PANEL DISCUSSION - HOW WILL GENOMICS AND AI ACCELERATE PRECISION MEDICINES

Transcript

• 00:01Yeah. I I had a
• 00:02question, I guess, for Clemens.
• 00:04You know, finding the the
• 00:06the relevant genes under those
• 00:08GWAS peaks is both really
• 00:09important and a challenge.
• 00:12And I was curious about,
• 00:13you you know, with the
• 00:14types of EQTL approaches that
• 00:16you were describing,
• 00:18whether these would pick up,
• 00:20I guess, variants that were
• 00:22conditionally
• 00:23dependent. Like, for example,
• 00:26you know, a variant that
• 00:27might be really important
• 00:29if you happen to be
• 00:30you know, have had a
• 00:31stroke or have been, you
• 00:33know, infected with a virus,
• 00:35but might be hard to
• 00:36capture just with thousands of
• 00:38samples out of a, you
• 00:38know, sort of random population.
• 00:40Yeah. Is so, I guess,
• 00:42how well does this type
• 00:43of analysis deal with that,
• 00:44and is there a way
• 00:45to somehow take into account
• 00:46these extra variables?
• 00:48Right. So so this is
• 00:49a very exciting avenue that,
• 00:53Zetsu and Lin, an instructor
• 00:54in the lab and I
• 00:55am very interested in, and
• 00:57and sort of it goes,
• 00:58is there any
• 01:00are there,
• 01:01loci
• 01:03that,
• 01:04are
• 01:05react to to pathology
• 01:07in in the brain. And
• 01:08we we've been focused so
• 01:10far really on loci
• 01:12quantitative trait loci that might
• 01:15be turned on or turned
• 01:16off,
• 01:19by Lewy body pathology.
• 01:21And
• 01:22and and so for our
• 01:24Parkinson's
• 01:26p d five d, you
• 01:27know, digital twin approach, so
• 01:29we're looking at Parkinson's brains.
• 01:31We we have the Lewy
• 01:32body
• 01:33pathology with the plaques and
• 01:34the tangles.
• 01:35We actually so
• 01:37so far, we've only found
• 01:39a few,
• 01:40target genes, but there's a
• 01:42clear interaction
• 01:43with Lewy body pathology. Interestingly,
• 01:46PM twenty d one,
• 01:48was one of the few
• 01:49genes
• 01:50where,
• 01:52Lewy body pathology actually,
• 01:54turned off the the regulatory
• 01:56f effects. So so there
• 01:58are some effects we're very
• 02:00interested in. There's much more
• 02:01to do, and,
• 02:03look forward to chat chat
• 02:04more with him.
• 02:06See,
• 02:08maybe as a follow-up question?
• 02:12Yeah. It
• 02:14worked.
• 02:15Yes. I follow-up follow-up question
• 02:17to Sean's question.
• 02:20What do you think is
• 02:21needed in terms of genomics
• 02:22to distinguish
• 02:24risk factors for onset
• 02:26and risk factors for progression?
• 02:28Because I've obviously, for the
• 02:29patients, if one has interventions
• 02:31for progression, you probably can
• 02:32save one third or half
• 02:33of the people that way
• 02:35because it's such a long
• 02:36duration.
• 02:37But you explained
• 02:39access to patient data with,
• 02:41early onset.
• 02:42But, is that sufficient, or
• 02:44do we need other type
• 02:46of genomics information to
• 02:48better distinguish these two factors?
• 02:50Right. So,
• 02:52didn't have a chance to
• 02:54talk about this. I'll touch
• 02:55up on this at the
• 02:56very end of the symposium.
• 02:58But,
• 02:58so,
• 03:00actually,
• 03:01we're very proud to
• 03:04really, in a way, have
• 03:05opened up the field of
• 03:07progression,
• 03:08genetics,
• 03:09or
• 03:10neurologic diseases.
• 03:12And so on that
• 03:14one slide that I showed,
• 03:15there actually the dimension of
• 03:17of of risk and the
• 03:18dimension of progression.
• 03:20And turns out
• 03:23there's there's some variants that
• 03:25regulate onset and progression, but
• 03:27then there was some variants
• 03:28that just regulate progression and
• 03:29others just,
• 03:31just risk.
• 03:32And and so,
• 03:35at the Adam Center, we
• 03:37are, in in this,
• 03:39multilayered,
• 03:41you know, Parkinson's Atlas or
• 03:43digital twin of Parkinson's.
• 03:44We are including,
• 03:46the time component
• 03:48both in terms of we
• 03:49have, ten thousand,
• 03:52longitudinal
• 03:53life patient cohorts where where
• 03:55we have longitudinal clinical data.
• 03:57And then on the pathology
• 03:58level, you know, we have
• 04:00to serious from on onset
• 04:02to progression. So so, yes,
• 04:04so this, I think, is
• 04:04a very interesting question,
• 04:07much to do, but we
• 04:08are we're
• 04:10we're building the platforms to
• 04:11do it.
• 04:18Oh, okay. Just a question
• 04:19for doctor Strippmatter.
• 04:21So right over here.
• 04:23Other side.
• 04:24Hi. Angela Kacasi, Arvinis.
• 04:27My question is about, per
• 04:29granuline
• 04:30reduction and how that impacts
• 04:32the aggregates that that form
• 04:35that are insoluble.
• 04:36And then you also mentioned
• 04:38that this rescues function
• 04:40and improves the damn microglia
• 04:42phenotype, etcetera.
• 04:44Do you believe that there's
• 04:45a biomarker
• 04:47for that type of soluble
• 04:48oligomer of tau that you
• 04:49think is causing the dysfunction
• 04:52that might be really meaningful,
• 04:54to really assess when you're
• 04:56perturbing lysosome dysfunction.
• 04:59We have a LARC two
• 05:00PROTAC that we're just going
• 05:02to talk about at ADPD
• 05:03next week, where we see
• 05:05reductions in soluble oligomers.
• 05:07We know we're impacting the
• 05:08lysosome system,
• 05:10but we're not impacting the
• 05:11insoluble tau fraction. So we
• 05:13think this would be really
• 05:14interesting
• 05:15to study in human disease,
• 05:17but we're interested in your
• 05:19insights on how would you
• 05:20go about doing that.
• 05:22Yeah. I think those data
• 05:24certainly you know, the simplest
• 05:26model is there's a one
• 05:27to one correlation between aggregation
• 05:30and disease, but it's I
• 05:32think these data amongst many
• 05:33others prove that that's not
• 05:35the case.
• 05:36And so as you're pointing
• 05:37out,
• 05:39understanding whether there's different
• 05:42types of misfolding, different aggregates,
• 05:46and which ones are most
• 05:47important for disease is an
• 05:49important question.
• 05:50This may be one handle
• 05:52on getting at that by
• 05:53studying,
• 05:54these animals.
• 05:57But I think, you know,
• 05:58the short answer is we
• 05:59don't know what are the
• 06:00most relevant
• 06:02toxic species of misfolded
• 06:05neurodegenerative
• 06:06proteins.
• 06:12Hi. This is a question
• 06:13that piggybacks off the first
• 06:14two.
• 06:15And maybe,
• 06:16doctor Liu,
• 06:18specifically, you mentioned missing heritability.
• 06:20And, certainly, there's, a lot
• 06:22of people in the Parkinson's
• 06:23field that feel that there
• 06:24isn't missing heritability. It's really
• 06:26environment. That's, that's the huge,
• 06:29sort of elephant in the
• 06:30room, and that could mean
• 06:31exposure to paraquat, which has
• 06:32talked about a tremendous amount
• 06:34in the advocacy world versus,
• 06:35you know, exercise, exposure to
• 06:37coffee,
• 06:38huge list. So,
• 06:40that's that's a huge undertaking,
• 06:42and I wonder how that's
• 06:43sort of feeding into many
• 06:45of your models. Sure. Yeah.
• 06:46No. Thank you for that
• 06:47question. You know, the the
• 06:49question of missing heritability is
• 06:50always hard because it's it's
• 06:51the unknown. And and you
• 06:52mentioned environment.
• 06:54So when we when we
• 06:56try to capture the missing
• 06:57heritability using these wearable devices,
• 06:59it could be the case
• 07:00that these devices are actually
• 07:03capturing something about the environment,
• 07:05for example,
• 07:07response to stimuli.
• 07:08So the lines get kind
• 07:09of blurred between what the
• 07:11missing heritability is versus what
• 07:13the environment is. But the
• 07:14good thing I think there
• 07:15is using these devices, like
• 07:17wearables and smartwatches,
• 07:19they are measuring continuously
• 07:22as someone kind of interacts,
• 07:24through their environment and with
• 07:25their environment. And so I
• 07:27I do think that
• 07:29by using such devices, unlike
• 07:31other technologies which only measure
• 07:33one time point, we're we're
• 07:35really actually able to capture,
• 07:37much more of that environmental
• 07:39effect. And I hope that,
• 07:40things like Parkinson's can be
• 07:42better characterized,
• 07:43through devices like this.
• 07:48Yeah. I have a question
• 07:49on the circular RNA
• 07:51that you discuss. Circular RNA.
• 07:53Yeah.
• 07:54Yeah. You say that there
• 07:55was an increase in particular
• 07:57or certain circular RNA in
• 07:59Parkinson disease.
• 08:01And,
• 08:02I thought do you see
• 08:03them as upstream or downstream
• 08:05of
• 08:06Parkinson's,
• 08:08phenotype? Because you say that
• 08:09you were thinking of treating,
• 08:14cells with circular RNA to
• 08:16see whether you rescue
• 08:17the Parkinson's.
• 08:18So it's unclear to me
• 08:20where you put them in
• 08:21the
• 08:22sequence of event. I think
• 08:23it's unclear to me too
• 08:25whether it's upstream or downstream.
• 08:28We do this is association
• 08:30study based the the published
• 08:31one, but
• 08:33my personal
• 08:36hypothesis there, it could be
• 08:37upstream.
• 08:39Why I think circular RNA,
• 08:41most of them don't translate
• 08:42to protein.
• 08:43But that but why they
• 08:44are,
• 08:45transport and enriching synapses, they
• 08:48must play a role there.
• 08:49Whether they function as RNA
• 08:51binding protein, like, help the
• 08:52transportation of other messenger RNA
• 08:55to the synapses
• 08:56because the local translation center
• 08:58in the synapses, as you
• 08:59know.
• 09:00Whether
• 09:01and, also, the known function
• 09:03for other, circular RNA, people
• 09:05show RNA binding protein,
• 09:07like and muscle blind, for
• 09:09example, a good example published
• 09:10already. This microRNA sponge can
• 09:12be another mere seven,
• 09:14like, the the first circulary
• 09:16defined in two thousand seven
• 09:18fourteen was c dot one
• 09:20a s. They have seventy,
• 09:22miR seven binding sites there.
• 09:24So it's like a sponge
• 09:25to release and solve miR
• 09:26seven. MiR seven controls the
• 09:27nuclei. So this is already
• 09:29kind of known pathway
• 09:31from circulating to macronally to
• 09:33snooply.
• 09:34But how this link to
• 09:35synapses, that's really something I
• 09:37mean, synuclein is part of
• 09:39the synapses,
• 09:40snare pathway. But how this
• 09:42caused the the synaptic
• 09:44disruption,
• 09:45I I don't have data.
• 09:47But this is I think
• 09:48it's just my, hypothesis. I
• 09:50think it's more likely to
• 09:51be upstream.
• 09:57I
• 09:59I would like to ask
• 10:00a question to Pablo Sardi
• 10:02in the audience,
• 10:03who
• 10:04is the head of,
• 10:06rare and neurologic disease research
• 10:08at Sanofi.
• 10:09Pablo, I know you have
• 10:12a deep pipeline around,
• 10:14GBA, GC, GKs, drugs,
• 10:18in Gaucher disease and Parkinson's
• 10:20disease.
• 10:21And and so I was
• 10:22very curious what your reaction
• 10:24was to the data,
• 10:26Steven showed,
• 10:28about,
• 10:30galactos ceramide and,
• 10:33neurofibrillary
• 10:34tangles.
• 10:36This is supposed to work
• 10:37the other way around where
• 10:38we ask questions. This is
• 10:40an inverted inverted classroom. So
• 10:42so the so I I
• 10:43did have a question. One
• 10:45but the first thing is,
• 10:46you know, congratulations
• 10:47to you and, and everyone
• 10:48for, you know, the symposium
• 10:50and for the open science.
• 10:51I think, you know, that
• 10:52that's something that I, I
• 10:53think it's, you know, it's
• 10:54super important for everyone,
• 10:56not just for, you know,
• 10:58academic
• 10:59people, but also for industry
• 11:00and how you know, and
• 11:01for patients as well.
• 11:03Then second, I'm gonna answer
• 11:05the question. And then third,
• 11:06I'm gonna throw a question
• 11:07at at you guys as
• 11:08well. So, from from that,
• 11:10I, you know,
• 11:11I think there's definitely and
• 11:13we were discussing earlier today,
• 11:16whether the lysosomes yeah. What's
• 11:18the function of the lysosomes
• 11:19and how do how do
• 11:20they impact,
• 11:21disease, not only risk, but
• 11:23also progression of the disease
• 11:24in general and how we
• 11:25target it.
• 11:27I I don't I don't
• 11:28think we're just scratching the
• 11:29surface. And if we think
• 11:31about even the trials that
• 11:32are go ongoing, that are
• 11:34related to lysosomal functions, whether
• 11:36it was, you know, our
• 11:37Vangustat trial,
• 11:39the trials from our lecture
• 11:40on programmiling
• 11:41or the try or, you
• 11:42know, programmiling or the Lark
• 11:44two, etcetera. I think there's
• 11:46a lot that we need
• 11:47to learn about how to,
• 11:48you know,
• 11:49target this from a progression
• 11:51point of view.
• 11:52It's super complicated. Even for
• 11:54Garcetti's disease, that it's a
• 11:55single monogenic
• 11:57disorder, we can see that
• 11:58there is, more complications
• 12:00than just the the enzyme
• 12:01replacement therapy.
• 12:04And then how do this
• 12:05play out in a in
• 12:06a sequential
• 12:07manner? You know, whether if
• 12:09we fix the pro granular
• 12:10defect, what are we gonna
• 12:11do? And this is a
• 12:12precision medicine approach that your
• 12:15you and many others are,
• 12:16you know, trying to put
• 12:18forward for Parkinson's and other
• 12:19neurodegenerative
• 12:20diseases.
• 12:21So I think we're
• 12:23just scratching the surface. I'm
• 12:25sorry. I don't have a
• 12:26good answer to, say, you
• 12:27know,
• 12:28I I don't think there's
• 12:29one only one enzyme that
• 12:31it you know, it's downstream
• 12:32of the program link or
• 12:34the TMN one that that
• 12:35we can go and say,
• 12:36you know, if we fix
• 12:37that defect, we're gonna fix
• 12:38it all.
• 12:39And that's why the precision
• 12:41medicine is gonna be important.
• 12:43And for the precision, I
• 12:45the way I foresee it,
• 12:47maybe that's more important,
• 12:49is we're gonna have different
• 12:51approaches that will move biomarkers
• 12:53at some point. And, you
• 12:54know, proteomic biomarkers are the
• 12:56ones that are closer
• 12:57today,
• 12:59to in in the clinical
• 13:00applications.
• 13:02And when we see that
• 13:03the biomarkers are moving in
• 13:05the right direction, then those
• 13:06are the patients that are
• 13:06gonna find a benefit, and
• 13:08we're gonna continue with those.
• 13:09And if the biomarkers are
• 13:10not moving, then we're gonna
• 13:12we have to stop the
• 13:12therapy.
• 13:13And that's as, you know,
• 13:16silly or or human intelligence,
• 13:18not AI, but HI
• 13:20as we can be, I
• 13:21think at this point.
• 13:23And then okay. Then I
• 13:25kind of answered the question.
• 13:27And for a quiet guy,
• 13:28I think I talked too
• 13:29much already.
• 13:30And then I'm gonna throw
• 13:31a question at you guys
• 13:32because, you know, we we
• 13:33think about risk factors and
• 13:34project
• 13:35and and,
• 13:37progression of the disease.
• 13:40How can we use the
• 13:41models that you're setting up
• 13:43in order to understand
• 13:45or to to make trials
• 13:47faster
• 13:48and smarter?
• 13:50And is there something that
• 13:51we can you know, if
• 13:52we leverage the models where
• 13:53you see the even
• 13:55project,
• 13:56how the different cell types
• 13:57are being affected,
• 14:00can we learn something about
• 14:01biology using GBA and Lark
• 14:03two as two very simple
• 14:06modifiers for progression of the
• 14:08disease? Both are risk factors,
• 14:09but the progression seems to
• 14:10be very different for those
• 14:11two patients.
• 14:13So I think it's a
• 14:13question for you and for,
• 14:15you know, doctor Dong.
• 14:17Yeah.
• 14:18Yeah. I have a lot
• 14:19to say about it, but,
• 14:21Jason, you might have something
• 14:22to say about this as
• 14:23well. Yeah. This is perhaps
• 14:25not exactly,
• 14:27related,
• 14:28a direct answer, but it
• 14:29it's very related, which is,
• 14:30thinking about you mentioned clinical
• 14:32trials and how we can
• 14:34evaluate whether a drug is
• 14:35working,
• 14:36and and kind of also
• 14:38the,
• 14:39kind of the trajectory of
• 14:40of, development
• 14:42is using these digital biomarkers.
• 14:44So let's just say we
• 14:45have a sensor, watch, whatever
• 14:47it may be, and we
• 14:48can build an AI model
• 14:50that really represents the severity
• 14:52of the progression of an
• 14:53individual.
• 14:55In the context of a
• 14:56a clinical drug trial, instead
• 14:58of waiting perhaps ten weeks
• 15:01for kind of a clinical
• 15:02endpoint to be measured, now
• 15:04you you have a live
• 15:05marker at every given hour
• 15:07how someone is changing and
• 15:09responding to that, to that
• 15:10drug. And so I think
• 15:11that's one of the ways
• 15:13that something like a digital
• 15:14marker, whether it be with
• 15:16wearables or other sensor technology,
• 15:18can can aid that.
• 15:20And and so, in in
• 15:22the drug development,
• 15:24platform that we're building up
• 15:26at the Adam Center, we
• 15:27are we are trying to
• 15:29do, two things. Number one,
• 15:31to do,
• 15:32clinical trials in a dish
• 15:34using personal personal
• 15:36stem cells from from patient
• 15:37from our Yale Harvard biomarker
• 15:39study
• 15:40to,
• 15:41sort of use this patient
• 15:43avatars, not just one, but
• 15:45a a a cohort of
• 15:48patient stem cells with with
• 15:50with, you know, for example,
• 15:51the PM twenty d one,
• 15:53variant
• 15:54to,
• 15:55to as an initial test
• 15:57for drugs.
• 15:58And then if they respond,
• 16:00recontact the patients and stratify
• 16:02clinical trials by targeting,
• 16:05carriers with with the genetic
• 16:07variants.
• 16:08And,
• 16:10we,
• 16:11also have,
• 16:13biofluids
• 16:14for for all of these
• 16:15patients,
• 16:16blood and impart CSF.
• 16:18And so then we, you
• 16:20know, go to freezers to
• 16:21develop tailored biomarkers.
• 16:23So that that's that's really
• 16:25sort of the in,
• 16:28the drug development
• 16:29strategy. And then we're
• 16:31planning to and hoping to
• 16:33work with Luke Lee, who
• 16:35will tell more about his,
• 16:37Parkinson's brain on a chip
• 16:39model for for for for
• 16:41for drug development.
• 16:44And and so so I
• 16:45think both at the level
• 16:46of preclinical
• 16:48trials and then,
• 16:49clinical stratification based on on
• 16:52genetics,
• 16:53I think we can make
• 16:54make a headway.
• 16:57So I would like to
• 16:58thank everyone. Jesse, is your
• 16:59question a short one?
• 17:02Because we have only this
• 17:03view between you and the
• 17:05country. I'll I'll I'll try
• 17:06the best short. So, Thomas,
• 17:08really congratulations
• 17:09on this notion
• 17:11of looking at what I've
• 17:12always called the dark matter
• 17:14of the genome.
• 17:15Yeah. Those seventy five hundred
• 17:17and six,
• 17:21genetic elements that may affect
• 17:23disease development progression.
• 17:25But when you my question
• 17:26for you is this. Conceptually,
• 17:28when you target one of
• 17:29those
• 17:31in the dish,
• 17:33how do you account
• 17:35for what's going on with
• 17:37the other seven
• 17:38thousand
• 17:40five hundred and five or
• 17:41a significant fraction of those
• 17:42that might also be
• 17:44affecting disease progression and the
• 17:46variability,
• 17:47the inter patient
• 17:49variability
• 17:50that those small differences
• 17:53might introduce.
• 17:54Right. It it sort of,
• 17:56looks like, you know,
• 17:59impossible
• 18:00sort of at the at
• 18:01the macro level. But, actually,
• 18:02if you dig into,
• 18:04specific
• 18:06variants and specific target genes
• 18:08and target pathways, it becomes
• 18:10all much more clear.
• 18:12For example, one of the
• 18:13GWAS functional GWAS target genes
• 18:16we found is SCARP two,
• 18:18which is,
• 18:19produces the lam protein,
• 18:21which is the transporter
• 18:23for GKs, GBA from the
• 18:26ER to the lysosome.
• 18:27And so, clearly,
• 18:31if you target this pathway,
• 18:33that should work for the
• 18:35ten percent of patients that
• 18:36carry the GBA mutation
• 18:38and and, you know, the
• 18:40whatever percentage of patient carrying
• 18:42the SCARB two mutation.
• 18:44So I think so that's
• 18:45that's, I think, one,
• 18:47way to make this work.
• 18:49The other is we are
• 18:51today, you know, we focused
• 18:53on sort of one variant,
• 18:56one well,
• 18:58pleiotropic
• 18:58target genes.
• 19:00But, actually, you know, the
• 19:01big vision is really to
• 19:03be able to to input
• 19:05to look at the whole
• 19:06genome,
• 19:08look at all the risk
• 19:09variants
• 19:10a person has that might
• 19:11be Parkinson's variants,
• 19:13might be some Alzheimer's variants.
• 19:15Right? And use it,
• 19:17use this polygenic
• 19:19input
• 19:20to to identify
• 19:22the polygenic,
• 19:24RNA consequences and and and
• 19:26treat them.