Design for Inference: From Random Experiments to Lab in the Loop

All stages of drug discovery and development are incredibly challenging, such that development of new medicines not only is long, complex, and costly, but also suffers from an exceptionally high failure rate, leaving much unmet medical need. These challenges are largely due to the complex, non-linear nature of the underlying scientific problems, from deciphering how cells malfunction in disease, predicting correct targets for therapeutic intervention, generating and designing molecules or other therapeutics to target them, and predicting which patients should be treated and in which dose and regimen. In each of these problems, we are posed with enormous spaces of possibilities, far exceeding those that can be measured in a lab or a patient population, such as the number of possible combinations of gene variants, or drug-like small molecules or therapeutic antibodies. The dramatic advances across different areas of machine learning, from representation learning to generative AI, now open an extraordinary opportunity to tackle each of these challenges to transform drug discovery. A true impact will require a shift across drug R&D, to become part of a “Lab in a Loop,” where experimental or clinical data is collected in order to train a model, the model is used to predict the next set of experiments, and the process is iterated, at scale, both to yield key predictions in any specific project and improve the model for all projects. In this talk, I will describe how we built such a Lab in a Loop of experiments and machine learning in Genentech across our target discovery, drug discovery, and drug development efforts to serve patients with autoimmune disease, neurodegeneration, infectious disease, and cancer.

Speaker: Aviv Regev, Genentech

Oak Conference Room