Latest / The BioHub - by Avetix / Episode 143 - Jenny Yang - CEO & Co-Founder at Outpost Bio

Transcript

Episode 143 - Jenny Yang - CEO & Co-Founder at Outpost Bio · The BioHub - by Avetix

• 0:06
  Today I'm joined by Jenny Yang, founder and CEO of Outpost Bio
• 0:11
  Jenny has spent the last decade working at the intersection of
• 0:14
  machine learning and personalized medicine, with
• 0:17
  experience spanning genomics research in Canada, CRISPR work
• 0:21
  at Stanford, and a PhD at the University of Oxford.
• 0:24
  Focused on clinical machine learning, Outpost Bio is a is
• 0:28
  building an AI platform to understand how human microbe
• 0:32
  interacts with drugs, food, and other interventions to unlock
• 0:36
  new insights into personalized medicine.
• 0:39
  Jenny, it's great to have you on the Biohub podcast today.
• 0:42
  Thank you so much, Amber. Really excited to be here.
• 0:45
  It would be great for our listeners if we could start off
• 0:48
  with your background, which really sits right at that
• 0:50
  intersection of machine learning and biology.
• 0:53
  What first pulled you into applying AI to healthcare and
• 0:56
  drug discovery? So I started undergrad back in
• 1:01
  in 2014, and this was really around the time when machine
• 1:03
  learning was just making a lot of breakthroughs.
• 1:05
  So you're hearing about it everywhere and I knew that it
• 1:09
  was a skill that I wanted to at least learn about if it was such
• 1:12
  a hot and upcoming field. And over my second year summer,
• 1:17
  I just decided to create a small program online on my own, pull
• 1:24
  data from a public repository of of cancer genomics data sets and
• 1:29
  then build like a really, really simple program that just
• 1:31
  organized this. I then cold emailed 20 different
• 1:35
  professors and then one professor, Professor Stephen
• 1:39
  Jones at the Genome Sciences Center, he actually offered me a
• 1:42
  job that summer and that's when I just fell in love with being
• 1:45
  in in the field. So specifically I was working on
• 1:48
  machine learning approaches for personalized oncology.
• 1:51
  I just really, really enjoyed my time at that lab working with
• 1:53
  different bioinformaticians, meeting clinicians that were
• 1:56
  thinking about implementing AI into their their work streams to
• 2:00
  to help patients. And yeah, just, it really
• 2:03
  started there. I just kept coming back every
• 2:06
  every year to that lab and then decided to jump into, you know,
• 2:10
  a PhD after that. And yeah, have just been
• 2:14
  continuing down that journey. Amazing, and a lot of precision
• 2:18
  medicine has historically focused on human genomics.
• 2:21
  What made you interested in the microbiome?
• 2:25
  Yeah, I, I think that there's been some incredible
• 2:28
  advancements over the last 30-40 years on the human genomics side
• 2:32
  and we've seen incredible kind of discoveries there, especially
• 2:36
  in the personalized medicine space.
• 2:37
  We can now say if you have this particular genetic mutation, it
• 2:41
  might mean you have this particular subtype of disease.
• 2:44
  And then now we can even choose targeted therapies and are
• 2:47
  actively designing targeted therapies towards those
• 2:50
  mutations. But I realize over time, like
• 2:53
  even though genomics has helped us solve and answer a lot of
• 2:57
  questions, it hasn't quite fully solved the question of why two
• 3:01
  people with the same human genomics, same subset of disease
• 3:05
  on the same therapies are still responding very differently to
• 3:09
  therapies. And I really believe that part
• 3:13
  of that missing piece is all the other DNA that makes up, you
• 3:17
  know, what it means to be human. So we're not just our own human
• 3:21
  DNA. We're really these ecosystems.
• 3:24
  We, our bodies have trillions of microbes inside of us, on us.
• 3:28
  And those microbes also affect how we experience health and
• 3:31
  disease. And they're so, so diverse.
• 3:34
  So between you and I and everybody in the world, we share
• 3:37
  99.9% the exact same human genomics.
• 3:41
  But if we look at the microbes in and on our body, we can be up
• 3:44
  to 90% different. And when it comes to
• 3:47
  personalized medicine, whether you take a medication or your
• 3:52
  intravenously, it's going to pass through your community of,
• 3:55
  of microbes and that will affect how you how you respond to those
• 3:59
  therapies. Same with like skin
• 4:00
  interventions. There are microbes on our skin
• 4:02
  and when we take interventions that are cream based, our
• 4:05
  microbes will also interact differently to someone else's.
• 4:09
  Yeah, that's really interesting. And thinking about that
• 4:11
  difference between two people. And so for people and our
• 4:15
  listeners who may not be familiar with the space, how
• 4:18
  would you describe what Outpost Bio is building and the problem
• 4:22
  that you're trying to solve? So really what I like to say is
• 4:25
  that Outpost Bio is trying to decode human microbiology.
• 4:29
  We do this both in the wet lab and in the dry lab.
• 4:33
  So essentially we're trying to generate data sets of community
• 4:36
  of bacteria X plus molecule Y leading to biological outcome,
• 4:40
  said. And then on this data set, build
• 4:43
  AI tools to allow other people to be able to run these type of
• 4:47
  analysis for their own applications as well.
• 4:51
  OK. And there was a big wave of
• 4:54
  microbiome companies a few years ago, but many of them struggled
• 4:58
  with that clinical translation and some manufacturing
• 5:01
  challenges. What do you think has changed
• 5:03
  now that makes this the right moment to revisit that
• 5:06
  microbiome space using AI? Yeah.
• 5:09
  So there definitely was a huge wave maybe like 6-8 years ago of
• 5:13
  a lot of microbiome companies coming out.
• 5:15
  And I think one of the biggest challenges of having jumped into
• 5:19
  the microbiome field back then was that the technology was just
• 5:23
  not at a point where it would have made it very tractable to
• 5:28
  create tools in this space that other people can use.
• 5:33
  So earlier I mentioned the point that everyone in the world can
• 5:36
  be up to 90% different in their microbiome makeup.
• 5:40
  So that's already a huge combinatorial space where you're
• 5:43
  looking at a lot of different combinations of bacteria that
• 5:46
  you want to study. Another huge complexity of the
• 5:49
  microbiome is that it's not just diverse, but it's also very,
• 5:54
  very large and complex. So there's only about 20,000
• 5:58
  human genes in the in the human genome.
• 6:01
  But if you look at our just our gut microbiomes, there's 150
• 6:05
  times to 250 times more genes in our gut microbiomes.
• 6:09
  So not only are we looking at a very diverse space, but it's
• 6:11
  also very, very large. So the computational tooling
• 6:16
  needed probably wouldn't have been ready until now.
• 6:19
  Now we have incredible cloud computing infrastructure, new
• 6:24
  machine learning architecture. So this allows us to actually
• 6:27
  dig into such a diverse and complex high dimensional space.
• 6:33
  Another technology that I think has really pushed for the field
• 6:35
  of the microbiome field forward is metabolomics.
• 6:40
  So you've heard of genomics, transcriptomics, proteomics.
• 6:43
  Metabolomics is really the the new and upcoming omic.
• 6:46
  So this is an omic that allows you to understand biochemical
• 6:51
  processes that occur. And back, back when micro, the
• 6:56
  first wave of microbiome companies came out, a lot of
• 6:59
  them were focused on metagenomics or metataxonomics.
• 7:03
  So really looking at the composition of data that of
• 7:05
  microbes that exist. But really to fully make the
• 7:10
  microbiome actionable, you want to understand what these
• 7:13
  microbes are doing, not just who's there, but what they're
• 7:16
  actually doing. And that's what metabolomics has
• 7:18
  allowed us, allowed us to do. And what I like to say is that
• 7:22
  the field of metabolomics is at a point where it's ready for
• 7:26
  scaling. So there's a lot better
• 7:27
  resolution. Costs have gone down
• 7:29
  significantly. So really it's in a point where
• 7:32
  genomics probably was 15 years ago.
• 7:34
  It's not quite a commodity yet, but it is at a point where we
• 7:37
  can now scale it up for machine learning purposes.
• 7:40
  OK. Yeah, that makes a lot of sense.
• 7:42
  And and many AI native biotech companies these days talk about
• 7:47
  their proprietary data sets as their competitive advantage.
• 7:50
  At Outpost, you're building both wet lab and computational
• 7:53
  capabilities. How important is that data loop
• 7:56
  for building the models that can actually predict biology?
• 8:00
  It's super important for us because microbiology itself is a
• 8:04
  very, very challenging field, and to generate standardized
• 8:07
  protocols for it is a challenge on its own.
• 8:11
  So for example, data generated in one lab, it's hard to
• 8:14
  reproduce at another just because there's so many
• 8:17
  variables that could be changed. That's why we wanted to make
• 8:21
  sure that we had control over the wet lab protocol because I
• 8:24
  really do believe in the machine learning world.
• 8:26
  It's garbage in, garbage out. If you want to actually generate
• 8:30
  biologically relevant models, you need to have a gold standard
• 8:34
  data set for this. So for us, we want to generate a
• 8:39
  high quality data self, the data set ourselves, build AI models
• 8:43
  and then because we have a wet lab, we're able to then validate
• 8:47
  some of the findings that we can predict through our models.
• 8:51
  So it is really this loop that we're trying to operate on.
• 8:54
  And when building that AI model around the microbiome, then how
• 8:58
  important is it to capture diversity across different
• 9:01
  populations? This is definitely top of mind
• 9:04
  for us. So like I mentioned, we can be
• 9:07
  up to 90% different. There's a lot of literature out
• 9:09
  there that points that to differences in even geography
• 9:13
  and where someone grew up, the foods they they ate.
• 9:17
  So we're making sure that when we source human derived samples,
• 9:20
  we definitely source from diverse areas around the world.
• 9:24
  And when you're building a company, especially within AI
• 9:28
  and biotech, it can be quite challenging.
• 9:30
  But you mentioned that you've had a positive journey so far.
• 9:34
  What has your experience been like as a woman building a
• 9:37
  company in this space? So I think I've been extremely
• 9:41
  fortunate throughout my career, like ever since my undergrad and
• 9:44
  Steve Jones bringing me into his lab through my PhD as well as
• 9:47
  other industry experiences I've had.
• 9:50
  I've been really fortunate to work with people that I feel
• 9:53
  like have judged me on the quality of my work rather than
• 9:58
  if I'm I'm female or, or some or like other traits that are more
• 10:04
  like demographically based. So I've I've definitely had
• 10:08
  extremely positive experiences, but I am very aware that
• 10:11
  sometimes I'll be maybe one of the few females in a room.
• 10:15
  And although I feel like we're getting better at closing that
• 10:20
  gap, there is still a ways to go.
• 10:22
  And I want to make sure that both I tell my story about
• 10:25
  having positive experiences, but also leave the the path wider
• 10:29
  for people who come come after me.
• 10:33
  But I, yeah, I think something that I have learned over time is
• 10:36
  if there's something you're passionate about, put in the
• 10:38
  effort to really master it. Because at the end of the day, I
• 10:42
  think your work and the output that you put into the world
• 10:44
  will, will be the most important.
• 10:47
  Or, you know, I personally feel like you should be the, the most
• 10:50
  important. And if there are people who you
• 10:53
  know would judge you on demographic features, they might
• 10:56
  not be the the right people to spend spend time with because we
• 10:59
  don't want to be recycling those kind of world views.
• 11:03
  Yeah, that is such a good insight as well about who you're
• 11:05
  working with and you know, when it comes to Outpost bio and the
• 11:10
  upcoming milestones, what are you targeting across the next
• 11:13
  year? Yeah, really the next year is
• 11:15
  going to be focused on scaling up.
• 11:17
  We have a high throughput wet lab approach that we've already
• 11:22
  optimized and now we're just really focused on scaling.
• 11:25
  We've started benchmarking initial machine learning models
• 11:29
  and yeah, are focused on demonstrating that they work
• 11:32
  really well in specific verticals.
• 11:35
  We'll also have some initial pilots and proof of concepts
• 11:37
  with some commercial partners. So really focused on scaling up
• 11:40
  and validation of both the science, the tech, and the the
• 11:44
  commercial opportunity. You know, that's really
• 11:46
  exciting. It sounds like you've got some
• 11:48
  really big milestones coming up with the business.
• 11:51
  And just to kind of round us off, Jenny, our usual last
• 11:54
  question at the Biohub podcast, if you could have dinner with
• 11:57
  any scientist or innovator who, past or present, who would it be
• 12:01
  and why? So I think for this question I
• 12:04
  would have to answer with Richard Feynman, scientist,
• 12:07
  physicist, my initial when I when I first joined uni, I was
• 12:14
  an engineering physics major and slowly moved my way into more
• 12:19
  biology and and AI. But physics has very much in my
• 12:23
  my history. I also really like this quote
• 12:26
  from him. That's about how this is
• 12:30
  paraphrase, but it's something along the lines of where like
• 12:32
  the easiest, the easiest people to fool are are ourselves, and
• 12:37
  we must not fool ourselves. And what he's really talking
• 12:41
  about is being rigorous in our science and like the science
• 12:48
  that we put out, or at least how I've interpreted it.
• 12:50
  And I find that really important in the fields we're in because
• 12:53
  we really are trying to create better health outcomes for
• 12:58
  people, not just patients, but really overall health for
• 13:02
  people. So we need to be really rigorous
• 13:04
  on the work that we're doing and the work that we choose to make
• 13:08
  public for other people to use because it has a huge impact at
• 13:11
  the end of the day. So we don't want to just believe
• 13:14
  we're getting good results when we see them because we want to
• 13:16
  believe that we should really put in the time and effort to
• 13:20
  rigorously validate. There's also pieces on him that
• 13:24
  talk about how he's a funny person and I, I feel like humor
• 13:28
  is and, and laughter is the best medicine.
• 13:32
  So it's it sounds like he would be a a great person to to talk
• 13:36
  to. Nice.
• 13:38
  I absolutely love that. Well, thanks so much for joining
• 13:41
  us on the podcast, Jenny. I'm really excited to see where
• 13:45
  Outpost bio goals in the next couple of years and follow your
• 13:47
  journey. Thanks everyone everyone for
• 13:50
  listening and please like and subscribe.
• 13:52
  Thanks so much, Amber.