University of Washington

Conditions a pretrained DNA sequence embedding on CpG methylation to predict gene regulation across cell types and alleles, generalizing zero-shot to imprinting, X-inactivation, and accessibility.

Diffusion-based backbone generation and sequence design method for programmable asymmetric transmembrane beta-barrel nanopores.

Discrete diffusion model for conditional antibody sequence generation that restricts learning to somatic variation via a germline-absorbing noising process.

All-atom diffusion model for de novo protein design conditioned on ligands, nucleic acids, and arbitrary non-protein atoms, enabling enzyme and DNA binder design.

Atom-level generative diffusion model for de novo enzyme design. Scaffolds arbitrary functional group geometries, solving all 41 benchmark active sites vs. 16/41 for prior methods.

A reinforcement-learning-enhanced general medical vision-language model that adds step-by-step reasoning for medical image diagnosis and visual question answering.

Protein sequence design method that explicitly models small molecules, nucleotides, and metals at atomic resolution, enabling ligand-aware design with 100+ validated designs.

Deep network that predicts structures of full biological assemblies containing proteins, nucleic acids, small molecules, metals, and covalent modifications simultaneously.

Diffusion model for de novo protein design that generates novel backbone structures conditioned on binding targets, symmetry constraints, and functional motifs.

AlphaFold fine-tuned on peptide-MHC and protein-peptide binding data for specificity prediction across MHC class I/II, PDZ, and SH3 domains.

Message passing neural network for fixed-backbone protein sequence design. Achieves 52.4% native sequence recovery, far surpassing Rosetta's 32.9%.