 PASIST is a novel deep learning framework designed to identify informative gene targets for spatial transcriptomics studies. It uses reference SCINasec data to select genes that capture more accurate predictions of the genome-wide expression profile, enabling more efficient and reliable identification of gene targets. The framework can be adapted to specific biological goals and can also be used to generalize from SCINasec data to spatial transcriptomics data. This article was authored by Ian Covert, Rohan Gala, Tim Wang and others.