Columbia University

A causal multimodal Transformer that embeds the do-operator within attention to predict single-cell responses to gene perturbations, including unseen ones.

A transformer that infers whole-genome DNA methylation landscapes from gene expression, generalizing zero-shot to unmeasured CpG sites and unseen samples.

EEG foundation model for brain-computer interfaces, using amplitude-aware masked pretraining and a progressive mixture-of-experts transformer across diverse electrode setups.

A neuro-oncology foundation model for brain tumor MRI that uses distributionally robust self-supervised pretraining to predict molecular markers and survival across institutions.

A graph-signal-processing foundation model that embeds gene regulatory network structure directly into its attention mechanism for parameter-efficient single-cell transcriptomics.

Protein language models trained on biophysical dynamics from MD simulations and normal-mode analysis; ESMDance fine-tunes ESM2 for strong zero-shot mutation-effect prediction.

An R package that uses GPT-4 to annotate cell types in scRNA-seq data from marker genes, matching expert accuracy across hundreds of cell types and tissues.