OCTO-vc is a multi-scale, multi-modal transformer trained via masked token modeling to predict spatial patterns of gene expression in response to changes in biological context, such as spatial ...