UCSF

Ordered-water prediction for protein structures by flow matching, adding the crystal symmetry mates that coordinate waters at lattice contacts.

Genomics foundation model that represents individual DNA fragments in a learned semantic space for cell-free DNA cancer detection and cell typing.

Multimodal reasoning LLM for protein function prediction, fusing protein language model embeddings to emit interpretable GO-term reasoning traces.

ADMET property prediction for CYP inhibition, PXR activation, hERG, microsomal clearance, and permeability, all trained on public assay data.

Self-supervised models that embed gut metagenomic abundance profiles for robust phenotype prediction in data-limited, cross-cohort settings.

All-atom generative model for de novo protein design using SE(3) flow matching over oriented residue rigid bodies.

Cytometry foundation model for immunology, pretrained by masked-marker prediction on over 100 million immune cells from 53 public ImmPort studies.

DNA-LLM reasoning model fusing genome foundation model embeddings with an LLM to produce step-by-step pathway and variant effect explanations.

Ophthalmic foundation model behind the GlobeReady platform, reading fundus and OCT scans by frozen-encoder retrieval instead of per-site fine-tuning.

Cell-free RNA language model for liquid biopsy, fusing RNA sequence embeddings with cfRNA abundance across 306 billion pretraining tokens.

Spatial transcriptomics foundation model continually pretrained on 30 million profiles, with a protocol-aware mixture-of-experts decoder.

Phenotypic screening model for neuronal activity, encoding network-level calcium dynamics of iPSC-derived neurons into single-cell embeddings.

Decoder-only genomic language model at single-nucleotide resolution, fine-tuned to predict which DNA contacts the nuclear lamina or nuclear speckles.

Codon-resolution language model suite pairing a bidirectional encoder with an autoregressive decoder over protein-coding sequences.

Cryo-EM heterogeneous reconstruction in real space, decoding density maps from a conformational latent space disentangled from pose and CTF.

Virtual staining models that translate label-free light microscopy into fluorescent-equivalent predictions of nuclei and plasma membranes.

Structure-conditioned protein language model aligned to experimental stability data, scoring variant stability and generating stabilized sequences.

CLIP-style vision-language model for echocardiography, pretrained on echo videos and cardiologist reports for zero-shot cardiac interpretation.