Efforts to use artificial intelligence to discover drugs have been underway for about a decade but industry watchers are predicting an inflection point is nearing for investors, who have been looking for ways to determine how AI-first drug developers should be valued. AI and machine learning offer the potential to speed up the hunt for new therapies by more quickly identifying compounds to treat disease. There’s also the promise of making clinical trial phases more efficient by improving patient enrollment and processing insights quickly as the information rolls in from studies. More tangible evidence of these capabilities is now being demonstrated. A high-profile example has been the effort to fight Covid-19, which forced biotech and pharmaceutical companies to bring all their capabilities to the effort of discovering vaccines and treatments in record time. Lidia Fonseca, Pfizer’s chief digital and technology officer, has discussed the role the pandemic has played in accelerating digital advancements during several conference appearances over the past year. “We believe that Covid-19 has advanced these trends by as much as five years,” Fonseca said in a virtual fireside chat with McKinsey in January. “It’s not so much that these are new technologies, more that we are applying them at scale.” Key points for investors According to Deloitte’s latest estimates, it can cost $2 billion to develop a new drug. Artificial intelligence and machine learning promise to lower that cost by reducing development times and increasing success rates. More advanced algorithms, increased computing power and richer data sets are leading to more progress. While most biotech and pharmaceutical companies are using AI and machine learning tools, companies that are native to the space are about to hit an inflection point that will help investors value these companies. Boston Consulting Group said in March that AI-first drug developers have identified more than 150 small-molecule drugs, with at least 15 already in clinical trials. The capabilities that will occur when quantum computing is widely adopted are unimaginable now, Fonseca added. But even with today’s supercomputing power, Pfizer is able to use modeling and simulation to screen millions of compounds to arrive at potential drug targets. The development of Paxlovid, an oral Covid treatment, in four months was helped along by deploying various machine learning techniques, Pfizer has said. ‘A great convergence’ A “great convergence” is underway throughout the industry, according to Julia Angeles, the portfolio manager of Baillie Gifford’s Health Innovation Fund. “It’s not just one technology that comes to play a role. It’s actually a combination of technologies,” Angeles said. In an interview, she detailed a number of improvements that have occurred with the advanced algorithms used to power machine learning, the richness of the data sets that can be examined for information and the efficacy of the computing power that is needed to bring it all together . But the critical change is the scale at which it is being done, Angeles said. “Many more companies can do it,” she said. “We have much more relevant data to mine biology, and we have much more powerful computers to do it much more effectively, and much faster than we’ve done it in the past.” One key component has been a steep drop in the cost of sequencing genomic data over the past 10 years, which has resulted in a trove of patient information that can be combined with other types of electronic health records. Separately, the release last year of source code for AlphaFold2 by DeepMind, the UK-based AI venture owned by Alphabet, has helped visualize the structure of proteins, which should also help development in that area in the coming years. So far, the technological progress has resulted in a wave of small-molecule drugs created by AI-native drug discovery companies. Combing through public records, Boston Consulting Group has identified more than 150 small-molecule drugs, with at least 15 already in clinical trials, from the top companies in the space. BCG said the pipeline is growing almost 40% per year. “Do these work in the clinic? We’ll have to wait and see. Hopefully they do. Because if they do, if they work as well as human-discovered drugs, that would be very exciting,” said Chris Meier, a managing director and partner at BCG. “If the success rate comes back much better, then of course it’ll get very exciting because all of a sudden we have something which is better than humans. We don’t know yet,” he said. The expected updates from a number of drug candidates over the next 12 to 18 months was a key reason Morgan Stanley analysts said they expect the sector is about to reach a turning point. In a research note published in late June, Morgan Stanley said readouts from early clinical work will help the market assign a value to AI-native drug stocks. The report said investors in the past have debated whether the group should sport the valuation of a technology platform or a biotech company. Indeed, the business models of these companies can vary. Some are more similar to the software as a service model, where the companies provide machine learning capabilities to partners for a fee. But many are also developing their own solo projects and have collaborations with pharmaceutical companies, where they will receive milestone payments and royalties as compounds meet objectives and are commercialized. The value of failing fast By Deloitte’s latest estimates, it can cost $2 billion to develop a new drug. That figure accounts for the vast majority of compounds that are studied, but fail in early clinical trials. Success rates can be less than 5%, and development times can span a decade or more. Morgan Stanley analysts estimate that an approximate 2% improvement in the pace of preclinical and phase 1 development could lead the industry to generate some 50 novel therapies over the next 10 years. This could equate to some $50 billion in net present value for the biopharma industry, they said. One of the key ways AI-enabled drug research can save money is by identifying the molecules that have the most and least likelihood of success early on in the research cycle. By doing this, the cost of failure is greatly reduced. Robert Burns, a managing director at HC Wainwright, said Schrodinger has described a 10-month time frame to identify a development candidate, while Exscientia has put its average time at around 12 months. By comparison, traditional drug discovery can take anywhere from three to five years. “That’s important, especially as you know, a lot of these companies within major pharma and biotech, they’re all trying to pursue very similar targets,” Burns said. Speed not only can save money, but it can provide a competitive edge. Despite the promise these companies hold, the stocks have fallen sharply along with the rest of the biotech sector. Most are now trading below their IPO prices. Baillie Gifford’s Health Innovation Fund reflects this trend. It’s down more than 26% year to date, but has gained nearly 7% so far this month, according to FactSet. Within the AI-first space, Angeles owns Exscientia and Recursion Pharmaceuticals, although neither ranks among the fund’s top holdings. Exscientia shares are down 39% year to date, and are trading 45% below its debut price last September. The company has collaborations with the Bill & Melinda Gates Foundation, Bayer, Sanofi, Bristol-Myers Squibb and others. The immunotherapy oncology drug, EXS-21546, is Exscientia’s most advanced compound. It’s in Phase 1b/2 trials to test the drug in patients with solid tumors. Recursion Pharmaceuticals has lost about 45% of its value since its April 2021 IPO. It’s very focused on using imaging technology to discover drug targets, and much of its focus has been on rare diseases. It has partnerships with Bayer, Roche and Takeda, and is already in a Phase 2 clinical trial to treat cerebral cavernous malformations, a disorder of the blood vessels in the brain, which can lead to seizures and fatal bleeding in the brain. Burns has a buy rating on Relay Therapeutics, which is off about 35% so far this year, and is trading just below its $20 IPO price. The company has several treatments for breast cancer in the works, and data on its lead compound, RLY-4008, should be released by the end of this year. Its partners include Roche and Genentech. On Thursday, Relay said it had sufficient funding to support its operating plan into at least 2025. As of June 30, its cash and investments totaled about $838 million, compared to $958 million at the end of 2021. Schrodinger reported it has $513 million in cash, cash equivalents, restricted cash and marketable securities, as of June 30, down from $529 million at March 31. At the end of its first quarter, Exscientia had about $719.8 million in cash, while Recursion at $591.1 million as of March 31. Until these companies offer updates on these programs, the investment case hinges on the potential value of the companies’ platforms. Once investors can see the progress being made in clinical trials there will be more confidence. “I think there really needs to be some sort of validation here,” Burns said.