In this inaugural episode of Biotech Breakthroughs, equity research group head Tim Lugo speaks about the impact of artificial intelligence and machine learning on the biotechnology industry, including the ways in which AI is revolutionizing drug development, impacting valuations, and the potential risks that lie ahead.
Podcast Transcript
00:25 - 00:45
Shantha Ozgen
Hi everyone. On today's first episode of Biotech Breakthroughs, we welcome Tim Lugo, partner, and group head of biotechnology, Equity Research. Back in May, Tim and his team dug deep into the ways in which AI and machine learning are impacting the biotechnology sector through a new report titled Putting the Tech and Biotech: A Deep Dive on AI in the Sector.
The report provides a very thorough analysis on the benefits and risks associated with AI, including the anticipated impact of drug discovery and drug development, along with the overall risks associated with the new technology. It's a thoughtful, in-depth report and one that provides us with plenty to talk about. With that, Tim, I'm going to hand it over to you.
So maybe to kick things off, we can go pull back up to 30,000 feet and you can tell us what inspired you and your team to really write this report and tell us how you think about AI's impact on the sector.
00:01:15:28 - 00:01:40:06
Tim Lugo
Sure. Thank you, Shantha. So with this report, we really just try to dig into AI machine learning. It's impacting every industry that we're seeing. It's obviously gained a lot of exposure starting, you know, earlier in the year. But we've been working on this report for probably about a year and a half. We have been hosting a number of private companies as well as public companies and webinars.
And just really digging into where is AI machine learning impacting biotech, where is it impacting the drug discovery and the drug development process? Our report is extensive, as you mentioned, 71 pages. So please, everyone listening to this. You don't need to read it all. Doesn't need to be consumed in one sitting, you know. Please reach out to Shantha or myself if you have any questions.
We really just wanted to dig into where is AI machine learning impacting biotech specifically right now? Where is the opportunities? And also combat some things that we continue to see in popular press on Where is it really going to have a major impact, in our opinion? We dug into that pretty in-depth in our report and also highlighted a lot of interesting companies in the space which we think are having to become the leaders in the AI drug development space or are about to become leaders with this.
02:36 - 02:54
Shantha O
Tim, let's just start with where the real need and crucial impact of AI and machine learning is in the biotech industry. Your report puts a really helpful framework around how it's really around the drug discovery process. So maybe we can talk about where the challenges are over here and how this technology can really help us overcome them.
02:54 - 03:24
Tim L
Sure. I think that's where AI and machine learning is going to have the largest impact in biotech it’s just higher quality candidates. New biology and also higher quality candidates. Those are the two areas where historically biotech has always created a lot of value. That's historically where it's thought recombinant protein technology created a lot of value or mRNA creating a ton of value or where antibody technology created a lot of value.
So we think that AI machine learning is just going to be a continuation of these large step functions we've seen in the biotech sector in the past. Biotech in general, as well as clinical development, generates a ton of data. We're talking head to petabytes of data. A lot of small molecules are screened against new targets. All of this creates data, often in very messy forms.
That is where machine learning and AI will probably have its most impact is really through all of this data, you know, taking out nuggets and also trading feedback loops, which ultimately should lead to higher quality preclinical candidates and less likely candidates to have less likely clinical failures as it progressed through phase 1 to 3 clinical trials.
So I think that's where these technologies can have the largest impact in the near-term and that's really where we see a lot of interesting companies kind of percolating up to the topic.
04:28 - 04:48
Shantha O
Tim, you guys talk about four ways in which AI and machine learning can really help with a drug development process here. You know, you talk about screening on the biology side, helping the screening on the chemistry side, but streamlining the candidates. There's obviously ways in which AI machine learning can help us on the clinical trial and then obviously the eventual commercial opportunity here.
Can we actually go through each one of these and dive into how you are viewing the landscape for AI and how machine learning is impacting in these different ways?
04:57 - 05:22
Tim L
Sure. I think first and foremost, we're seeing a lot of interesting new targets which are being identified from AI machine learning techniques. There's a lot of genomic and clinical data that has been produced over the past decade. AI algorithms and AI technologies are really being hardest to draw out these new targets from looking at these large datasets.
These genomically sometimes can be in locations which are very unobvious to a human researcher, sometimes very far away in the genome. And you know, I've heard from some of the companies I work with who are applying these technologies that, they're finding new targets that are just not even going to be conceived from a human researcher.
Then they're taken, then validated by human researchers, obviously. But a lot of these new targets which are being drawn out from algorithms are completely novel and would not be obvious at first glance from traditional genomic technologies. So I find that very exciting. Those are any time you have new targets in discovering new biology that usually leads to a ton of value because setting up the walls for competition and just making those data sets internally within companies that much more proprietary.
Secondly, you know, new chemistry, AI algorithms are being employed for new candidates, screening candidates and also how these candidates interact with targets. I think we're going to get much higher quality, better targets, and better interactions with this AI discovery compounds. So, any time you have a more specific and more sensitive assays, you're going to have less off-target effects by these small molecules and usually just safer, more potent compounds.
I think AI and machine learning is definitely having an impact there with some of the early-stage clinical candidates we're seeing coming out of seven colleges and maybe less impactful right now, but probably more impactful going in the future is going to be its impact on clinical trial design and maybe even monitoring of some of the clinical data coming out of trials.
So those trials could be tweaked during the process. This is an area where it's very much nascent or these A.I. machine learning companies, although some of them are employing it as we speak, we're still really in the first inning of AI/machine learning being applied on the clinical side, in my opinion.
And then ultimately where we don't see it right now but I think there's a lot of opportunity in the future on the commercial side, post-marketing studies, you know, having AI algorithms and machine learning feedback loops, really processing all the data coming out of the real world setting and ultimately leading to, you know, maybe better therapies in the future and also picking up side effects that could be popping up in the real world, which are not completely obvious until you get out into, you know, millions of patients and hundreds of thousands of patients not just have a small, defined clinical trials, which is characteristic of early stage drug development.
08:09 - 08:31 Shantha O It sounds like we're just at the tip of the iceberg on where this technology really gets used across the spectrum in the sector. Would you agree with that, that that's really this is just early innings at this point. And you know, what we're really looking at right now is just a sliver of where this technology can then broadly be used across the whole sector and in a lot more different ways than just the drug development process?
08:31 - 08:52
Tim L
Completely. And, you know, there's only a handful of these companies right now, a lot of them are being funded now, early stages right now. Be kind of leaders in this space were found maybe ten years ago, but it's still a very early-stage part of the industry. Still, something that I think on the large pharma side being explored.
08:52 - 09:20
Tim L
A lot of large pharma and a lot of large biotechs are going to have machine learning groups and bio computing groups, almost siloed within the large organizations. And then on the share play side, a lot of those companies are still very young. They still are coming up with their first drug targets. Their first compounds are, you know, in the clinic. And ultimately AI machine learning is about iteration, about feedback loops.
And, you know, in the biotech sector that leads to clinical success or failure. So those feedback loops really don't come full circle until five, ten years sometimes after drugs are discovered. So we're very much in the early stages right now. Those feedback loops are just being refined as we're speaking. Probably over the next few years, we'll still going to be an early part of AI and machine learning really impacting the sector.
09:44 - 10:03
Shantha O
So, we do know that a lot of what AI and machine learning relies on is the quality of the data. And so, it cannot obviously be understated how important it is in the drug development discovery process. Can you just maybe walk us through, and I know your report talks about this also, what makes data so crucial in drug development? What are the challenges with it right now and what are ways in which AI and machine learning can help us overcome those challenges?
10:09 - 10:47
Tim L
Data is absolutely critical. A lot of companies are founded on unique platforms that are proprietary to them and the data being produced from these platforms needs to be mined and needs to be found efficiently. It needs to be stored as well, very efficiently and then mined efficiently. And that is probably one of the largest challenges for some of the companies that were founded on non-tech enabled technologies transitioning to the current era where tech must be deployed throughout their organization.
It also builds a moat for some of these pure-play AI biotech companies, which are bounded based on the idea that they will be producing a large amount of data. They will be industrializing biology and producing just massive amounts of data which will be mined. And again, going into the feedback loops which AI/machine learning are so adept at pulling nuggets out of.
So that's one of the kind of issues we pull out on our port report a lot is the value for these pure play companies and how different it is versus the large pharma companies which are just really coming around to building out this infrastructure internally, and also building out the scale internally. And not to mention that it's probably going to be a lot of cultural issues which occur when the typical white lab scientist is approached with some novel insights that come out of just in silico technologies and in silico assays, which AI is very good at making choices that that are, you know, not emotional obviously.
And we have people who've been dedicating, you know, 20 years plus of their life into a certain framework of drug discovery. So, it's going to be interesting to see how those kind of tugs and falls happen within organizations and how drugs are prioritized internally. One thing we do know which is occurring is the volume of data is only increasing.
Integrating that with very messy decades-old data requires algorithms and a lot of powerful computing. So, this is only becoming more important as biotech kind of progresses on to its next stepwise, I guess, drug discovery.
12:34 - 12:56
Shantha O
As we are still talking about data, do you think this is a rate limiting factor near term for how quickly this technology of AI and machine learning really gets used by the industry or all of this is happening simultaneously where the quality of data is improving, and it goes hand in hand with increased use of the computational technology that you're talking about?
Or do you think this is really just something that's going to take a lot more time, particularly in drug development versus maybe some of the other industries where it's being used?
13:04 - 13:29
Tim L
It's going to take time just because drug discovery takes time. One thing we talk about in our report a lot is we push back a bit where it is commonly thought that AI and machine learning techniques are going to shorten cycles or drug development. But really, drug development occurs mostly during the development time. It occurs a lot in the clinic and the clinic. You still need to enroll patients. You still need get sites up and running. You still need to find appropriate patients for clinical trials. That's not really going to be impacted by AI and machine learning in the near term. What we will get is a lot of probably higher quality preclinical candidates and so shortening, they're very much the front end of the process, probably shaving years off of that timescale but that could be up to 20% faster.
But the clinic still have a lot of blocking and tackling of drug development. And then on the regulatory side, you know, that isn't going to be stormed anytime soon. The FDA isn't thought of as the most flexible agency when it comes to regulation for good reason. I mean, we are developing drugs that are being dosed to millions of people and their mandate is to only approve safe and effective therapies.
I'd say the timeframe is still going to be a decade type of timeframe for biotech, even if you are deriving these compounds through AI and machine learning at the end of the day we will probably just have higher quality compounds and, you know, probably higher hit rate, which is very important for investors and the industry in general.
14:40 - 15:02
Shantha O
So maybe we can take a pivot here. Your report says that this is time to maybe rethink how valuations for these tech-enabled biotechs that are in the space should be thought about. You also say there's no one size fits all, that there's no one way to succeed in this industry. There are multiple different variations where different companies can obviously exist and different strategies that can exist.
Maybe walk us through what some of these variations are that are successful or that can succeed. How should we think about the companies and these various different verticals? And then I guess a crucial question comes around valuation. I mean, how should we think about ultimately how to value this vertical, which is more tech-enabled versus the more traditional biotech companies?
15:21 - 15:54
Tim L
Valuations are going to be difficult in this area and in small cap, biotech valuations are always a bit of a flash point and a bit of a controversy on the long and short side. But in AI derived biotech, it's going to be it's going to be interesting to see how this plays out over the next decade because arguably if we're talking machine learning, we are talking about iterations, we're talking about failure ultimately lead to a more valuable internal data set.
But in the traditional small cap biotech world, and Shantha, you know this extremely well, anybody who invests in biotech knows that really well, the first setback for a small company is usually a death blow to the valuation. You know, we always see gaps down in the 50% range. We look every day on the Nasdaq, which companies are the most volatile?
It's always small cap biotech. So, it's going to be interesting. It's going to be, you know, to argue when a company's first candidate takes a step back, is the investor that had been primed on valuing that as a major setback for valuation? Would the investor really come to the conclusion that that data set within that company is now more valuable?
It's going to be hard to argue that, but really that is likely what's occurring. It's going to be very interesting to see how these companies are not just valued when everyone is excited about AI, everyone is excited about machine learning being applied to biotech. These are cutting-edge technologies in health care as well as in tech. But when things go a little bit wrong at these companies to see what occurs with the valuations, especially in the current environment. We are in an environment where small-cap biotech is just not as highly valued as it was back when interest rates were at 0%.
That's a very widely known phenomenon. Yeah, valuation is going to be extremely interesting to see how that how that plays out over the next few years as these AI driven companies get some success, but also some setbacks in the clinic. Other struggles out there or other issues, which I think some of these companies could face are definitely, on the regulatory side, the FDA has been very slow to change its review process in general.
You look at areas like even in accelerated approvals or biomarkers that is constantly in struggle for, you know, nonclinical endpoints, which may allow for quicker drug development and quicker approvals for those to be adopted by the agency. That's always a source of friction between the AI companies and the regulators. It will be interesting to see how in silico experiments are included in the regulatory process or even allowing for IND filings and then eventually for drug approvals.
18:27 - 18:49
Shantha O
So maybe we're on the topic of companies that are looking at this space. Your report also talks about, you know, obviously they're smaller cap companies, they're private companies, but then there's also big pharma that's involved in drug development. Can you just share some perspective, at least like on the big pharma side, what you've heard? Because I think one of the things your report also mentions is their challenges.
And I think you alluded to it yourself earlier in sort of culturally integrating these technologies. So your view may be the difference between how this is being used and smaller cap startups versus the more traditional established companies. You know, where is the use really happening or is it happening across the board?
19:07 - 19:52
Tim L
I think in the larger companies, it's still you hear one or two companies really deploying these new techniques, really embracing them internally, putting some I'd say, significant focus behind machine learning techniques. And this idea of industrializing biology. I'd say it's not broad though and definitely in the large pharma sector, what is I think probably more common is you have some sort of computational biology team, and they are tasked with coming up with new insights, with, you know, publications, with tracking the players in the field, especially the smaller, more nimble players in the field.
But my feeling is they're not really being integrated into the entire process. I think these small companies have a real chance, either A) becoming next kind of biotech flagship type of companies like we've seen across the sector. You know, we often seeing companies that jump in that, you know, 1 to $3 billion range become a $50 billion company over, you know, the short time span.
So we do see the potential for some of these smaller players to eventually become one that's biotech leaders and join the ranks of the other large biopharma in the sector by really leveraging these technologies. It's going to be a very interesting area to track, I'd say, over the next 5 to 10 years. Again, we're still very much at the beginning of this and B) computing power looks to only be improving.
That's something that is a real tailwind for this specific subsector in biotech, whereas, you know, computing power continues to have its j-curve in adoption and being deployed. Drug development still continues to be slow, while expensive. So, there's a lot of you know, there's a lot of room for improvement with all of these technologies.
21:14 - 21:30
Shantha O
So I know you talked about a lot of ways in which AI and machine learning can help improve and, you know, facilitate better drug development. But the discussion wouldn't be complete if we didn't at least touch on some of the risks. You flagged some of them already for us on the regulatory side, quality of data.
I know your report talks about it extensively. Anything else you think that's, you know, should be on people's minds as they think about how this technology gets deployed in something that's, you know, as integral to our livelihood and our daily lives as drug development?
21:45 - 22:13
Tim L
Definitely on the regulatory side, we already talked about the FDA is how the number of adcoms, specifically digging into this area and holding outcomes is one thing, but really integrating it into the regulatory process is another. So we'll see. I think that's a risk is maybe the agency is as quick as adopting their own opinions. They have come out with these outcomes internally and kind of embracing these new technologies.
There's a clear risk that, you know, garbage in garbage out type of data is very much something that everyone in this field is probably battling with on a daily basis. We do have these massive data sets which are available either publicly or as a proprietary dataset. But some of these data sets are over ten years old, some over several decades old.
When you're looking at medical records from large country-level type of medical record datasets, these are decades old. It's very hard to sift through that and to really come up with high quality data out of these large legacy sets. And ultimately, this is not a chatbot. This is biotech. These are developing drugs which are going to be dosed to humans and will hopefully, you know, cure, and manage diseases.
But, maybe similar to gene therapy, we do have some setbacks where we will see some side effects occur with these therapies. And I know that the gene therapy space that has a lot of hesitation. That is definitely a risk that there's almost a broad painting of the AI machine learning biotechs because when you really dig down into them and we do a great job of that, our report especially, I should note, my associate, is definitely not sparing in the field.
He did great job in digging into the various companies in the sector. All of these companies have very different approaches. So, there is a real risk that we just kind of broadly paint, you know, a sector in a very unrefined manner and that they are, unfortunately, could when there is a setback with one company, could be a spill over, at least to investors, it being a whole sector.
So that's definitely a risk. And I should note that there is a real push around diversity in data sets, you know, diversity and patients that are being included in the clinical trials. And when you look at a lot of these legacy datasets, there is a real risk that diversity, which has been occurring here over the past few years isn’t reflected in these datasets which are decades old.
Hopefully we're not just developing therapies for subsets of subsets, hopefully we’re broadening, you know, developing new therapies for the most representative population.
24:34 - 24:40
Shantha O
Thanks Tim. We’ve touched a lot of many of the highlights from your report today. I appreciate you joining us and look forward to doing this again.
24:40 - 24:53
Tim L
Thank you, Shantha. I know it's extensive, but thank you for reading it. And again, thank you for hosting me for the inaugural Biotech Breakthroughs podcast. It's been fun.