Princeton University

A pan-cancer pretrained latent diffusion model that enhances spatial transcriptomics, imputing genome-wide expression from sparse panels with zero-shot generalization.

A transformer language model pretrained on 5' UTR sequences across five species to predict mRNA translation efficiency, ribosome loading, and expression levels.

A foundation model for fMRI brain activity recordings, pretrained with masked autoencoding on ~6,700 hours of data for clinical prediction and zero-shot network discovery.