Isomorphic Labs builds AI engine to replace empirical drug discovery

Isomorphic Labs, backed by Alphabet and founded around 2021, is developing an in-house AI drug design engine to move drug discovery from empirical trial-and-error to computer simulation. President Max Jodberg stated the goal is to achieve experimental accuracy in simulation models to reduce reliance on real-world assays, which are slow and expensive. Jodberg noted that current drug development takes about 15 years and costs $1.5 billion, with most failures occurring in clinical trials due to unforeseen toxicity or off-target effects. He cited rat models as only about 30% predictive of human clinical toxicity and argued for using AI, organoids, and in vitro endpoints to better model human biology. The company does not license its models but uses them internally for its own pipeline and partnerships with pharma companies, with internal focus areas including oncology, immunology, and inflammation. To build this capability, Isomorphic Labs is actively hiring core AI researchers and deep learning experts who may lack life sciences backgrounds. The hiring strategy involves pairing these AI specialists with domain experts in biology and chemistry in a ratio of roughly five AI people to one domain expert. Jodberg noted that the intersection of frontier AI and drug discovery is new, meaning there are few existing experts in the specific field, necessitating the recruitment of top-tier talent from core AI fields. The company aims to create a culture where these disparate experts learn from each other to develop predictive models that can generalize across disease areas and modalities. Isomorphic Labs generates its own proprietary data to avoid biases in public repositories, aiming to achieve compounding improvements in drug discovery efficiency and safety.