Artificial intelligence is touching all parts of our lives, from digital assistants to solving math problems. But we have yet to see a significant impact on the success rate, cost, or development time for new medicines. The limitations are straightforward – even if AI can develop a new drug candidate, it still takes years of testing in humans to validate the approach. By the time that validation is complete, the underlying AI model is hopelessly out of date.
Earlier this year, we asked 44 drug developer executives and 50 U.S. doctors whether AI will significantly impact the over $1 billion and 10+ years it takes to develop a new drug, or raise the <25% probability of approval of a new drug entering clinical trials. The short answer is that we don’t know yet, but the respondents are excited about AI’s ability to process large data sets, identify new drug targets, and leverage better biomarkers to identify the right patients for clinical trials. Concerns include typical AI issues such as privacy and errors, but also the potential of AI to reduce innovation by recommending the least risky path.
In answer to which companies are leaders in AI assisted drug development, the drug development executives cited a range of pharma, biotech, and academic centers, but also included big tech companies like Google / Alphabet, and spinoffs / partners of big tech like Recursion (partnered with Nvidia), Isomorphic (subsidiary of Alphabet), and DeepMind (research division of Alphabet).
Soon after our survey was fielded, Anthropic acquired AI biotech startup Coefficient Bio for $400 million. Also this year, Amazon launched Amazon Bio Discovery. Microsoft released BioEmu.
Experts describe big tech as “just poking around” at how they want to be involved in drug development, but note the incentives are strong: As the public is warned of the existential threat that AI poses to humanity, and concern for AI-related job losses increases daily, AI companies are drawn to areas everyone where AI assistance would be welcome. We can all get behind new medicines to help us live longer and healthier lives.
In order to train AI models to be good drug developers, we need data. Big tech companies will look for large data assets to pair with their compute power. Fortunately, there are a range of companies developing such huge, well-curated data sets; leaders are Grail, Tempus, Guardant Health, and Natera. The goal is to leverage these data sets not only to develop new compounds, but to speed up clinical trials by testing these drugs only on patients likely to respond. While we still need to wait for slow humans to respond in clinical trials, the combination of big tech compute power with big biological data sets will revolutionize the speed, cost, and efficiency of how we make new and important medicines.
Over the next year, we’ll see more partnerships between emerging “tech bio” startups and big tech. We also expect partnerships between healthcare data companies and big tech companies. But perhaps most interesting will be partnerships between the large pharma and large tech companies as they work together to revolutionize drug discovery.
.jpg "When AI Joins the Drug Discovery Lab")
