M3-Gen Generates Gene Expression Profiles from Clinical and Imaging Data
July 24, 2026
M3-Gen uses a GAN conditioned on histopathology images and clinical metadata to generate biologically coherent gene expression profiles. The framework employs contrastive learning to build a unified latent representation, demonstrating realistic data generation on the TCGA dataset.
HOW THIS AFFECTS YOU
●
researcherYou can use this to augment limited datasets with synthetically generated molecular profiles.
●
healthThis could potentially reduce costs and privacy hurdles in multimodal biomedical research.