0:01
the drugs that we have are not fully utilized to treat every disease that they possibly can and it's our mission to take all that incredible work and just do the little bit of more effort it takes to find the additional diseases
Search this video
0:13
that they could potentially treat this kind of R&D would cost trillions of dollars for traditional Pharma the pharmaceutical industry I don't demonize them at all they do an incredible job at generating novel compounds it's not
0:29
their fault that that there's just no incentive to pursue additional research around drugs that are off- patent and are generic we take those drugs and we go after diseases that are totally neglected by the pharmaceutical industry
0:42
there's so much loow hanging fruit with the 3,000 drugs that already exist we are going to spend years and years unlocking the life- saving potential of those you make it very easy to want to give you money well hello everyone it's Jim o shasy
1:04
with yet another infinite Loops today I have a very special guest a member of my family well kind of a member of my family uh Dr Grant Mitchell who man Grant you have more degrees and qualifications than virtually anyone
1:20
else in this family you are you you are a medical doctor you are an MBA you are ex McKenzie you are an expert in machine learning and Ai and today we're going to talk about the company you founded every cure which whose mission is essentially
1:40
to say that there are lots of people out there with diseases for which there is no acknowledged treatment and you and your partner are on a mission to take a look at not only all existing drugs but also combination of new drugs to try to
2:01
find cures for the what is it 3,000 diseases or more that have no effective treatment Grant welcome thank you so much for that kind introduction Jim and uh the the uh problem is dire more dire than you even uh portrayed there it's
2:18
actually around depending on the definition of disease you choose 22,000 diseases and uh the vast majority of which are both rare and have no single approve treatment so that means that one in 10 people uh in this world will
2:35
develop a disease that has no FDA approved treatment so the doctor will have to say I'm sorry there's there's nothing we can do quite yet for this and and the hypothesis that we have that we're very certain is true is that the
2:50
3,000 or so existing drugs that we already have can treat a lot of these diseases and there's certain barriers that exist that disincentivize this kind of res research but the organization that we've launched called every cure is
3:03
designed to overcome those barriers and and pursue that research and make sure that patients get these treatments I recall a lunch many years ago that you and I had in Union Square and you were working on something similar then talk a little bit about
3:20
that because we chatted uh before starting the recording uh uh about it but you've been at this a long time yeah the origin really goes back before I knew what you know the word machine learning was um and it all started in medical school uh I went to medical
3:38
school at Penn and and I did a combined degree for an MBA there at Warden as well and my roommate at the time a guy named David Fagen bomb who's now a co-founder at every cure he uh was first of all he's an incredibly healthy person
3:54
he was used to be the quarterback at Georgetown and he was uh you know a weightlifting fanatic and we were in med school together and he walks into the living room in our in our apartment and he says Brant something is just wrong I I
4:10
feel like I'm G to die I said said what how what are you talking about David you're the healthiest guy I know um and just a few short weeks later he was in the ICU uh and nearly in a coma his his organs were shutting down uh he could he
4:28
could barely function and uh he nearly lost his life he was so close to death that a priest came in and read him his last rights uh I literally hugged him goodbye never thought I was going to see him again and uh but amazingly he
4:42
survived probably because he had such a strong Baseline to begin with uh and the only way to stop what was happening in his body at that time was to obliterate his immune system with seven agent chemo so that his immune system can stop
4:54
attacking every uh organ in his body well the Cadence of that disease the way it end up working was you'd have a flare where a cyto kind storm and an immune reaction attacks everything in your body then you have full-blown chemo to stop
5:13
it and then you're healthy for a while and then it happens again about every 8 to 12 months and you can only go through that cycle so many times until you know the uh it kills you and so he was given probably three years to live and we
5:29
being naive early 20 year olds um 20-some year olds we looked at this problem and said well if no one else is working on this I guess we will uh you know David himself said well I have nothing better to do I've got three years to figure this
5:47
out and so you know we called every researcher in the world uh invited them to a a conference that we thought they might be interested in already attending anyway and we sat at the front of the room and said hi uh this is David and
5:58
this is Grant and we're here to cure castan disease and and they all kind of chuckled and said good luck kids um but what we realized is that these 20 or so re researchers were the only ones in the world that knew anything about this
6:12
thing but they weren't working together and just as much as it was a science problem to try and figure out how this disease worked in David's body um it was an organizational and stakeholder alignment problem to get everyone to
6:24
work effectively in a in a in a in a cohesive constructive way that was made sense to elucidate our understanding of the disease in the most efficient way possible and uh as we did that we would study David's blood and try and identify patterns in
6:42
the way that the sakin were expressing themselves and to see if maybe we could figure out which pathway was actually going wrong in his body and then we would start testing drugs on him that uh ultimately after about 3 years with you
6:56
know almost no time left found a drug that ultimately put him into remission so given these flares that came every 8 to 12 months it was we were on the edge of our seat you know for you know month 8 month 12 went by month 13
7:11
14 16 18 two years went by and I said my God I think he's okay and so it's been it's been over 10 years now just last month we marked 10 years that he's been in remission has been completely healthy he's gone on to get married have two
7:27
kids uh he runs the he helps run the orphan Disease Center here at the University of Pennsylvania where he also has his own lab that studies sakine disorders and got on to repurpose another 15 or 16 drugs in both his disease which is called camine
7:42
disease uh other diseases like angio saroma and even covid and so the essence of what we were doing was you know trying to understand the underlying mechanisms of these disease and then finding drugs that already exist that
7:59
line with those mechanisms so we don't have to spend a billion dollars in 10 years to develop a whole new drug which we didn't have the time or money to do it was to move quickly to identify drugs we already have that might hit these Pathways that are shared uh with with the original indication that they were approved for and that's where we really uncovered this issue and so David was off in you know the lab doing this in a more um you know
8:30
uh lab focused scientific basic science way and my career took me off into a different place I I uh launched a health technology company a couple different ones uh right after during med school and during business school and then I
8:47
went off to work for McKenzie for a number of years in the Silicon Valley office and then later on in their machine learning division called Quantum Black based out of London and there I was leading teams that were really pioneering
8:59
the use of large medical record databases to identify subpopulations where a drug might perform better might be higher in efficacy or or better in safety and and we realized that that's really in a way it's kind of drug repurposing it's
9:18
taking a drug and finding a population where it works a little bit better and a drug that already exists and as David was working in the lab and I was working in the data we kind of came together and we say can we can we Auto what we've
9:30
done can we scale what we've done in just one disease and given the explosion and the amount of data that exists out there and the explosion or the the improvements in the way that we can harmonize and integrate the data into
9:42
one place and then the models that have been built to analyze that data we we we thought that maybe it would be possible now we and we would check in every few years 2016 2017 wasn't really possible you know we had this dream for a long
9:55
time 2018 2019 is probably when I was talking to you and I was thinking about you know can we can we do this and really uh late lately it's become it's become possible especially with u you know like I said before more data structured better you have models like
10:11
these large language models that that are able to digest all of medical literature output it in a structured fashion compile it into a biomedical Knowledge Graph these really interesting ways to to display and and analyze this
10:24
kind of data and ultimately that's how every cure uh was formed was was the was the concept that the drugs that we have are not fully utilized to treat every disease that they possibly can and we can use utilize artificial intelligence
10:38
to unlock their life-saving potential wow um just so incredibly impressive and you know a million questions bring to mind uh as you know my oldest sister Le died of Lupus um and when you said the the cocine storm she had a kind of a similar
11:00
thing where she would go into remission and then there'd be a massive attack and and it wasn't like clockwork uh like your colleagues uh but um when when she died in 1971 it was like nobody knew very much at all about the about the disease
11:22
and in the in this case did you find that um H the The Cure that worked for your colleague was that transferable to other people with the similar uh disease yeah so the Cure that worked for him um you know we we we studied his
11:39
blood we sampled his lymph nodes um we did imunohistochemistry and and flow cytometry and basically found that there satchin were elevated another molecule called VF is elevated there's t- Cell Activation this all pointed towards something called the mtor uh pathway and started looking at different drugs that would uh uh would would hit that pathway settled on a drug called cilus Ser cus has been around for decades it's uh it's actually isolated
12:11
from a fungus found in the soil on Easter Island it's amazing right and uh and it it shuts down the overactivation of this pathway that leads to this Cascade that causes this whole cyto kind storm and for David it works perfectly and it also
12:27
works for about a third of the other patient Pati s that have disease like David and so that's resulted in you know the benefit to countless you know thousands and thousands of of patients lives so um it's it's been it's a pretty
12:41
thrilling and uh satisfying and motivating thing to be able to figure something like that out and to be able to do it uh to have the opportunity to do it more and at scale and and have the opportunity to save potentially millions
12:54
of lives is uh is a huge motivation for my team yeah yeah and that's one of the things that uh I'm I'm pretty involved through investment and other areas in Ai and AI development and it's actually one of the use cases that I find the most
13:12
exciting uh be because uh the number of people who are suffering and need not be suffering just through efforts like your own is in the millions and uh what what an amazing use case for this technology you you mentioned that it uh it helped
13:32
about a third of other uh people with the disease and when we were chatting earlier you mentioned something about looking at patients who have a certain disease uh but are in some way resilient uh and and even though they have that
13:50
particular disease their genome has some kind of pattern in it that um either blocks or diminishes or ameliorates the condition are you doing are you doing that type of work as well for example when you find one of uh uh the and vice
14:09
versa when you find one of the two-thirds that it doesn't work for are you seeing things in their genome that gives you the reason for why it's not working yeah so we are we aren't doing that work specifically but the way our organization designed is that we
14:27
incorporate that work into our data set and so um we we're actually a nonprofit organization and we designed ourselves as a nonprofit you know for years we thought okay how do we make this into a business what we did with with CMA
14:41
disease and we just kind of couldn't figure it out because we're going after drugs we don't own and we're going after diseases that are neglected in small populations there's just not really a business there so we couldn't quite we
14:52
couldn't quite piece it together and it and it was really an aha moment that this this is design it should be designed as a nonprofit and it should be an AI company because if you're if you're want to build the world's best AI
15:06
platform for drug repurposing you're going to need the world's best data set to train it and you're not going to get your hands on all the data that you want to get your hands on if you're a competitor to all these people that are
15:19
trying to use this data so we're collaborative we're non-competitive we are not profit seeking we our primary goal is to to relieve patient suffering and save patient lives so uh I'll get to your question about how we're utilizing that
15:37
kind of uh resiliency data that I mentioned before but first i'm going to to help you understand how we use that I'm going to describe the kind of data set that we're constructing and it's something called a biomedical Knowledge Graph it's well
15:51
known in in the uh areas and the fields that that we're in but maybe not a commonly known term to to the Layman but it's AFF ly a a uh representation in uh 3D Vector space of all of the biomedical knowledge we have as Humanity every drug
16:12
every Target every protein every Gene every pathway cell type organ system Etc and how they relate to different phenotypes symptoms and and diseases and so every one of those biomedical Concepts that I just described would be
16:27
represented as a node and then every relationship it that concept has with another relationship like a drug treats a disease there would be an edge and so that's kind of they call it a semantic triple drug treats
16:44
disease so you've got uh a node an edge and a node and imagine a graph of every known you know signaling molecule and protein and and and a concept you can imagine tens of millions of nodes even more edges representing all of human knowledge in in
17:05
biology and uh that's what uh multiple people have constructed uh the actually NIH funded a a program called the inats translator program where a number of these knowledge grafts have been constructed other groups are doing it a
17:20
lot of private companies have their own we were compiling them and integrating with with an integration layer that kind of takes the best from the the top public ones and then layers in additional proprietary data that we get
17:32
from other organizations or data that we generate on our own and and the example that you just mentioned a company that is working on tracking uh genetic diseases and groups of people with the same genetic disease and looking at
17:50
subpopulations within that group where there might be some resilience to the the mutation and then studying their genome to say okay what other uh proteins are being transcribed that might be protective against this mutation and then going out and
18:06
designing drugs that uh might mimic that protection well how's that data going to fit into my knowledge graph well you can imagine that I now if I have the the data set that they're working with I know that there's a mutation that
18:21
results in a disease so a gene um associated with disease that that's that no Ed in a node and I also know that this other protein is protective of that disease so that just information that goes into the graph and the more truth
18:36
that I put to that graph the more I can train that graph to identify patterns of successful examples of a drug working for a disease and then it can try and find that pattern elsewhere where it finds it it either identifies a a and
18:53
nodes and edges that should already be connected or or are connected in our knowledge space but no one has actually acted on or it can maybe even generate a hypothesis on a totally new Edge that is novel and and not has never been considered before by experts before so
19:09
to answer your question again is you know we're not doing that work ourselves but we integrate the knowledge from that work so it can train our models and so we can pursue drug repurposing ideas which is absolutely fascinating to
19:23
me um it seems to me that when you look at traditional Pharma uh you know this kind of R&D would cost trillions of dollars for traditional Pharma uh pursuing a for-profit new drug is there is is that a mindset that is difficult for you to overcome or is
19:46
Pharma embracing what you're doing you know the pharmaceutical industry is it I would I I wouldn't I don't demonize them at all they do an incredible job at at generating novel compounds uh that they can market and return shareholder
20:04
value with and it's not their fault that there's just no incentive to pursue additional research around drugs that are off patent and are generic and no one can make money on and so that actually represents around 80% of all
20:22
drugs so you have these precious molecules of life-saving capability that TR ions of dollars has got have gone into producing that now sit on our Pharmacy shelves and are sort of ironically or paradoxically not pursued further in
20:40
terms in terms of uh significant research efforts um and so we take all of we stand on the shoulders you know of that incredible effort to to just push it past you know we feel like we're on the 99 yard line with all of these drugs
20:57
and it and and it's our mission to to take all that incredible work and just do the little bit of more effort it takes to find the additional diseases that that they could potentially treat uh so I I am uh more than happy to to
21:12
work with the pharmaceutical industry and they're generally happy to to work with us because we're not we're not a threat uh to them we are we are enhance enhancing the use of the things that they've already built yeah and uh how do
21:25
you handle there there must be uh in the in the V morass of how we're handling Medical Data privacy issues ETC and then throw on top of that regulatory issues intellectual property issues legal issues how are you traversing all of
21:42
those it seems like it could be quite an impenetrable maze for a nonprofit yeah it's a really complicated landscape there's all sorts of different as you mentioned IP issues there's different routes to FDA approval um one
22:00
way we stay out of it is by staying out of it if if we do if we do the research um and we're not looking to pursue IP we're not looking to capture IP and we're not looking to then license it out or or uh or sell it to someone
22:17
else uh we're just simply identifying the viability of a drug and a disease and we generate the evidence package that is sufficient to convince the uh physician and research community of of its Truth uh then it can update guidelines and it can be uh prescribed
22:38
by physicians doctors can prescribe off label they can prescribe any drug for any purpose they see medically fit um and it doesn't have to necessarily be FDA approved for that initial purpose the real trick is making sure that
22:53
insurance companies uh actually cover it when when it is prescribed in that way so hypothetically would one of the potential outcomes from you building these amazing graphs and data sets be where you could find a not only a new
23:11
use for an existing drug that's been approved for something else but a combination of existing compounds that leads to a brand new drug and if that happens then how do you deal with the IP issues ETC Cent nonprofit uh do do they just become part
23:32
of the common data set uh for others to pursue yeah so it's really not our Focus you know we're not designing novel compounds um we we we think that there's so much loow hanging fruit with the 3,000 drugs that already exist that we
23:49
we're going to spend years and years unlocking the life saving potential of those and the reason why we're focused there is because that is the fastest way to save human lives because you don't have to if you if you develop a novel
24:05
compound you have to go all the way through the entire clinical development of an approval process indd phase one phase two phase three trials um this takes years and years and hundreds of millions of dollars whereas in a in certain scenarios in drug repurposing
24:22
just like with my co-founder David within weeks of Us coming up with the hypothesis that this drug might work for him as long as we could find a physician that would prescribe it to him it went directly into his human body just weeks
24:36
later so that brings me to sort of this issue that I think we're going to see and you as an investor might have uh make yourself aware of is that there's going to be lots and lots of failures in the world of AI driven drug Discovery
24:54
and that's because not only are you an AI company that's generating hypothesis you're also a biotech company that has to validate a novel compound and bring it all the way through the clinic uh through clinical trials and through
25:08
regulatory approvals into patients so here you are an AI company you've hired up your team of you know 50 data scientists and experts and you come up with your hypothesis you say okay great you're not Amazon that gets you know to
25:20
AB test where they're going to put a button like on the user interface and then they get feedback by the end of the day and okay move the button here instead of here when when you come up your hypothesis after your AI team says
25:32
okay this is where we're going to this is what the drug we're going to uh move forward with you now have to go through potentially 10 years and and hundreds of millions of dollars of of additional development so you don't know if your AI
25:44
team built anything of value you don't have that validation feedback loop that you do in other AI consumer-based organizations so now you're juggling uh sustaining an AI corporation that doesn't have a feedback back Loop while you have to also pay for uh the
26:00
clinical development of of a drug and so it's just it's a it's a tension that that's hard to hard to manage and Drug repurposing solves that tension it allows us to go from hypothesis to validation in a much tighter feedback
26:16
loop so what we're doing is something that both helps patients in the fastest and cheapest way possible but also the happy accident is that we push forward the field of data driven drug Discovery because we can inform our models in a
26:30
faster feedback loop yeah to me it's just like almost mindblowing the way that you can take the new AIS that we have developed and take this huge body of knowledge and repurpose things that we wouldn't even I think that's one of
26:48
the use cases of AI I you and I can't look at a data array of a 100 million items and and figure out which seven fit together in that wonderful lock and key mechanism whereas of course AI uh that is kind of what it's designed to do have you have
27:08
you looked at um I I presume but correct me if I'm wrong uh are you extending it to not just existing drugs uh in the marketplace but to the you know the historical use of plant Medicine of a variety of those things
27:28
and if so what where are your data sources there yeah these and some of the N drafts that already exist they do include some of these you know neutraceuticals and and other sort of active pharmaceutical ingredients that
27:42
that that could be relevant to this kind of work but in terms of getting it directly into patients we're going to only um pursue Concepts that are readily available to either to be prescribed or to a consumer yeah I think another interesting thing about
27:58
uh you know this this particular application in artificial intelligence is that you know you can't trust the AI completely we're certainly not there you know I think someday in the future maybe a hundred years from now we'll say my
28:16
God we tested drugs on humans that's how barbaric you know like like didn't like what a dangerous and awful thing to to do um but uh we're not there and in the meantime we do have to test it on humans and and so uh the way we do that now is
28:34
that there has to be an incredibly tight interface in your development of your um your algorithms and your your platform with medical expertise and so one thing I learned when I was at Quantum black and at McKenzie is and we would go up
28:52
against other machine learning organizations I remember one time we we they put us head to-head uh with group and they said okay whoever comes with the best insights in the next three months we're going to we're going to
29:03
pick to go with a a longer contract going forward and you know two seemingly similar teams working on two uh the same data set we came up with a totally different uh recommendations than the other team did and what what what was
29:20
the actual differentiator between the teams was that we had five medical degrees on our team not just not just a bunch of data scientists but data scientists plus medical experts and in every step of the way that you're building um these uh these knowledge
29:36
graphs and designing these algorithms you're interfacing with medical expertise to make sure you imbue it with clinical understanding with biological rationale of how this is actually going to work and how to interpret the the
29:49
typically really messy medical data and so if you think about what you know the The Matrix that we're producing this this uh heat map of 3,000 drugs cross reference was 22,000 diseases creat 66 million possibilities and we then score those
30:09
possibilities from zero to one and and normalize them across you know the whole landscape so that's a that's a that's a a tricky thing to do is you know disease uh drug a for disease X compared to drug B for disease Y how do you compare the
30:26
possibilities of each of those 0 to one so we create that normalized score and then we start looking at the highest scores and then filter down from there to say okay of all the of all the highest possibility opportunities here
30:43
highest probability of success opportunities here which ones are going to impact patients the most and which ones can we prove out quickly and efficiently in a lowcost trial with a few am of patients and high signal so we can do this in 3 to six to 12 months births and suppos a five-year trial times and the the thing to think about back to the uh comment about we need medical expertise highly integrated uh with what we're doing is that even if
31:10
you take the top thousand scores there you're still in the 0.001% of the highest ranking of scores and you and you now you got to pick amongst your thousand to down get down to the top five to get down to the top one what is my first shot on goal going
31:26
to be that better be successful for all the things that I'm working on here and and it better help patients and it really better work so uh the AI can't do that you need a a really smart uh you know head of translational translational
31:44
science to make that last sort of decision that what's going to go into patients and and how how it's all going to work yeah I'm a big believer in the human and the loop use of AI particular in very complicated and sophisticated uh
32:01
research like this you absolutely at least in my opinion uh I agree 100% with you you need human experts in the loop uh that can use the tools which are incredible tools absolutely but yeah as you rightly point out you can't just
32:20
leave it to the tools because there are a lot of hallucinations you mentioned um the uh 100 years from now we probably be looking back and saying oh my God we tested drugs on actual living human beings what what are your thoughts on uh
32:39
where are we rather than your thoughts I would like to ask you where are we in our ability to take somebody's entire genome create a digital twin and then do experiments with those drugs in silico as opposed to in actual real life
32:58
we are we're getting close we're not there yet um and you know so right now when I say I'm map we're mapping you know 3,000 drugs vers 22,000 diseases I think in the future it'll be we're mapping 3,000 drugs to8 billion people and it would be you know okay layer on
33:15
in this knowledge graph that we're constructing Jim's exact genome and prodium and transcriptome and latest lab samples Etc to come up with a drug that is fit for whatever how however the disease expresses itself specifically in
33:33
in Jim's Body um but in terms of creating that digital twin that is a uh requires us to create the perfect Knowledge Graph the knowledge graph that is is effectively the exact representation of of reality of of of the human body and and we're just not there yet there's so much unknown uh there's a guy uh named Larry Hunter kind of one of the the the fathers of of of This datadriven research and he uh I don't know if he coined the term but he
34:05
uses it and I love this it's called the you have the genome the transcriptome and the proteome and all these SS and there's also the ignorome there's this there's this area of of of biology uh that we simply don't know and we don't know what we don't
34:22
know we don't even know the shape of it when we look at the graph so uh I I I really I really like that and and and the work that we do the compiling of all this information all these continued assays and and and Drug Discovery
34:34
efforts all work to continue to uh elucidate that that area that that ignor room so we can eliminate it and ultimately make these perfect digital Twins and therefore make these perfect uh predictions I love that term and it it
34:49
seems to be uh part of the basic DNA of human OS the the ignorance uh the ignorance part um the idea uh of extending the research I'm I'm interested in for example have you considered like using uh medical chat Bots with existing patients where
35:13
they could take uh an interrogation or a conversation Andor uh do a metaanalysis of all the online Health forums where people are talking about yeah I have disease X they put me on uh drug y uh here are the complications or work
35:35
perfectly for me is that type of metaanalysis and aggregation of a more colloquial way of people just talking to one another uh about the particular disease could that be helpful absolutely I think um the way I think about it
35:53
though is less consumer facing and more research focused uh uh so I'll tell you kind of and I'll maybe give you a step back and further like how are we using large language models so um one way is to actually digest and interpret all of
36:13
medical literature so imagine so you've got pmed and pmed Central those are public resources for uh papers that are not behind pay walls or anything but then you have other organizations like ele these uh medical literature
36:27
publication organizations that have 7 million extremely high quality highly curated life science um scientific papers and we're working with them to use large language models to read effectively every single paper and extract all those biomedical Concepts
36:48
that I described before so if it ever mentions you know a protein or or a drug or a disease Etc not just extracting the the concept but also it the Rel ship that is described and understood within each paper so this is clearly something
37:02
that a human could never do but to read every single one of those seven million papers and extract out all that knowledge but not just uh understand it uh and then be able to report it back to you in a chat byot interface but to be
37:14
able to extract it in a tabular fashion that shows all the specific relationships between every biomedical concept that is embedded in all those papers so that it can then be layered into a Knowledge Graph pretty incredible then in addition to that once you have
37:30
the knowledge graph built if you recall previously when I was describing these these nodes and these edges in these semantic triples well semantic triple is effectively a sentence so could you then interpret this knowledge graph as a body
37:43
of language that is that is connecting long strings of sentences that that describe how things uh are are connected with each other and their relationships additionally in terms of you know using these kinds of models you
37:57
can train them not just on Words and text uh you know as you know they they they work by predicting the next word in a sentence but what if you train them on amino acid SE sequences and could then ask it to predict the next amino acid
38:15
and in a in a protein sequence or you trained it on patient Journeys the uh the the events that occur in in in a patient's journey through their disease longitudinally over time and then predict the next sequence that might
38:32
occur in that Journey based on what the inputs on the on the previous part of their Journey were so the the these models aren't just used specifically in language they can be used in other clever ways and and we're exploring
38:43
those those now I think I think uh in terms of the chatot interface example of that I think it's really interesting to use it there's company called atrio Health atrp that has designed uh a way to uh query large medical record databases
39:01
using a chatbot so you could you could ask it to scour 160 million patients medical records and as a physician you could say have you ever seen a particular patient that has this particular medical history and has this disease and taken this drug what was what was the outcome and it will scour all of uh existing medical uh Records to see what is happening in the real world not just what's a hypothesis like in coming out of a textbook but what
39:32
actually happened what's what what what what is the reality show we call that real world evidence so there's a lot of different uh clever uses of these models and it's not just chatbots I think that uh another use case that we've explored that we think
39:47
might be advantageous and helpful would be when we were discussing various use cases in the medical stack uh what about theide AA of AI simply generating null hypotheses and Publishing publishing them to an electronic database that
40:06
would be available to all researchers uh you know sort of knowledge via negativa do you think that that would be valuable in and intersect with the work you're doing yeah this is now we're getting to some really sci-fi
40:20
stuff here where the where they or use AI not only to to uh you know digest and understand medical literature but use it to generate its own hypotheses and then if you combine that with robotics and automated Labs it could potentially even
40:39
test out those hypotheses see if whether it was correct or not utilize the output of them to feed back into its own knowledge set to then refine its hypothesis generation and rinse and repeat until it actually advances science on its own so that that is
40:54
definitely in the realm of possibility and not very far off that's like within 10 years which uh I I often uh just remark that what a time to be alive I mean like the knowledge explosion that we are are undergoing right now and the
41:12
Innovations and and whatnot certainly I you know I'm 63 and this is the the most I've ever seen in my lifetime have you ever read uh waight but why the the blog yes I've actually had him on the podcast a couple of times oh Tim irvan yeah oh
41:32
that's great yeah he wrote a great uh article or really blog post it was almost 10 years ago now and it was on AI and it was on you know it was it was reflecting on the concept of the of the of the hockey stick and sort of the
41:46
compound growth and and AI capabilities and where are we now we're at the just the bottom of the edge of the hockey stick and and we're starting to feel it you know it's it's it's starting to happen faster and faster and faster and
42:00
it's almost like we're kind of like clinging to the edge of the this the side of the Cliff of of the hockey stick that it's starting to accelerate and it's only it's only going to come faster and it's going to and I think he created
42:12
this image uh in his in his uh blog post where he's like there's a group of people being looking off in the distance like wow looks like AI is kind of coming here it should be here soon and all a sudden it's like it's over it's over
42:25
there you know so it feels like we're we're in that moment yeah yeah the other thing like I in doing some reading around this in anticipation of talking with you like there's all sorts of I mean like it makes me just feel like a complete idiot
42:41
right as I'm going through it all and I obviously which is why you need the specialized knowledge that you referred to earlier but how do you how how is it to interact your classic case here you're a medical doctor you have an MBA
43:01
and you have uh a a pretty deep understanding of AI and machine learning how do you as a medical doctor translate it into tech for the AI researchers is that a is that a problem that's a great question and I think you uncovered or or you touched on
43:24
something that I think is going to be uh an interesting transition of of what our Workforce needs and what is going to be most valued um at Quantum black we actually called this role a translator so you use the word how do you translate
43:41
uh that that knowledge and so there's this person that needs to exist and we're actually hiring multiple of these people in my organization um that needs to be the synthesis point of the medical uh knowledge and medical jargon and the
44:00
technical and machine learning knowledge and and jargon and being able to understand both move information across the organization think strategically and elevate it you know to to the executive team so that is a uh position that's
44:17
going to be needed in every industry so you know AI Plus for me medicine AI plus whatever else for any other industry is going to need these strategic think thinkers that aren't necessarily doing the coding but at least understand it
44:30
enough to know when to use Which models where um and then and then can think and communicate clearly so that these Concepts can be Advanced and and people can that are making the Strategic Investments can make the right ones um
44:45
so it's it's it's a rare skill and I think it's going to be highly valued going forward and for people that are out there trying to figure out you know what they want to do in life uh you know you don't necessarily have to be a Cod
44:57
but if you can speak the language and then you can be a great communicator and think strategically then there's going to be a place for you yeah I agree and I I would of course take it a step further and once that person is found I would
45:11
like try to record their every interaction and have the AI train on them so that the AI itself could be improved in the communication skills of that unique and special individual so true so true um just knowledge capture
45:28
in organizations is a really interesting uh concept on its own I our our head of data engineering um is an incredible guy he also comes uh from from Quantum black uh and uh he is having us put every bit of our uh all of our notes and knowledge
45:49
and uh Team meetings and and everything else into something called notion it's one of these you know uh you know organizational tools and the incredible part about it is that then you can there's an embedded in notion an AI um
46:04
function where you can ask it anything and as long as you as long as you're good about putting all of your organizational content into this platform um you can uh it'll answer it as if it's like chat GPT uh but with full knowledge of the internal workings
46:21
of every cure so our U our head of data engineering Pascal brm he's he really spearheading that and it's been it's been an amazing uh I feel like advantage to our organization to have that quick understanding of everything that's going
46:34
on and anyone can ask it a question and how do you as a nonprofit uh obviously I see the reason you're a nonprofit makes total sense because you're not competing with any of the four profit but um who who are your main uh supporters uh donors and and tell our listeners and Watchers how they could get involved how how they could make donations to every cures thanks for asking that question yeah so again we're nonprofit because we want to build the
47:06
world's best AI platform and we need the best data set to do it to save as many lives as we possibly can with drugs that already exist so since the drugs already exist I it's kind of a funny thing I say we're the the smallest and the biggest
47:22
farmer company in the world we're the we're the biggest because every single drug that already exists is in our pipeline we're the smallest because we don't own any of them and so and then and then we take those drugs and we go after diseases
47:37
that are totally neglected by the pharmaceutical industry so it's by design has to be a nonprofit um and so uh you know our backers then are actually kind of interesting groups because the typical person that wants to
47:52
fund or support a a a medical nonprofit is usually going after a disease disease that has affected them or a loved one and so they're laser focused on on figuring out that disease we're agnostic to drug and disease we go where the data
48:07
tells us to go that's going to save the most lives so if you're interested in just having the highest Roi on your dollar to save human lives and you don't necessarily care what disease it was then we're the right group for you or if
48:20
you're interested in solving a systemic issue in healthc care and how R&D is done in drug discovery and you want to think that big then we're the organization for you so that doesn't always fit everyone's philanthropic intentions but and it
48:36
requires us to find really big thinkers or or people that just really care solely about saving human life and relieving suffering from from disease and so if if if either of those you know fit what you're looking for you can just
48:49
go to our website it's every cure.org we have a a donate uh function or just reach out to me personally my email is Grant at every cure.org and right now our our backers are uh some some really great philanthropies uh Chan Zuckerberg
49:07
initiative laa Hill uh Arnold Ventures um the Carolyn Smith Foundation uh uh Newar at uh at Flagship pioneering the founder of madna uh he he's he's supported us uh and some others that have remained Anonymous so we've had some some really great um uh
49:28
philanthropic uh supporters and we we could not have gotten this far without them absolutely not um so we're so grateful that they've taken a chance on us in early days um and then we've also been really really grateful to have
49:43
secured a an agreement with arpa h this is a new federal agency uh Advanced research projects kind like DARPA but for healthcare um and this was launched by the Biden Administration with $3 billion to fund uh Innovative technological advancements in in the
50:04
delivery of Health for our society uh and so we secured a an agreement with them that should uh provide us with around $50 million over three years to build out this artificial intelligence platform uh but additional funds are
50:17
needed to like I said before take those insights from the platform and actually deliver them into patients so the additional money is needed for validation in labs and then actually uh performing the clinical trials which as
50:31
you know can be an extremely expensive Endeavor so uh every every dollar counts we're an extremely Frugal organization we we rely on free office space and and uh and and as much support as we can get um but even then we are still one of the
50:47
the the best funded drug Discovery AI efforts uh around so we're really thrilled to be in the position that we're in to save lives yeah I think it's an amazing uh I idea and and you're executing against it uh incredibly well I will absolutely
51:04
support it and we will include a link to your website uh for show in in the show in the show notes I just think it's uh it's just such an amazing idea and project and I love how focused you are on we're just listen we're we're going
51:24
to go for the low hanging fruit we're to try to save as many lives as possible I I just don't I mean like I've heard a lot of pitches in my life and that's one of the strongest ones I've ever heard I'm glad to hear that thank you so much
51:38
for that support Jim I just really appreciate it yeah absolutely what what what was the biggest surprise as you uh did the research um what what uh What uh emerged as a solution to a disease that had you kind of as a doctor and your colleagues
51:58
scratching your heads and thinking how how can this drug be working for this yeah there are there's a few examples of that like a a uh a a flu drug that seemed to relieve uh dementia symptoms and they only figured it out because it was in
52:18
the enough people that had dementia were also getting the flu so you got to be to able to dig it out of the research at least empirically not just biologically via connections but empirically you have to have people actually taking the drug
52:31
to find the real world evidence but I think is more shocking than these sort of Novel um uh novel associations is the fact that sometimes there's good research that shows a drug is going to work for a disease that is published in
52:47
the literature and it just sits there because people just don't get access to it Vis other researchers and visions don't know was ever published so that was the case in in angio saroma uh that we found it was a pd1 inhibitor uh was U
53:06
was suggested to maybe work in in angio saroma and then a patient came to us that had angio sarcom and said what do I do well that paper had been published three years prior and it was sitting neglected in in the medical literature I
53:22
call this we call this latent knowledge that that if if the physician community at large had really read and adopted would have known and so thousands and thousands of people died while we actually already knew the answer and now it's been elevated so
53:40
that it's part of the guidelines and and people get prescribed this drug but um but that's a solution or a problem that we're solving as well because when you compile that knowledge into one space and you generate the hypothesis it
53:53
utilizes that that Laden knowledge so to me that's the most shocking thing is is that is that sometimes we already know the answer but it just hasn't been adopted yet wow and that kind of leads me to I there there are so many
54:06
externalities involved here like you see people being for or against a particular uh regime or drug uh therapy for reasons that have nothing to do with its efficacy and or as we saw during covid how do you cut through all of that
54:26
extraneous noise uh and just focus on on solving the problem yeah well we actually did something called the corona project during covid where we uh we evaluated the evidence of every single drug that was attempted in in Co amazingly I said
54:45
before there's only around 3,000 or so FDA approved drugs actually a little less than that five to 600 of them were attempted in covid wow so like like a like a sixth of all drugs were tried for Co it was the greatest drug repurposing effort ever
55:03
globally and they probably ever will be ever again so um there's just a ton of noise when and and an effort like that and when you combine it with the politicization of the time it was just kind of a nightmare for for for the
55:17
public messaging of of what was working and so we we tried to solve that by evaluating the evidence uh it was over 75 some thousand different papers that was compiled just a short amount of time and you can you can if you uh you know
55:30
if you Google the criminal project and and and you look up our evidence you you'll see us grade from you know a a uh you know to F the um the the level of evidence that we think is was been compiled from literature in in in an
55:45
objective way so to me the answer to that question is is just do your best to uncover objective reality with the proper scientific research tools that we have yeah and uh keep your eye on the prize exactly because there just seems
56:01
to me uh we are we are increasing signal exponentially but we are also increasing noise maybe even more expon I know that that's acious looking at it but the amount of noise that we are now confronted with is massive uh much more massive than uh any
56:24
humans of the past certainly but but uh projects like yours bring the signal uh hopefully uh to to the front of the analysis so so what would be uh I know that you've got a very long-term plan here but what would be the next
56:44
Milestone that like when you achieve it you're like okay now we're now we're getting much much closer to where we want to be yeah so just a general time frames of of what we're doing here is you know so we signed the uh government
57:03
agreement on April 1 of this year so just earlier this month um and we or that or that's when the the contract launched so so now we're really after to the races because that was a much more significant amount of funding than than
57:18
than previously previously the funding had really helped us um build sort of the the beginning stages of our platform kind of a beta version of the knowledge graph and the and the scores and the ranking systems but now we're building
57:31
this at scale uh with a large team and uh and and an infrastructure that is uh professionalized um and so year one is all about building the database and the knowledge graph uh compiling all of the public sources compiling all the additional uh private
57:53
and proprietary data sources and generating some of our own data so by the end of year one we should have constructed this biomedical Knowledge Graph and we're actually going to give that away to uh researchers in society
58:07
so we're going to have everyone can access that so they can uh explore it query it do their own research on top of it uh we hope that's a real Boon to the scientific and research Community to have access to that resource year two is
58:21
developing the scores that heat map that that I that I mentioned to you by the end of year two we actually plan to publish that as well so a visualization tool that's Dynamic and navigable by both researchers and Physicians and even
58:39
patients um to zoom in on their drug of Interest or their disease of interest to say okay nothing else is working for this patient what else might be possible and have it actually and to zoom in on on their disease you'd see a rank ordered
58:54
list of all three th drugs and the probability of of each one of them to work for the disease of of of your focus uh and so that we hope is another Boon to researchers to say okay where's the heat if I'm agnostic drug and disease
59:10
where should I look how should I allocate resources if I'm focused just on one disease where should I look across the drugs to allocate resources and do further research or even help inform what I'm going to do for a patient and then once those scores are
59:22
built and and there and both the database and Knowledge Graph and heat map are released to the world that's when we start moving our drugs into the clinic the the every cure drugs that we select uh that we think are going to
59:36
have the highest impact on patients people can pursue all the other things that are in that graph on their own and we hope they do we hope that it it accelerates research globally but we're going to go for the ones that help the
59:47
most patients and so to me the the biggest thing I'm looking forward to is that first time uh that we uh confirm that a drug Works in a disease and it goes into that patient population and starts saving lives that's absolutely
1:00:01
what motivates all of us in the meantime you know we we actually use the graph and and the preliminary results to inform some of our thinking and and we've been able to help a a small amount of patients with with that but uh we're
1:00:14
uh we have to stay focused on on getting to this scaled version so that we can we can repurpose drugs really in perpetuity uh what what you know we hope to have 20 or so uh repurposing opportunities packaged and ready for clinical trials
1:00:30
and beginning to be validated by 2030 and then at steady state hope to have you know two to five drugs being repurposed every year that's a I know that didn't sound like much but that's actually extremely ambitious the entire
1:00:45
uh pharmaceutical industry as a whole approves maybe 50 some drugs a year and so you know if we're doing a tenth of that as as this little nonprofit that's that's pretty incredible you know even like a big one single large Pharma
1:01:02
company might do you know two to five in a year and that would be a lot for them so um but like I said there's this faster more scalable more efficient way to to get these drugs to patients and we hope to to Really automate that and and
1:01:15
and and build an engine that works in perpetuity that just doesn't rest until every drug is unlocked to treat every disease that it possibly can yeah and and that brings in the uh idea of using this as open- Source AI which I I think
1:01:32
uh I'm a huge fan of because of everything you've listed but also because it allows for the tinkerers and cognitive diversity it's it's like the old I quote a lot uh the quote about you know you could have somebody who's the
1:01:46
smartest human in the world they uh know the most about everything but you could never ask them to make a list of things that would never occur to them and and and so those things often do occur to others and the only way to capture that
1:02:05
in the knowledge is through the open source through the ability for many many people to work on uh the project and suggest contributions and obviously the contribution will stand or fall on its efficacy uh in in the trials but uh I
1:02:22
think that that too is just an incredibly Noble and and worthwhile project so you you kind of achieved the hatrick here at Grant like uh you if uh if there are current Saints you might get nominated and your partners I sh that but I you know
1:02:45
there's there's a there's a concept I ever heard this I I I someone told me this a few years ago there's this Japanese concept called eeky guy have you heard of this I have yeah this reason for being being this this this
1:02:58
the center of the VIN diagram of of what are you good at what do you love what does the world need and what can you get paid for if you can find something that hits all four of those and you can be right in the middle of there well then
1:03:10
then youve really found something uh that that can get you out of bed every day and and for me this is this is certainly what that is and and C certainly for my co-founders as well so uh we're we feel just really privileged uh and blessed to be part of an
1:03:26
organization like this and I'll tell you what it has been a delight to build this organization because it attracts really smart but really mission-driven people and when you have everyone that's aligned rowing together in the same
1:03:45
direction and you've got this you know North Star of these patients that are suffering that we are helping and we're relieving that suffering and saving their lives um the the the Team Dynamics and the culture is really fabulous uh so
1:03:58
we've been lucky to be able to trct a lot of really high quality um High performing people that that care about this and are and and work really well together to to achieve uh what we're trying to accomplish here yeah it
1:04:12
reminds me of the quote that probably wasn't from Mark Twain although it's often attributed to him he said the two most important days of your life are the day you were born and the day you discover why so so I think grant that you and your
1:04:29
team have discovered the why and the and are at that perfect uh pitch point of the eeky guy uh graph I think that this is an amazing thing you're doing I will support it I urge all of our listeners and viewers to to do the same this is
1:04:46
lifechanging stuff and uh I just think it's so incredible how far you've come and yet how far you have to go and and why you need the support well if if you listen to the podcast in the past you know that we do ask a final question where we make you the emperor
1:05:03
of the world for just a day you can't kill anyone you can't put anyone in a re-education camp but we do hand you a magical microphone in which you can say two things and the two things that you say will incept all 8 billion plus
1:05:20
people of the world for whenever their next day is when they wake up they're going to wake up and they're going to think I've just had the two greatest ideas but this time I'm going to act against both of these things I'm going
1:05:33
to start acting immediately against these two things what are you going to incept in the world's population Grant wow that's a big question and I and I I hadn't thought of this yet so give me a moment this is sure that's uh and I'm
1:05:47
I'm incepting it as if uh like it's like the movie Inception I went into their dreams and I implanted this idea okay exactly exactly okay gotcha gotcha you know I think there's one idea is this this notion that that you are
1:06:05
they and what I mean by that is you know when when David and I first started thinking about cman disease and whether we could solve it or not whether we could cure his own rare disease we we first said well surely they must be working on
1:06:24
this and did a little research and we said oh my God there's really there's really no one out there so where are they and then we look at like could we be there you know and there's no reason whatever problem you're thinking about whatever it is
1:06:44
that you can't be they you know they don't assume that they's out there in a little Factory tinkering working on all the problems in just the way you think they should be they're not you know you can be they so so get and go and go do
1:06:56
it I love that one and and I think I think the other thing that I would say is that the world responds to Clarity of vision so you know the eeky guy concept is one thing but it's like you can't have youry guy until you have the
1:07:15
clarity of vision and and when and when you tell the world exactly what you need exactly what you're doing and exactly why you're doing it resources fall in line if you if you tell the world I don't really know what I want to do like can you can you help
1:07:32
me can you tell me you know no one really gets behind that you might get you might get a mentor or two but resources are not going to start flooding your way when you when you when you take a stand when you when you have
1:07:45
the courage to say okay I'm going to put my line of the sand and say I'm going to do this and I'm going to take a risk to tell everyone that I know that I'm gonna do it I'm gonna make the ask I'm gonna I'm gonna ask that guy I know should
1:08:01
that would be the world's best in this to work with me I'm gonna ask this guy that I know's got the funds to finance it and I'm and I'm gonna tell the world I'm going to do this that's when things start happening and no one's going to do
1:08:14
that for you you have to make that decision to to to to express that Clarity and have the courage to stand behind it and refine the idea as it goes but but but until you tell the world what you need the world's not going to give it to you wow
1:08:30
I'd love both of those they might actually rank among among some of the best inceptions that that I've ever had I've had lots of guests Grant this has been amazing uh I wish you absolutely the best of luck with the project as I
1:08:47
said I will contribute I will urge others to contribute uh I I don't know like Clarity of vision boy you make it very easy to want to give you money so congratulations congratulations I'm not H thank you Jim I really appreciate it