MIT

Protein foundation model predicting a ten-modality joint embedding from sequence alone. Reaches 69.99% enzyme class accuracy on dark proteins.

Histopathology foundation model predicting spatial gene expression from H&E slides at single-cell resolution via linear whole-slide attention.

EEG foundation model pretrained on raw multi-channel waveforms via continuous-time flow matching. Leads 7 of 9 clinical and BCI benchmarks.

Unified all-atom generative model for biomolecular structure prediction, binder filtering, and controllable protein and nanobody design.

Flow-matching framework that translates omics signatures across biological domains, such as mouse to human transcriptomics, without paired samples.

Multimodal foundation model integrating spatial transcriptomics, H&E histopathology, and pathway scores for single-cell niche discovery.

Cell-centric microscopy foundation model that distills morphology and microenvironment views into a unified embedding for virtual spatial omics.

Prime editing efficiency prediction from pegRNA sequence, with every biochemical step of the editing mechanism modeled as its own learned rate.

Inverse folding model that emits a pairwise Potts energy function from a backbone, so the same network designs sequences and predicts mutation ddG.

Energy-based model of protein conformational space, turning a diffusion model into a statistical potential for structure ranking and mutation scoring.

All-atom generative model for de novo protein and peptide binder design against diverse biomolecular targets, wet-lab validated across 26 targets.

Graph diffusion transformer for in-context molecular design, adapting to new tasks from a few molecule-property demonstrations without fine-tuning.

Protein deimmunization framework that aligns a protein language model with GRPO to design sequences carrying fewer MHC Class I T-cell epitopes.

Multimodal diffusion transformer for de novo protein design, jointly generating sequence and structure conditioned on 465 Gene Ontology functions.

Protein dynamics model predicting per-residue covariances and residue-residue coupling from a static structure for fast RMSF and ensemble generation.

Patient-level representation learning from scRNA-seq: a transformer set encoder with a diffusion decoder, fine-tuned for clinical prediction.

Structure-based virtual screening model that jointly predicts protein-ligand complex structures and binding fitness from sequence and SMILES.

Protein surface tokenizer that turns surface-exposed residues into codebook tokens, lifting SKEMPI binding affinity change prediction to r = 0.600.

MSA-based protein language model for unsupervised contact prediction, outperforming ESM2-15B with 111M parameters and leading on interface contacts.

De novo protein backbone generator trained on low-confidence AlphaFold structures as corrupted data, reaching 86% designability at 700 residues.

Single-cell perturbation-response model that predicts transcriptomic and imaging outcomes of unseen genetic perturbations via a VAE with attention.

Plasmid characterization and retrieval model aligning DNA sequences with property text across ten facets, from antimicrobial resistance to host range.

Protein backbone generation from voxelized 3D densities, pairing a 3D CNN autoencoder with latent flow matching for shape and motif conditioning.

Molecular property prediction foundation model pretrained on computed descriptors. Wins 75% of Polaris tasks and 97% of MoleculeACE assays.

MSA design model generating alignments from protein language model embeddings to improve folding accuracy on orphan and low-homology proteins.

Open model that jointly predicts biomolecular structure and small-molecule binding affinity, approaching FEP+ accuracy in seconds on a single GPU.

Multimodal clinical foundation model reasoning jointly over 2D and 3D medical images, ECG time-series, and text reports across nine clinical domains.

Distribution-level representation learning that embeds whole cell populations, perturbation responses, and sequence sets, not individual data points.

Self-supervised transformer pretrained on millions of tandem mass spectra, giving embeddings for spectral annotation and fingerprint prediction.

Surgical video foundation model pretrained by entropy-maximizing compression on 0.78M unlabeled frames from 2,122 minimally invasive procedures.

End-to-end protein structure alignment from 3D coordinates, matching TM-align quality at LDDT 0.56 and beating Foldseek on SCOPe40 search.

Peptide-MHC structure prediction by SE(3)-equivariant diffusion, sampling 10 Cα conformations in under 6 seconds at 0.45 Å best-of-10 RMSD.

Multi-agent LLM framework for autonomous discovery in protein science, pairing generator and critic agents over design and simulation tools.

Zero-shot RNA design pipeline that ranks variants by genomic language model likelihood combined with inverse-folding structural compatibility.

Protein binder design that inverts the frozen Boltz-1 all-atom predictor, targeting small molecules, nucleic acids, metals, and modified residues.

2D/3D X-ray to volume registration for surgical navigation, aligning live fluoroscopy to preoperative CT or MRI after a five-minute fine-tune.

Single-cell model inferring which developmental signaling pathways are active from scRNA-seq, trained on combinatorial stem-cell perturbation screens.

Protein-protein interaction language model that embeds whole sets of interacting chains through cross-chain attention, not one sequence at a time.

Clinical imaging encoder multitask-pretrained across X-ray, mammography, dermoscopy, fundus, ultrasound, CT, and histopathology for few-shot transfer.

Latent diffusion model generating 3D drug-like molecules and inorganic crystals from one shared all-atom autoencoder and Transformer denoiser.

Generative design of protease substrates, producing 10-mer peptides conditioned on a target cleavage profile across 18 matrix metalloproteinases.

De novo protein sequence design conditioned on target vibrational dynamics, pairing a diffusion designer with a predictor that ranks candidates.

Anti-phage defense gene classifier pairing protein language model embeddings with genomic features to find immune systems outside defense islands.

Structure-to-sequence protein design network inverting trRosetta, predicting coevolution features from a backbone to generate stable sequences.

Geometric foundation model matching enzymes to the reactions they catalyze, trained on 1.5 million structure-informed enzyme-reaction pairs.

Histopathology patch encoder turning 512x512 tiles into 768-dimensional features, trained with a CoCa objective on 1.26 million captioned images.

Motion foundation model for wearable accelerometry, trained with relative contrastive learning on 1B segments from 87,376 participants.

Target-conditioned peptide binder design model that samples hot-spot residues from an energy-based density, then extends fragments autoregressively.

Macrocyclic peptide binder design against protein targets, cyclizing a diffusion backbone generator's positional encoding so it closes rings.

Open-source structure prediction model for proteins, nucleic acids, and small molecules, trained on public data to AlphaFold3-level accuracy.

De novo enzyme design conditioned on the reaction to be catalysed: substrate and product SMILES in, catalytic pocket, enzyme, and docked complex out.

Cell Painting image generation conditioned on a control well image and a compound's structure, covering cell lines and chemicals never trained on.

Molecular property prediction from 3D structure for assays with dozens of labels, pretrained by extreme denoising plus DFT and LLM auxiliary labels.

DNA methylation foundation model over 49,156 array CpG sites. Imputes missing values, embeds samples, and predicts epigenetic age and disease risk.

Poly(A)-tail length change predicted from mRNA 3' UTR sequence in maturing oocytes, scoring how single-nucleotide variants disrupt tail lengthening.

Spot detection and quantification in 5D fluorescence microscopy. Pretrained 2D and 3D U-Nets segment foci, then Gaussian fitting measures each one.

Diffusion model that jointly generates a 3D molecule and its shape, electrostatic surface and pharmacophores for ligand-based drug design.

Walk-jump sampler that runs molecular dynamics in a smoothed, noised space of all-atom coordinates to generate peptide conformational ensembles.

Predicts binding free energy change (ΔΔG) at protein-protein interfaces by scoring bound and unbound states with an inverse folding model.

All-atom protein dynamics simulator that bridges consecutive MD snapshots directly using SO(3)-equivariant two-sided stochastic interpolants.

Multimodal LLM for inverse molecular design, interleaving text and graph generation with a diffusion transformer and A* retrosynthetic planning.

Perturbation target identification for single-cell transcriptomics, reading intervened genes off the difference between two inferred causal graphs.

De novo protein binder design pipeline that hallucinates binders through frozen AlphaFold2 weights, averaging a 46% experimental success rate.

Molecular dynamics surrogate that generates whole trajectories, switching between forward simulation, transition path sampling and upsampling.

Mixed-modality genomic language model encoding protein coding sequences as amino acids and intergenic DNA as nucleotides in native genomic context.

Interactive foundation model for biomedical image segmentation, prompted with scribbles, clicks, and bounding boxes to segment unseen structures.

Structure-based mutational effect prediction from local atomic environments, scoring how substitutions change protein stability and binding affinity.

Keypoint-based foundation model for brain MRI registration, pretrained on over 100,000 3D volumes for rigid, affine, and deformable alignment.

Protein language model that predicts which of twelve subcellular compartments and biomolecular condensates a human protein partitions into.

Genomic language model trained on metagenomic scaffolds that learns protein co-regulation and function by modeling gene context and operon structure.

Blind protein-ligand docking that transfers to binding domains absent from training, scoring 22.6% top-1 on DockGen and 50% on PoseBusters.

Protein conformational ensemble generator that fine-tunes AlphaFold 2 with flow matching, sampling protein dynamics beyond a single static structure.

Spot detection for single-molecule RNA FISH and fluorescence microscopy, trained on a differentiable F1 approximation, needing no threshold tuning.

De novo protein backbone generation by SE(3) flow matching, with motif-scaffolding built in. Samples a designable backbone in seconds on one GPU.

Medical image segmentation model that solves unseen segmentation tasks in context from a few labeled examples, with no retraining or fine-tuning.

Rigid protein-protein docking by diffusion over the rigid-body pose, with a confidence model ranking sampled complexes. Median C-RMSD 4.85 on DIPS.

Cyclic peptide structure prediction and de novo macrocycle design, by wrapping a frozen structure predictor's positional encoding into a ring.

De novo protein backbone generation by SE(3) diffusion over rigid-body frames. Samples designable monomers up to 500 residues without pretraining.

Blind protein-ligand docking by diffusion over ligand translations, rotations and torsion angles, reaching 38% top-1 accuracy on PDBBind.

Rigid protein-protein docking model that predicts a complex from two unbound structures in one pass, with no candidate sampling or refinement.

Chemical language model pretrained on up to 77 million PubChem SMILES strings for molecular property prediction on the MoleculeNet benchmarks.