How is AI changing the drug development process?

Sascha: [00:00:02] From Equity Mates media, this is the dive. I'm your host, Sascha Kelly. We hear a lot about artificial intelligence, but as a non-scientist, I'm the first to admit I don't understand it at all. Coming out of Covid, one of the most remarkable things about the past two years was the speed at which we developed drugs and vaccines. Many in the field are saying that the pandemic was the ultimate test of artificial intelligence and machine learning, and the technology passed with flying colours. It's Wednesday, the 30th of March, and today I want to know how artificial intelligence is changing the drug development process. And what does the future look like? To help me understand this new and exciting fields and one that I have got to say, I have absolutely no understanding of. I'm joined by my colleague and co-founder of Equity Mates, Alec Renehan. Alec: [00:00:55] Hi, Sascha. Good to be with you today. It feels like we're both maybe a little bit out of our circle of competence, but I'm excited to explore it with you. Sascha: [00:01:03] I know we're going to be swimming upstream today, but that's the exciting thing. We're going to find out something that we didn't know before. So artificial intelligence, machine learning, drug discovery, where do we even start with dissecting a topic this big and complex Alec: [00:01:19] as to non-scientists? We're going to tackle it in a very overly simplified way. Sascha: [00:01:24] I love that Alec: [00:01:24] if there's any scientists out there listening, tearing their hair out, we apologise. Well, you know what? We don't apologise. We're going to attack this in our own way. But I think this topic is something that everyone, scientists and non-scientists alike should be really excited by and to start with Sascha to frame this conversation. I want to take you back to the early 2010s and introduce you to 20 year old Alec watching vice documentaries while at university. Sascha: [00:01:50] And I've got to say 20 year old Alec doesn't sound particularly cool right now, honestly. Alec: [00:01:56] Thirty year old Alec not particularly cordial, but I remember in my university days watching a vice documentary on the coming crisis in antibiotics and how scientists were searching the jungle for new compounds to develop new antibiotics, to stay ahead of antibiotic resistant bacteria Sound bite: [00:02:15] in terms of finding new antibiotics. Nature offers the most promising new Alec: [00:02:19] compounds, and the predictions are dire, according to some experts. Antibiotic resistance bacteria could kill 10 million people a year by 2050, surpassing cancer deaths. So Sascha, as you can imagine, that scared 20 year old me Sascha: [00:02:34] that scares, you know, thirty something year old me so I can understand why scared you back then? Alec: [00:02:38] Do I look, it's always stayed in the back of my mind. This idea that we're fighting a losing race against antibiotic resistant bacteria, which is why I was so excited last year when I first learnt about a new antibiotic called holism. Sascha: [00:02:52] You've given me the perfect set up there to just say, All right. Tell me about Allison. Sound bite: [00:02:57] The first new antibiotic to be discovered in nearly 30 years has been hailed as a game changer. Alec: [00:03:02] Well, Allison was the first antibiotic identified with artificial intelligence and machine learning. Now it was a drug originally developed for diabetes, and it didn't do that well. It failed in testing. But in 2019, A.I. researchers at MIT found that it had potential to work as an antibiotic. In testing, it showed success against antibiotic resistant strains of different bacteria. Now, Sacha most exciting because its mode of action, a.k.a. the way it kills bacteria in the body, is different to most traditional antibiotics. Scientists suspect it could be difficult for bacteria to develop resistance to. Sascha: [00:03:41] I am still confused about how artificial intelligence is helping change the game. How is it working into this process? Alec: [00:03:49] Well, the process of drug discovery is complex and expensive. And Sascha, one thing we're going to learn in this episode is the names of drugs and drug companies are difficult to say. I think their names are outside my circle of competence. But according to Taconic, bringing a drug to market can cost approximately $2.8 billion and take up to a decade now. AI and machine learning can cut that cost and reduce that time. Sound bite: [00:04:14] Instead of looking for a needle in a haystack, we can use the giant magnet of computing power to find many needles in multiple haystacks on target. Sascha: [00:04:25] So you've given the numbers there, but how exactly does A.I. help in that process? Alec: [00:04:31] So there are a variety of rabbit holes we can go down here. Sascha, we Sascha: [00:04:34] love rabbit holes. Alec: [00:04:36] Scientists are currently exploring uses of light at every stage in the drug development process, drug discovery, pre-clinical trials and clinical trials. But today, I think we focus on the drug discovery process. And really, I want to focus on one word data because over the past few years there has been an explosion of data in bio sciences because it has become much. Easier and cheaper to collect sequences of genes and proteins and always just better at digesting huge data sets than humans, according to reporting from the Financial Times. I can quickly identify many compounds in a relatively short period and add a quarter of the cost of traditional methods. Sascha: [00:05:19] That's amazing. Can you tell me about these companies? Alec: [00:05:22] BioNTech, the $40 billion German pharmaceutical company that shot to prominence during Covid has invested in this space. Their CEO has one of my favourite quotes about A.I., he said. A.I. is like kids. You really have to teach them a lot until they start to come to conclusions. Sascha: [00:05:40] That's such a great quote. Alec: [00:05:42] That really is what I guess we're saying that with all this data, these big companies are teaching A.I. how to read the data. They're setting up the data to to be processed Obel by AI, and they're starting to yield results in drug discovery. Sound bite: [00:05:57] We can use machine learning to detect very subtle changes in cells that are produced by different compounds and thereby find out how compounds work. Alec: [00:06:04] So as well as BioNTech Pfizer, the $300 billion American giant has really been out in front of this space. In 2020, they realised that I couldn't screen its libraries of 4.5 billion commercially available compounds, so it invested in cleansing this data, making it A.I. friendly. And now I can scan this entire database in just 48 hours. Sanofi $116 million French pharmaceutical giant has embraced A.I. and is signing deals with two A.I. focussed start-ups. But Sascha does much more than to focus start-ups out there. Funding for these start-ups has risen 3800 percent in the past five years, now, surpassing $2 billion a year. And these venture capital backed start-ups are signing so many deals with Big Pharma. Sascha: [00:06:55] As soon as you start throwing billions with a B around, you'd know that the money is like a magnet just going to go straight into that space, right? Alec: [00:07:03] Like an absolute magnet. And we can't talk about ally and money flowing into a space without talking about the undisputed leader in A.I. and that is Google, a.k.a. Alphabet. The latest emblem for existential dread is Google's DeepMind project, which created AlphaGo, an A.I. programme that's become unbeatable at the most complex strategy game on the planet. Late last year, Alphabet announced it would use the research from its AIS subsidiary, DeepMind, to create a new business isa Morphic Labs. That's going to use DeepMind's A.I. to accelerate drug discovery, and they hope, well, I guess we all hope ultimately find cures for some of humanity's most devastating diseases. Sascha: [00:07:46] What a sentence to finish on there, Alec. And as we said, it can't be an episode of the dive without talking about Alphabet or Google. They seem to just pop up and every single story that we look at. Alec: [00:07:57] Well, to be fair, a lot of our research for these episodes is on their platform, so they deserve a mention. Sascha: [00:08:03] Look, we're going to take a quick break and hear from our sponsors, but afterwards we're going to go look at the returns from the investment in this space. And then also, I'm going to get you to look into your crystal ball and imagine what the future will look like. Welcome back to the dive, I'm Sascha. And today we're talking about artificial intelligence and how it intersects with drug development. I'm joined by my colleague Alec Ren, a hand and ALEC. Before the break, you explained how many of the biggest pharmaceutical companies in the world are investing in AI drug discovery. Our favourite Google popped up there as well as it alongside a couple of drug companies that I can't pronounce, so I'm going to be going to avoid that. But here's the big question are we actually seeing the returns from this investment? [00:08:53][50.5] Alec: [00:08:54] So to date, we haven't seen a regulator approve a drug discovered entirely by A.I.. We have seen A.I. find new uses for existing already approved drugs, but it still is early days. There are now 15 drug candidates designed by A.I. in clinical trials. But as we said at the top. Drug development can take more than a decade with all the different stages of clinical trials, so it'll be a little while before we truly say the returns. But Sascha, I want to give you an idea of the two ends of the spectrum. I guess you could say the best and the worst of what the future could offer. Sascha: [00:09:29] I love this. It's like a crystal ball. Or if you're a movie fan, the sliding doors moment of what's going to happen. So you know me, I'm an optimist. Let's go with the positive first. Alec: [00:09:41] When Covid took the world by surprise in January 2020, doctors struggled to treat patients. They only had a few clues on how this virus operated. So benevolent A.I. turned to this technology artificial intelligence machine learning to help they set their algorithms to work and manage to raid 50 million medical journal articles in just a few hours. How fast can you read, Sascha? Sascha: [00:10:04] Well, not that fast, that's for sure. Alec: [00:10:07] So over the next few days, researchers continued to deepen their understanding of how Covid operated, and they fed this information into the algorithm. In just four days, the A.I. had identified an Eli Lilly drug used to treat rheumatoid arthritis baroque in the tip apology. So we did say, though hard to pronounce. And this drug, this rheumatoid arthritis drug, was found to tackle both Covid and the body's inflammatory reaction to COVID. Sound bite: [00:10:38] We, benevolent and others have clearly demonstrated that in early drug discovery, artificial intelligence can help with better target selection and reduce the time it takes to get familiar to a candidate molecule. Alec: [00:10:52] Now, this was the first time that I had discovered a drug to repurpose. That was then used widely. Sound bite: [00:10:58] So I think if we look at the hype cycle, I think we've definitely gone over the peak of expectations and through the trough of despair. And now we're really on the beginning of the slope of enlightenment. So I think it's a Sascha: [00:11:10] really exciting time. So this is an example of how I went and researched existing information that we have already discovered and said, Hey, this drug is probably going to help this new problem. This isn't about developing a new drug from scratch in four days. Alec: [00:11:27] That's it. That's it's not about creating new compounds, although they're doing that, it's in clinical trials. This story is about taking an existing drug, existing compounds and finding new and novel uses for it. So a drug used to treat arthritis was repurposed and found to be incredibly effective against Covid. The World Health Organisation strongly recommends Perec natives use, and a UK government trial found it significantly reduces COVID deaths. Sascha: [00:11:55] Well, that's a great story. All the other ones as well. Am I tempting fate by asking if I can get some more good news stories? Alec: [00:12:02] Yeah, well, Sascha Covid was a real good news moment for this A.I. drug discovery process more generally of Celera, sorted through six million cells in just three days to find an antibody that could be mass produced in this case by Eli Lilly. That has helped more than one million COVID patients. Similarly, a supercomputer helped in Pfizer's quest for an antiviral that could be taken orally. And although I did not design COVID vaccines, it helped optimise them. Moderna's platform, the technology it uses for all of its vaccines, learnt from producing 20000 unique marinade sequences, which helped it design and manufacture the first batch of its COVID vaccine for testing in just 42 days. Sascha: [00:12:49] Wow, Alec, I think you can kind of start to get overwhelmed by the numbers when you list them off like that because they're just so huge and this technology is just reading through masses of information in such short periods of time. So I'm a bit reluctant to hear about it, but I guess I need to. I want to hear what could go wrong. What's the negatives about this technology? Alec: [00:13:11] As quickly as I can discover new drugs and cures, it can also use that same processing power to sift through biological. Research data to find, I guess you'd call it the opposite of a cure. Sascha: [00:13:24] OK, that sounds really ominous. Alec: [00:13:27] Yeah, this is a pretty ominous story. Collaborations Pharmaceuticals, a company based in North Carolina, uses machine learning to identify drugs for rare and neglected diseases. But its scientists were asked to contribute to a conference discussing the effect recent scientific breakthroughs like AI and machine learning would have on the Biological and Chemical Weapons Convention, which is a multilateral treaty banning biological and chemical weapons. So Collaboration's pharmaceuticals set their mind to the task. They trained their A.I. with a starting set of molecules known pesticides, known environmental toxins, and left it alone, basically to calculate how to adapt those molecules to become more deadly within six hours. The A.I. had discovered 4000 potential killer molecules, many of which could be used as biological weapons. Most scary, some compounds scored more highly for toxicity than any known biochemical weapon. Thankfully, the researchers withheld crucial details of the research method for security reasons. So it's not like you can pick up a journal article and repeat this process, but it does show just how scary this technology can be. Sascha: [00:14:42] Yeah, and I guess at least that uncomfortable question, which is if someone's done it before someone can do it again, right? Alec: [00:14:50] Yeah, that's it. But look, Sascha, as scary as that thought is, it's important to note that life sciences has always had the potential for misuse. Biological and chemical weapons could be made before A.I., and they can be made after A.I.. II doesn't change anything. It just makes everything a lot faster. Sascha: [00:15:09] That's made me feel a lot better about it, actually, because you do sometimes get caught up in the breakthroughs of the technology, but you forget the fact that humans have been breaking things and making things for generations. And so this is just the next step in this particular avenue of discovery. Alec: [00:15:27] That's it. We are entering a amazing decade for bio sciences, from gene sequencing and editing new therapeutics, m RNA vaccines. It feels like there has been and there will continue to be a lot of breakthroughs in the coming years, and all of this will be sped up and supercharged by AI and machine learning. So there will be risks and there are scary thoughts like these researchers at Collaborative Pharmaceuticals demonstrated. But I think on the whole, it's a really exciting time and a really exciting story. We're going to see some amazing breakthroughs. We're going to see some fascinating new and novel drugs, and hopefully all of it will lead to step change improvements in people's quality of life. Sascha: [00:16:10] Well, I think if we started with terrified 20 year old Alec 10 or so years ago watching vice documentaries and getting pretty nervous about the future of the world, hopefully we've put his and yours and mine our minds at ease after today's episode, and again, a call out to any scientists who've been listening, tearing their hair out. Don't just be angry. Write to us and come join us on the show. You can always ask us to talk about a story that interests you by contacting us at the dive at Equity Mates dot com and follow us on all of the social media channels. All those details, as usual in the show, notes below. All podcasters say it, but it really, really does help remember to write and review us and your favourite podcast app. It helps us discover ability, and we love being discovered by other people and subscribe so that the next time there's a new episode, which will be later this week, it'll be delivered to the moment it drops. Alec, thanks for helping me get up to speed with AI today. It was a fascinating topic. Alec: [00:17:12] Thanks, Sascha. I think we are still outside of our circle of competence, but it is a fascinating story. Sascha: [00:17:18] Amazing. OK, until next time.