GENOME LANGUAGE MODELS CAN DESIGN CANCER VACCINES

Summary from the TLDR newsletter

Modern cancer-vaccine design involves identifying mutations, predicting presentation, predicting immunogenicity, ranking targets, and designing RNA. However, biology is complicated, and whether a target makes a good vaccine candidate depends on all kinds of information. This makes cancer vaccines an unusually direct test of whether AI models can take what it has learned from biology and apply that knowledge to design a medicine for one particular human being.