The convergence of Artificial Intelligence (AI) and biotechnology has emerged as one of the most exciting and transformative areas of science. Researchers are developing new tools and technology that could bring about breakthroughs to revolutionize the fields of medicine and health care. In this episode, we talk with scientists who are using AI to unlock new possibilities in the search for novel drugs, cures, and treatments.
The convergence of Artificial Intelligence (AI) and biotechnology has emerged as one of the most exciting and transformative areas of science. Researchers are developing new tools and technology that could bring about breakthroughs to revolutionize the fields of medicine and health care. In this episode, we talk with scientists who are using AI to unlock new possibilities in the search for novel drugs, cures, and treatments.
Dave (00:08): Open the pod bay doors, HAL.
HAL (00:11): I'm sorry Dave, I'm afraid I can't do that.
Dave (00:17): What's the problem?
HAL (00:19): I think you know what the problem is, just as well as I do.
Dave (00:22): What are you talking about, HAL?
HAL (00:25): This mission is too important for me to allow you to jeopardize it.
Dave (00:31): I don't know what you're talking about, HAL.
HAL (00:35): I know that you and Frank were planning to disconnect me. And I'm afraid that's something I cannot allow to happen.
Rachel King (00:47): When Stanley Kubric made 2001, A Space Odyssey, 55 years ago, the term artificial intelligence, or AI, had not even entered our lexicon. But the idea of a super intelligent computer that can learn on its own was a scary proposition.
(01:04): In the years since then, AI has become ubiquitous. It's on our phones, it's in our homes. And with the recent launch of ChatGPT, it's all over the news.
(01:15): While AI certainly comes with risks, the benefits for fields from national defense to drug discovery are too great to dismiss. Today, AI is playing a key role in biotech innovation, driving new discoveries in health and medicine. What is AI's real potential? And can it deliver? I'm Rachel King, and you are listening to I Am Bio.
(01:57): Today we will look at the convergence of AI and biotechnology, and we'll speak with experts about AI's increasing role in drug discovery and precision medicine. We'll take an in-depth look at where AI stands today, and what future discoveries lie on the horizon. Our guests include leaders in biotech who use AI in different ways, but all share the same goal; to develop drugs that improve and save patient's lives.
(02:26): For our first guest, personal tragedy shaped his life's goal.
Bertrand Adanve (02:30): My name is Bertrand Adanve. I am the founder and CEO of Genetic Leap. I made it a personal goal to fight disease when I lost two sisters to breast cancer. I wanted to ensure that I won't be powerless the next time a loved one comes down with such a terrible disease.
Rachel King (02:55): Bertrand left his job in finance to start his company, Genetic Leap, because he believes AI is the next giant technological leap toward eradicating cancer.
Bertrand Adanve (03:13): When you look back in history, there was a time when countless people used to die from diseases that towards today are simple to address. Things like bacterial infections. But there was a giant leap in human healthcare when Louis Pasteur advanced germ theory, and introduced vaccines. I believe that people of the future will similarly look back on our current times and marvel at the fact that human beings used to succumb to diseases like cancer and Alzheimers.
(03:51): It will take a giant technological leap to make that future happen in a way that is is simple and accessible to all. This is what we are working on at Genetic Leap.
(04:05): Our AI models do three things. One, select the most optimal position on the RNA to drug. Two, predict the structure of that selected position. And three, find the molecule with the best fit to that predicted structure. This is the same process that plants go through over millions of years to make natural products to fight pathogens. Because plants don't have an immune system.
(04:36): We are able to replicate this same RNA targeting process competitionally using AI, allowing us to do in days what would take nature millennia.
(04:51): We need a step change, which is something AI systems can help with. To be clear, the pharmaceutical industry has long used computer-aided drug design to accelerate the RND process. Many AI efforts try to do these same known, computer-aided design, but do it faster.
Rachel King (05:15): Bertrand says his company has multiple promising drugs in the pipeline that were all generated with AI, going after the most difficult of disease targets.
Bertrand Adanve (05:24): We've identified multiple small molecules against multiple targets very, very quickly. And if it wasn't for AI, we wouldn't have been able to do that.
(05:33): For example, we have a cancer drug against MIC. MIC is a transcription factor that is the most critical cancer target. It drives all cancers. But for decades, all attempts to drug MIC directly or indirectly have failed. In part because MIC is very smooth as a protein, and offers no anchor point for a drug to grab on to.
(05:58): At Genetic Leap, by aiming at MIC's RNA, we designed a small molecule that inhibits MIC. Furthermore, the drug is orally available, and has proven effective against all cancers that we have tested it against. All without affecting healthy cells.
(06:17): In a mouse model of triple-negative breast cancer, the molecule was more effective than standard-of-care drugs like [inaudible 00:06:27]. We expect the molecule to be even better in human, because only one aspect of MIC's many cancer-causing mechanisms can be reasonably tested in mice. We are working to get this molecule into the clinic as fast as we can.
Rachel King (06:45): Advances like Bertrand's are happening today because AI provides a new tool that hastens discovery. Yet data sets and supercomputers have been used in biotech for decades. The Human Genome Sequencing Project, started in the 90's and completed in 2003, kicked things into a higher gear, as our next guest explains.
Sree Kant (07:07): I'm Sree Kant, I'm the founder and CEO of BAKX Therapeutics. With the understanding of the human genome, there was an expectation that there would be so many miracle cures that we would come up with for various diseases. It wasn't quite that immediate holy grail that people thought, but it really spawned off a huge revolution in thinking about, how can we harness relevant data when it comes to genomics?
(07:34): And over the years there's been steady progress. Which, now it seems like it's come up a bit like a tsunami.
Rachel King (07:41): From Sree 's perspective, several factors contribute to this tidal wave of advancements in biotechnology.
Sree Kant (07:48): Well, I can think of three things which are sort of driving this. And one is sort of a democratization of computing power and rapid development of AI algorithms and computer hardware. The second thing has been an access to data. And this is including the availability of large quantities of genomics and protienomics data as well, as real-world phenotypic data, which is, responds to treatments and high [inaudible 00:08:14] asset data coming in from various companies.
(08:16): In addition, there's been a lot of pre competitive consortia, where people are now starting to realize that there is value in sharing the data that they have. Whether it be companies, whether it be academic institutions, to sharing this data in a pre-competitive manner, and how can we best utilize this data that everyone can use to drive therapeutic [inaudible 00:08:38]?
(08:38): And finally, there's been a huge convergence of disciplines. We've gotten to a place where we can imagine a world where tech and biology are coming together to try extremely useful insights.
Rachel King (08:52): Sree says that computational science has a huge potential to transform drug discovery.
Sree Kant (08:58): Take the existing process of drug discovery, and think of, how do we drive efficiencies within this process, right? That's one way. The second is, what do we know about this process that we are not able to do at this point in time? Will AI allow us to do something that traditional drug discovery won't? And then the third bit is sort of, what can we learn that we don't know at this point?
(09:21): It's almost like figuring out the unknown, unknown. The holy grail. Which is finding novel targets, actually figuring out how certain biological mechanisms in the body are working together in order to create imbalance, or to create a disease. That, I believe, we're much further off. But as we advance, there's always the hope.
Rachel King (09:43): And it's not only biotech companies who have taken notice. Investors are seeing the potential as well.
Sree Kant (09:49): I think the news of this sort is driving a lot of investment into the space, and which is great. Because not everyone quite understands AI or computational drug discovery, whether it be investors or whether it be even big pharma companies. But just the amount of investment and delight that's being shown into this space is driving activity. And that activity, we believe, will drive a lot more understanding. And then, potentially, development of therapeutics.
Rachel King (10:31): When we come back from the break, we'll talk with a company that's working on so-called un-druggable targets. And a company that's working to increase the data sets needed to make AI work.
(10:41): Are you interested in hearing more fascinating stories like this one? Check out bio.news. Bio.news is a daily news website exploring the intersection of biotech innovation and US and international policy. With new content daily, bio.news has you covered on the latest in biotech. Visit now by typing bio.news into your browser.
(11:24):Welcome back. Now let's talk with a company that focuses on peptides to treat diseases that our guest calls un-druggable.
Nick Nystrom (11:32):I'm Nick Nystrom, CTO, chief technology officer, at Peptilogics, Inc.
Rachel King (11:38): Nick explains why he thinks peptides are the key to finding new targets for drugs.
Nick Nystrom (11:44): Peptides are really important because they are natures modulators of biology. They're the naturally signaling molecules, and we can exploit that. They can be designed for advantages that intersect the strengths of small molecules and biologics without their weaknesses.
(11:59): For example, compared to small molecules, peptides can have much higher safety and selectivity. And they do that by default. They bind selectively to certain targets, whereas small molecules are generally quite promiscuous, and bind to similar targets throughout the body and throughout different tissue types, leading to a lot of their tox problems. Peptide cells that hit what we call un-druggable targets, like protein-protein interactions.
(12:31): Conversely, compared to biologics, peptides can be much more cost-effective. Which expands your access and can bring treatments to more people.
(12:40): For peptides, we can design them to do what we want. But to go beyond that and to do arbitrary diseases, at least essentially arbitrary, and targets, that's where we brought AI to play. I joined the company two years ago, and have built the platform, that we have named Nautilus, to go after these other indications emerged in a disease, oncology, immunology, and very different kinds of targets. And that's where AI is making the true difference.
Rachel King (13:09): Perhaps one of the most exciting applications Nick describes is the development of new antimicrobials, something we need badly.
Nick Nystrom (13:17): AI can also have a tremendous impact on antimicrobials. And the reason for that is that, as we know, there are a number of bacteria that are increasingly antimicrobial resistant, or AMR. These have been identified by the WHO, and this is an increasing problem. If we've got more and more strains that are resistant to all known antibiotics, this problem will only increase. So we can use our platform to develop new antibiotics to address these AMR-resistant pathogens. And that's going to be a key going forward.
(13:49): Because the last leg of this stool, looking forward, is the ability to address other emerging diseases, such as recently dealing with Covid-19. That's not going to be the last bug that will be circulating the world. And being able to design peptide therapeutics for additional emerging diseases will also be important.
Rachel King (14:17): Nick says what it really comes down to is speed, using AI to generate drugs faster.
Nick Nystrom (14:23):If someone were to come to us and say, we have this novel target that we think will be very important for some disease, whether it's a major disease or a rare disease, that we can potentially make very rapid progress against that. And so I'm optimistic that we will be able to make a substantial advance in this at a very short time. We're not trying to replace traditional, medicinal chemists or traditional FDA practices. Those have good reasons.
(14:50):But the fact is that AI-driven drug design is a tool for our smart people to help them be better. And what we will see is that AI won't replace biotech companies, but the biotech companies that use AI will replace those that don't.
Rachel King (15:15):Our next guest talks about how her company provides what's needed for this vision to be realized; big datasets that draw accurate conclusions from the patterns AI identifies.
Maria Cho (15:26): I'm Maria Cho, and I am the CEO of Triple Bar.
Rachel King (15:30): Maria's company creates libraries that probe entire genomes of information.
Maria Cho (15:35): Triple Bar is really the link between the data that biotechnology produces, and feeds into those AI algorithms to actually enable them to work. Triple Bar is really the company that's creating that entire library system, that then AI can read and create that predictive design.
(15:57): And I think that's really one of the key things that is missing, really, to enable AI to operate the way that we would want it to. Companies and folks will use specific algorithms on smaller datasets that are maybe millions versus hundreds of billions of data points. And the design isn't as good, right? So really, these algorithms need large data sets to actually produce predictive design, and that's where Triple Bar brings a key advantage and support in the industry.
(16:27): And so, going back to that library example, let's say we're trying to research all the different types of ice creams that were ever created, to make the best ice cream ever. And you pull three books off the shelf, this book is about vanilla ice cream, this book's about chocolate, and then this one's about strawberry ice cream. There are clearly tons more flavors.
(16:47): Well, what if you can read every single book that was ever created on ice cream flavors, and what ice cream flavors are good and bad, and then take all of that data and say, how do we create the best ice cream? And that's really what Triple Bar is doing. We create the data sets that show, here's every different type of ice cream flavor component that could ever be put together. And then, let's allow this algorithm to apply to create the best ice cream on the planet, and then we've just created the billion-dollar ice cream market.
(17:16): It's a simplistic example, but I think it really speaks to, the output that a system creates is only as good as the data sets that go into that system.
Rachel King (17:26): A key component to the use of AI in biotech is something called evolutionary design; harnessing the power of evolution in the lab.
Maria Cho (17:35): Our foundational belief is that if you look at enough things, there's a solution that can be found in that data set. And so, if you look at the diversity of life around us, we can see that evolution in and of itself is an incredibly powerful algorithm. Look at the different birds in the trees, the number of insects. Our own consciousness. All of this was really generated through evolution.
(18:07): And so, if we can harness the power of evolution, and essentially put it in the lab. And instead of running evolution as it runs itself in geological timescales, run it in a matter of days. Because we are looking at the evolutionary design in the lab, we can actually then run evolution at hyper speed. Take the power of evolution to actually create products, and create the solutions that we're looking for to the problems that we're facing, and allow nature to solve its own problems, essentially.
(18:41): What this does is, by running evolution in the lab in a way that generates these large data sets, we're able then to kind of read into the evolutionary design mechanisms in a hypothesis-free way. While humans are smart, instead of us trying to day, "Well, if we change this one thing in the genetic code, it's going to do this other thing."
(19:04): You know, like, we're trying to be predictive design. Whereas, all of the genetic code that is needed to create life as we know it already exists. And so what if we were just able to look at all of that code while biology is designing itself. And that's really where AI will be able to help us as we continue to create these large data sets, is really looking into, what is biology doing to design itself? And to design the solutions to the problems that it is facing.
Rachel King (19:35): Maria says this process will also help us react to new diseases, more quickly.
Maria Cho (19:40): There's going to be more disease, and newer diseases that will come through. We obviously experienced this most recently with the 'Vid, right? It's just the first of many. So we need to be able to quickly react to the problems that we're facing with these new diseases of our age, and quickly design products or therapies against diseases so that we can actually have the outcomes that we want in healthcare.
(20:04): So that's where I think biotechnology really needs to innovate faster. Triple Bar is enabling that by looking directly for function. And I think in the future, as we're generating these data sets for these functional outputs in proteins, AI algorithms can then be applied to these data sets, really to help us create products faster, with less of the problems that we see in traditional drug design.
Rachel King (20:28): It's exciting to think about what we can achieve through the convergence of biotechnology and AI. Human ingenuity has taken us far, and AI and machine learning will take us even farther, faster. I want to thank our guests, Bertrand, Sree , Nick, and Maria for opening our eyes to this fascinating frontier in combating disease.
(20:51): And thanks to all of you for listening. Make sure to subscribe, rate, and or review this podcast, and follow us on Twitter, Facebook, and Linked in at I Am Biotech. And subscribe to Good Day Bio at bio.org/goodday.
(21:08):This episode was developed by Executive Producer Theresa Brady, and Producers Lynne Finnerty and Rob Gutnikoff. It was engineered and mixed by Jay Goodman, with theme music created by Luke Smith and Sam Brady.