in the context
• 03:27of autoimmunities.
• 03:29So shown here is a
• 03:29lupus patient
• 03:31having varying disease activities over
• 03:34time, a single patient. And
• 03:35you can see that during
• 03:36some periods, this patient has
• 03:38high disease activity while others,
• 03:40this patient was relatively
• 03:42clinically quiescent.
• 03:44So we show that by
• 03:45looking at this specific baseline
• 03:47immune set point during clinical
• 03:48quiescence,
• 03:49it actually provides predictive information
• 03:51about the extent and the
• 03:52intensity of the flares in
• 03:54a subset of patients whose
• 03:55flares resemble
• 03:57how a healthy individual is
• 03:58responding to a vaccine in
• 03:59this case. So lupus is
• 04:01very heterogeneous. For other patients,
• 04:03we actually saw that this
• 04:03doesn't apply. So it points
• 04:05out to the points to
• 04:06the very specific nature of
• 04:08these set point variables.
• 04:10And then we as I
• 04:11mentioned, we use single cell
• 04:12technology to dissect the cellular
• 04:14basis of this and uncover
• 04:15a circuit involving multiple cell
• 04:17types. And I won't go
• 04:18into the details because this
• 04:19was published, but I wanna
• 04:21mention one, point that's important
• 04:23is that once we sort
• 04:24of figure out the cellular
• 04:25basis and what are the
• 04:26important states of cells to
• 04:28measure,
• 04:29we're actually able to reduce
• 04:30this down to ten parameters
• 04:32that we now can measure
• 04:33in blood. And and now
• 04:34we can start to monitor
• 04:36people using these ten parameters
• 04:37that are quite easy to
• 04:38measure and start to look
• 04:39at them longitudinally,
• 04:41over time and and in
• 04:42different populations.
• 04:44Now what about beyond classic
• 04:45immunity of infection and and
• 04:47and and vaccination? Right? So
• 04:49immune cells, as many of
• 04:50you know, they actually circulate
• 04:52around our body, many of
• 04:53them, and they can sense
• 04:54sort of deviation from homeostasis.
• 04:57And therefore, they actually are
• 04:58collecting
• 04:59and processing a lot of
• 05:00information about health and disease
• 05:02and within our bodies. And
• 05:04so given this fact, it's
• 05:05not surprise surprising to to
• 05:07sort of see that the
• 05:08immune system basically,
• 05:10to my view, involving basically
• 05:12all, disease and conditions. So
• 05:14you can see that in
• 05:15neurodegeneration,
• 05:16in pain,
• 05:17in in definitely in aging,
• 05:19and metabolic health.
• 05:21And so so the the
• 05:22key question then boils down
• 05:23to what exactly is immune
• 05:25health. Can we actually measure,
• 05:27the health of the immune
• 05:28system and then using that
• 05:29to link to specific outcomes
• 05:31and understand how the immune
• 05:32system orchestrate,
• 05:34proper responses?
• 05:35So to start to look
• 05:36at that, a number of
• 05:37years ago, we worked with
• 05:39colleagues at the NIH,
• 05:40who study various,
• 05:42rare monogenic diseases. So we
• 05:44convinced them to sort of
• 05:45come together
• 05:46and, look at all of
• 05:48those patients that hit diverse
• 05:50genetic pathways and then measure
• 05:52them using longitudinally in some
• 05:54of them using the same
• 05:55systems immunology tools and also
• 05:57matching,
• 05:59clinically healthy individuals. So a
• 06:01key question we ask is,
• 06:02given this these diverse conditions,
• 06:04can we actually detect and
• 06:06and and and sort of
• 06:07sense get these common deviation
• 06:08from health by looking across
• 06:10all these,
• 06:11conditions?
• 06:12So in terms of data,
• 06:14we collected a variety of,
• 06:16immune profiling and omics data,
• 06:17including circulating proteins,
• 06:19cell type specific gene expression
• 06:21profiles, and more recently also,
• 06:23a subset of patients, we
• 06:24have collected single cell data.
• 06:26And by looking at actually
• 06:27these people, some of these
• 06:28people over time, it was
• 06:30again, the same picture emerges
• 06:32despite the high penetrance and
• 06:34the high effect size of
• 06:35some of these DCs, how
• 06:36they affect health. The individual,
• 06:38it's typically the unit, actually.
• 06:40They if you look at
• 06:40the immune system as a
• 06:41whole, these individuals actually don't
• 06:43cluster necessarily into patient groups.
• 06:45So based on that, we
• 06:46sort of decided to sort
• 06:48of take two machine learning
• 06:49approaches to look at the
• 06:50data. On the left hand
• 06:51side, I'm showing you an
• 06:52approach where we look,
• 06:55look at each individual as
• 06:56a vector of numbers, basically
• 06:57a vector of their phenotypic
• 06:58profiles, molecular cellular profiles, And
• 07:01then we try to compute
• 07:01the relationships among these individuals
• 07:03and try to ask what
• 07:05kind of variation exists in
• 07:06these multimodal data. On the
• 07:08right, we're using a more
• 07:10supervised approach and ask, can
• 07:11we compute a probabilistic
• 07:13measure of whether somebody is
• 07:15healthy or not? Right? So
• 07:17based on so for that,
• 07:18we know the label of
• 07:19of everyone. So on the
• 07:20left side, we actually don't
• 07:21have the label of anybody.
• 07:22We don't know what disease
• 07:23they have and so on.
• 07:24Right? So the surprise was
• 07:26that when you look at
• 07:27the emergent
• 07:29axis of variation,
• 07:31coming out from the left
• 07:32approach, unsupervised
• 07:34approach,
• 07:35it came out with exactly
• 07:36the, same
• 07:37principle axis of variation
• 07:39as this probability of being
• 07:41healthy, basically. So in other
• 07:42words, even if you don't
• 07:44know who's healthy and who
• 07:46has what disease, there's natural
• 07:48variation in the population that
• 07:49resembles precisely this probabilistic
• 07:52measure of of being healthy.
• 07:53So in other words, we're
• 07:54learning immune health actually by
• 07:56learning from a lot of
• 07:56pathologies. Right? What are the
• 07:58those are negative examples, so
• 07:59to speak, of of what
• 08:00it means to be having
• 08:01a healthy immune system.
• 08:03So and then the question
• 08:04is, what does it mean?
• 08:05So when you plot, these
• 08:07individuals, so each dot is
• 08:09a person here, and each
• 08:11row,
• 08:12corresponds to a monogenic
• 08:14disease,
• 08:15group, you can see that's
• 08:16the clinically healthy group at
• 08:17the top, they span a
• 08:19huge range, and some of
• 08:20them actually extends well into,
• 08:22the the the sick patients.
• 08:24Right? So as I mentioned,
• 08:25what does it really mean,
• 08:27when you have such a
• 08:28wide range of of of
• 08:29this metric, basically?
• 08:31So then we look at
• 08:32this in an independent set
• 08:33of people.
• 08:34This is the Baltimore Healthy
• 08:35Aging cohort where they collected,
• 08:37data
• 08:38from, individuals across each decade
• 08:41of life. And what you
• 08:42can see when you compute
• 08:43this metric on these people
• 08:44is that there's a decline
• 08:46in the immune health metric
• 08:47as you go,
• 08:48get older.
• 08:49But, so suggesting that this
• 08:51is actually tracking healthy aging.
• 08:53But you can also see
• 08:54that on the y axis,
• 08:56there's significant remaining variation.
• 08:59So given a a sixty
• 09:00or seventy year old, there's
• 09:01actually a a still a
• 09:02y variation
• 09:03in in according to this
• 09:04health metric. So we look
• 09:06into that a bit, and
• 09:07and that actually can predict
• 09:08vaccine responses within the especially
• 09:10within the elderly.
• 09:12It also, correlates with with
• 09:14variables like BMI, for example.
• 09:16And then it also tracks,
• 09:18in other context, like whether
• 09:19the heart is working well
• 09:20and also tracking drug responses,
• 09:22in this case, RA patients
• 09:24responding or not to, a
• 09:26therapy, basically.
• 09:27So the picture I wanted
• 09:29to paint is that, of
• 09:30course, this is a one
• 09:31dimensional measure that we uncover.
• 09:33There are multiple dimension that
• 09:34one can uncover, especially one
• 09:36specific to disease.
• 09:37But this one seems to
• 09:38be emergent
• 09:39independent of whether we look
• 09:40at we can drop out
• 09:41different patients. We can even
• 09:43drop out all the healthy
• 09:44individuals in the cohort. We
• 09:45still get the same
• 09:48metric
• 09:49back, basically.
• 09:52This is It's a public
• 09:53service.
• 09:56Stop. My time is up.
• 09:57Right? Okay.
• 10:00Okay.
• 10:02So the
• 10:03the the concept I'm getting
• 10:05to is that, often when
• 10:06we look at individuals in
• 10:07the in the human population,
• 10:09there are obviously some of
• 10:10us who are maybe really
• 10:12at this end of healthy
• 10:13healthy,
• 10:14and then there are ones
• 10:15that we know with disease
• 10:16or pathology
• 10:17that are far off from
• 10:18the optimal,
• 10:19set point. But most of
• 10:21us, I think, are still
• 10:22healthy clinically, but inching towards
• 10:24some sort of pathology. Right?
• 10:26And so by by developing
• 10:27this kind of immune health
• 10:29metric and using the immune
• 10:30system as a sensor, I
• 10:31think we can start to
• 10:32really start to quantify,
• 10:34where someone may be at
• 10:36and so on. So that's
• 10:37one of the key
• 10:38challenges that we would like
• 10:39to tackle,
• 10:40both in my lab and
• 10:41also working together with other
• 10:43others in the in the
• 10:44CSCI.
• 10:45So so this concludes the
• 10:46part of the talk for
• 10:48talking a little bit about,
• 10:50science from my lab and
• 10:51and and how so now
• 10:52I'm gonna move sort of
• 10:53onto
• 10:54motivating
• 10:55some of the, the key
• 10:56themes at the CHEI. So
• 10:58I hope I've convinced you
• 10:59that,
• 11:00with the immune system and
• 11:02what it's sensing,
• 11:03we can really move beyond
• 11:05the genome, which in a
• 11:06way, it's a static information
• 11:08store of of what may
• 11:09be possible.
• 11:10But but the immune system
• 11:12is really sensing, detecting what's
• 11:14going on right now and
• 11:15and in the past as
• 11:16well and integrating that information.
• 11:18And it's it's it's I
• 11:19I would argue that that's
• 11:20why it's actually gonna tell
• 11:21us a lot more about
• 11:23both the status of the
• 11:24of the body right now,
• 11:25but also what's gonna happen
• 11:27in the future. Right? So,
• 11:28therefore, the the grand challenge
• 11:29is can we uncover the
• 11:30connection between the immune system
• 11:32and physiology
• 11:33and thereby predict and modulate
• 11:35personal immunity and health.
• 11:37So at the c h
• 11:38CSCI,
• 11:39we
• 11:40help to develop and bring
• 11:41together both people and interdisciplinary
• 11:45approaches to monitor, predict, understand,
• 11:48and modulate personal immunity and
• 11:50health for the benefit of
• 11:51all.
• 11:53So to achieve this mission,
• 11:54we have some philosophies and
• 11:55and and and and and
• 11:57and pillars.
• 11:58So one is that,
• 12:00many of you heard of
• 12:01systems immunology,
• 12:03and some of you may
• 12:04think of it as immunology
• 12:06with lots of omics data.
• 12:07But I wanna argue that
• 12:08it it's more than that,
• 12:10actually.
• 12:10It it I think thinking
• 12:12the cyst immune system is
• 12:13a system,
• 12:14and then and then designing
• 12:16unique studies and and and
• 12:17and and questions to address.
• 12:19For example, how
• 12:21do immune cells and also
• 12:23going up the hierarchy, multiple,
• 12:25populations of cells integrate information?
• 12:28And can we understand how
• 12:29the interactions among these components
• 12:31actually give rise to outcomes?
• 12:32Right? So that's a grand
• 12:33challenge in understanding something as
• 12:35complex as an immune system.
• 12:36And to address that, I
• 12:37think we need to think
• 12:39of unique approaches and and
• 12:40and and and study designs.
• 12:42And then the second aspect
• 12:43is people.
• 12:45We we want to enable
• 12:46and cultivate.
• 12:47So we wanna collaborate with
• 12:49at Yale, there's a lot
• 12:50of exciting,
• 12:52groups and also, programs such
• 12:54as HTI, YCIO, and the
• 12:56engineering school. We wanna bring
• 12:57together across Yale and the
• 12:59and also the world to
• 13:00push the frontiers of this
• 13:02field,
• 13:03and and and and and
• 13:04really,
• 13:05drive forward also this predictive
• 13:07immune cell engineering,
• 13:09theme that I'm gonna highlight
• 13:10and again for the benefit
• 13:11of all. And in terms
• 13:13of people, also education and
• 13:14training is another mission. We
• 13:16wanna develop a new generation
• 13:17of computational and systems immunologists,
• 13:19immune cell engineers,
• 13:21and technicians and engage citizens
• 13:23in research.
• 13:24And then on the science
• 13:26side, I'm listing sort of
• 13:27four pillars.
• 13:29Pillar one, I think I
• 13:30mentioned that already. It's it's
• 13:31don't forget physiology. Sometimes we
• 13:33get too hung up with
• 13:34measuring just the immune system,
• 13:36obviously, but I think we
• 13:37have to measure physiology at
• 13:38the same time. And second
• 13:40is utilize human immune variation.
• 13:42So human immune variation is
• 13:43often sometimes thought of as
• 13:45a as a as a
• 13:46as a as a as
• 13:46a roadblock.
• 13:47But in a way, it's
• 13:48the it's the only way
• 13:49to actually utilize natural variations,
• 13:52and it's a human it's
• 13:53an experiment, natural experiment that
• 13:54we can utilize and learn
• 13:56a lot using these new
• 13:57technologies.
• 13:58The pillar two is measure.
• 14:00Right? We knew new technologies,
• 14:03and and and so, therefore,
• 14:04the application and fine tuning
• 14:06and development of cutting edge
• 14:07immune monitoring and molecular profiling
• 14:09approaches
• 14:10are are central.
• 14:11And then the other pillar,
• 14:13it sells as as as
• 14:14tools.
• 14:15So I I mentioned briefly
• 14:16on how immune cells, some
• 14:17of them can traffic and
• 14:19can sense information. So how
• 14:20can we harness them as
• 14:21both natural sensors? Because some
• 14:23of them already are sensing
• 14:25things, but we don't have
• 14:26the full book to decode
• 14:27what it means. For example,
• 14:29I can see a certain
• 14:30T cell in the blood
• 14:31and and may may have
• 14:32information about the liver, about
• 14:34the kidney that we haven't
• 14:35been able to decode those
• 14:36yet. So that's what I
• 14:37call natural sensing. The other
• 14:39one, it's it's more ambitious,
• 14:40but at the same time,
• 14:40much more controllable. Right? Can
• 14:42we engineer these cells
• 14:44so that they go to
• 14:44specific locations, actually collect data
• 14:46for us? That's one. And
• 14:48second is one is once
• 14:49those data are collected, can
• 14:50we act? And third, it's
• 14:52eventually, can you actually make
• 14:53the cells,
• 14:55act by themselves? So I
• 14:56hope Wendell is gonna highlight
• 14:57some of the pioneering approaches
• 14:58that he has,
• 15:00done in his lab on
• 15:01that on that front. And
• 15:02the last pillar, it's compress
• 15:04and predict. Right? Because we
• 15:05generate lots of data,
• 15:07but data is just data.
• 15:09And and if without compressing
• 15:11and and transforming them into
• 15:12causal predictive models,
• 15:15we we we cannot really
• 15:17act on it, and also
• 15:18we cannot understand. So I
• 15:19hope it's we can develop
• 15:20models that can allow us
• 15:21to compress the data, transform
• 15:23them into tools, but also
• 15:25understanding.
• 15:27So with that, I wanna
• 15:28highlight just a few things
• 15:29in the last,
• 15:30five minutes
• 15:31that sort of highlights some
• 15:33of the efforts, we have
• 15:34started to push in these
• 15:35these directions.
• 15:37So on the on the
• 15:37AI front, if you think
• 15:39look look at the the
• 15:40immune system as a sort
• 15:41of a multiscale machine. Right?
• 15:42So going from molecules
• 15:44within cells like DNA, proteins,
• 15:46and RNA,
• 15:47Together, they they orchestrate
• 15:49the behavior of cells, and
• 15:51cells, of course, exist in
• 15:52different types states, and they
• 15:53they have different compositions and
• 15:55different tissues, and then so
• 15:56on. You can move up
• 15:57the the sort of chain
• 15:58all the way to the
• 15:59individual and then human population
• 16:01level. And when you think
• 16:02about machine learning and and
• 16:04AI,
• 16:05it machine learning and AI
• 16:06is actually a mapping problem.
• 16:07Right? So machine learning, you
• 16:08try to develop a mathematical
• 16:10model that can map certain
• 16:12types of input to an
• 16:13output. Right? So you've seen
• 16:15the phenomenal alpha four, for
• 16:17example, which maps sequences, protein
• 16:20sequences
• 16:20into structure, right, and map
• 16:22sequences into function. So if
• 16:24you have a mutation in
• 16:25the protein coding gene, we
• 16:26wanna predict, is it gonna
• 16:28be bad, for example. So
• 16:29you're quite familiar with these
• 16:30kinds of mapping functions. So
• 16:32the challenge in the in
• 16:33the immune system, it's it's
• 16:34it's the mapping problem across
• 16:36these scales, so to speak.
• 16:37Right? So in each one
• 16:38of these scales, you can
• 16:39sort of measure different features.
• 16:41So we often now see
• 16:42gene expression data. We often
• 16:44see cell composition and cell,
• 16:47phenotype data. But can we
• 16:48map those,
• 16:50features onto
• 16:52type ontogeny,
• 16:54onto dynamics and trajectory,
• 16:56and onto functions, right, and
• 16:57all the way to, the
• 16:58individual organismal level that, I
• 17:01I mentioned earlier today, basically.
• 17:03Right? And the challenge is
• 17:04data, actually.
• 17:06We talk about big data
• 17:07a lot these days, but,
• 17:08actually, we have relatively limited
• 17:09data, partly because they're expensive
• 17:11and and so on.
• 17:13So that's where, one of
• 17:14the efforts, we're teaming up,
• 17:15we're we're we're driving towards
• 17:17is this project called the
• 17:17human immunome project
• 17:19that sort of,
• 17:22sort of checks off this
• 17:22both these pillars in terms
• 17:24of leveraging the human,
• 17:26population in in in across
• 17:27the world in this case
• 17:28and connection to physiology,
• 17:30but also the measurement aspects
• 17:32that I mentioned. So we
• 17:33think that the time is
• 17:34ripe to do this given
• 17:36the public. It's it's becoming,
• 17:37really, in a in a
• 17:39way understand better about the
• 17:40the the the the importance
• 17:41of the immune system, and
• 17:42technology is at a point
• 17:43where we we can really
• 17:44start to measure quite a
• 17:45lot. And and then, you
• 17:47know, the AI revolution that's
• 17:48going on. And so this
• 17:49is a project,
• 17:50that, I've been working together
• 17:52on, on the science side
• 17:53together with Shai,
• 17:55Shenor of Technion, for the
• 17:56past couple of years with
• 17:57contributions from a number of
• 17:58people.
• 17:59So the goal is really
• 18:01to map, this kind of
• 18:02personal immune states along major
• 18:04axis of variation
• 18:05across the world. And we
• 18:07wanna create an open,
• 18:08data resource, to empower both
• 18:10research and train the kind
• 18:11of AI model I mentioned.
• 18:13So we wanna sort of
• 18:14cut across
• 18:15populations
• 18:16in in terms of environment,
• 18:18across the whole lifespan,
• 18:19the gender and and geography,
• 18:21of of course. Initially, we're
• 18:22gonna focus on predisease.
• 18:25And I I think that's
• 18:25actually a huge area to
• 18:26fill because we we're gonna
• 18:28monitor these people longitudinally,
• 18:30as well.
• 18:31And so this actually needs
• 18:32to be a a truly
• 18:33global effort for both scientific,
• 18:35so leveraging the variation across
• 18:37the globe, and also for
• 18:39equity reasons. So so far,
• 18:40a lot of the kind
• 18:41of data I mentioned that
• 18:41we're collecting in academia and
• 18:43some companies, they tend to
• 18:44be mostly in the US
• 18:46and in EU. We have
• 18:47actually very little data, dense
• 18:49datasets,
• 18:50in other from other parts
• 18:51of the world.
• 18:52So just briefly, the way
• 18:54we're gonna do this will
• 18:54be we're gonna start with
• 18:56the phase one sort of
• 18:57ten sites or so across
• 18:58the world. We're gonna deploy
• 19:00the state of the art
• 19:00technology to enroll about ten
• 19:02thousand to fifteen thousand people
• 19:04and follow them over time
• 19:05and also give them a
• 19:06vaccine to probe them. And
• 19:07then with that data, we
• 19:08hope to be able to
• 19:09learn
• 19:10sort of a compression scheme,
• 19:11so to speak. Right? How
• 19:12how can we measure all
• 19:13of these ten to the
• 19:14seventh or so parameters, but
• 19:16with maybe a few thousand
• 19:17measurements
• 19:17that can still capture most
• 19:19of the variation
• 19:20in the population.
• 19:21And then we're gonna create
• 19:22this new monitoring kit that
• 19:24will be cheaper and and
• 19:25and simpler to deploy
• 19:27across the whole,
• 19:28many more sites to to
• 19:29scale up the number of
• 19:30people we can look at.
• 19:32So the goal right now,
• 19:33it it's it's it's really
• 19:34the the scaling
• 19:36down. It's really important because,
• 19:37otherwise, this is still a
• 19:38very, very expensive project. So
• 19:40based on the data we
• 19:41have so far,
• 19:42based on our and other
• 19:43studies,
• 19:44we think this kind of
• 19:45compression is possible.
• 19:47So this is just showing
• 19:48you that some of the
• 19:49sites, that that's under discussion,
• 19:52and and we hope to
• 19:53launch one, here at Yale,
• 19:55to to coordinate and to
• 19:56build this kind of site
• 19:57model that can be replicable
• 19:58across the world.
• 20:00The other pillar I wanna
• 20:01quickly highlight is cells as
• 20:02tools and which it's the
• 20:04Chan Zuckerberg Biohub's goal. It's
• 20:05the harness immune cells as
• 20:06natural and engineer sensors and
• 20:08modulators.
• 20:10We wanna
• 20:11take these cells eventually for
• 20:12early disease detection and prevention,
• 20:15sort of motivated by the
• 20:17immune health metric sensing I
• 20:19mentioned earlier. Right? So you
• 20:20can see that clinically healthy
• 20:21people, they have underlying
• 20:23potential pathologies that we can
• 20:24we can we can detect.
• 20:26So one of the ways
• 20:27it's it's you can actually
• 20:29engineer cells. You can sort
• 20:30of, like, guide them into
• 20:31specific locations
• 20:32and express certain sensors, and
• 20:34they can they can sort
• 20:35of collect data on tissues.
• 20:37And that would be a
• 20:37dream in in also from
• 20:39the basic science standpoint. Imagine
• 20:40you actually got tools now
• 20:42that you can go into
• 20:43the tissues without invasiveness
• 20:45and be able to collect
• 20:45data, basically.
• 20:47So,
• 20:48yeah. So this was also
• 20:50reminds me of the magic
• 20:51school bus in in case
• 20:52some of you have seen
• 20:52that episode. I remember that,
• 20:54when I when I watched
• 20:55it with my son many
• 20:56years ago.
• 20:57So
• 20:58the Biohub, it's it's a
• 21:00sort of a two pronged
• 21:01structure. One is a physical
• 21:03hub. So the main hub
• 21:04is in New York. There
• 21:05will be a component in
• 21:06New Haven. There's also the
• 21:07investigator program that will be
• 21:09spread out through the three
• 21:10universities. So the partners at
• 21:11Yale, Rockefeller, and Columbia.
• 21:13So an investigator,
• 21:15RFA just went out. So,
• 21:17I encourage, you to apply
• 21:19and also go to the
• 21:20webinar. If you have scientific
• 21:22questions, after the webinar, please
• 21:23feel free to contact,
• 21:25Alisa and also myself.
• 21:27So and finally, please become
• 21:30look look at our website
• 21:31and consider becoming a member
• 21:32and join our contact list.
• 21:34And you can also,
• 21:36check check us out on
• 21:37social media. And, also, you
• 21:39can check out the members
• 21:40and various other, information on
• 21:41the website. And And we
• 21:43have a seminar and Chalk
• 21:44Talk series. And for the
• 21:45past year and a half,
• 21:46we have both outside and
• 21:48internal speakers.
• 21:49It's been quite well attended
• 21:51and a lot of, interesting
• 21:52discussions, especially during the Chalk
• 21:53Talks because we encourage folks
• 21:54to come and interrupt and
• 21:56ask lots of questions, and
• 21:58that can lead to both
• 21:58new collaborations and also new
• 22:00understanding.
• 22:01So with that,
• 22:03yeah, send us any ideas
• 22:05and suggestions.
• 22:06I wanted to sort of
• 22:07end by saying that,
• 22:09CSCI, of course, cannot do
• 22:11what I
• 22:12propose alone.
• 22:13So we are, already actively
• 22:15engaged as you saw with
• 22:17both Yale colleagues and also
• 22:18partners around the world. So
• 22:20today, we're gonna highlight, each
• 22:21one of these pillars with
• 22:23the speak with the invited
• 22:24speakers. So first on pillar
• 22:26one would be, Ling
• 22:27and Steve. They will talk
• 22:29about,
• 22:29human,
• 22:30immunology and variation in response
• 22:32to COVID nineteen,
• 22:34and also vaccine and infection
• 22:35response signatures. In terms of
• 22:37measurement and monitoring,
• 22:38Wei Ka, who's a pioneer,
• 22:40in in in random laser
• 22:41technologies, she's gonna touch on
• 22:43deep tissue imaging.
• 22:45In terms of engineering cells,
• 22:47Wendell Lim is one of
• 22:49the pioneers, in this field,
• 22:50so he's gonna tell us
• 22:51about their efforts.
• 22:53And finally, in in AI
• 22:55and also AI models,
• 22:57we're lucky to have, Deep
• 22:58Jaitley,
• 22:59who's one of the early,
• 23:02pioneers in deep learning on
• 23:03generative models as well as
• 23:05Maria who just, joined us
• 23:06at Yale. She's gonna talk
• 23:07about TCR specificity prediction.
• 23:10With that, thank you for
• 23:11your attention.
• 23:20Any questions?
• 23:23And visit us. And this
• 23:25is also where the CSCI
• 23:26seminar
• 23:27is being held.