At a health-focused event in New York on Tuesday, Google announced it was developing a collection of “open” AI models called TXGEMMA.
The AI model, which Google said will be released later this month through the Health AI Developer Foundations program, allows the company to understand both the structure of various “therapeutic entities” including chemicals, molecules and proteins.
“Because developing therapeutics from concept to approved use is a long and expensive process, we are working with the wider research community to find new ways to make this development more efficient,” Google’s chief health officer Karen Desalvo wrote in a blog post provided to TechCrunch. “(r)explorers can help them ask TXGEMMA questions to predict the key properties of potential new treatments.
Google does not say whether the model license allows commercial use, customization, or tweaking. TechCrunch will contact the company for more information and will be updated if the company responds.
Countless companies, including Google Spinout’s isomorphic lab, have promised that AI can one day revolutionize biological discovery by dramatically accelerating the earliest R&D steps. There have been some successes, but AI doesn’t offer an immediate magical solution in the lab.
Several companies employing AI for drug discovery, including Exscientia and Benevolentai, have suffered from the failure of well-known clinical trials in recent years. Meanwhile, the accuracy of major AI systems for drug discovery, such as Google Deepmind’s Alphafold 3, tends to vary widely.
Still, big pharmaceutical companies and investors appear to be keen on the potential of technology. In January, Isomorphic, which has a partnership with Pharma Giants Eli Lilly and Novartis, said it expects testing for AI-designed drugs to begin later this year. One estimate shows that over 460 AI startups are working on drug discovery, and investors have put $60 billion into the space so far.