Vanderbilt University

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

Electron microscopy foundation model segmenting mitochondria, ER, nuclei, and lipid droplets across tissues and species from one pretrained encoder.

Few-step generative model for de novo protein backbone design, extending MeanFlow to the SE(3) Lie-group geometry of residue frames.

Protein conformational ensemble tokenizer that learns a discrete alphabet of states from molecular dynamics, reusable as a frozen feature layer.

Multimodal transformer predicting alternative splicing outcomes across C. elegans neuron subtypes, reaching Spearman ρ = 0.88 on held-out exons.

Diffusion model that locates zinc binding sites in protein structures at 94% precision, without needing the number of ions specified.

Protein language model that generates paired heavy and light chain human antibodies from an antigen prompt, with binders validated in vitro.

Renal pathology segmentation model resolving 14 glomerular tissue, cell, and lesion classes in human and mouse whole-slide images.

Hydration-site prediction placing water molecules on protein surfaces by score-based diffusion, within 0.3 Å of crystallographic positions.

Lasso peptide language model that adapts ESM-2 to threaded RiPP core sequences, supplying embeddings for cyclase substrate and activity prediction.

Structure-based RNA virtual screening that scores small molecules against a binding site's base-pairing graph, around 10,000x faster than docking.

Pathology instance segmentation for glomeruli, nuclei and eosinophils, deforming a bounding circle into a contour rather than a box into an octagon.

Renal pathology segmentation covering six kidney tissue types across 5x to 40x magnifications from one network, and transferring from human to mouse.

Glomerular detection, segmentation, and glomerulosclerosis subtyping from renal whole-slide images, pretrained on web-mined glomerular figures.

Glomerulus and nuclei detection in whole-slide pathology images, predicting a bounding circle rather than a box for rotation-consistent localization.