All Competitors

Genomic language model for tomato pangenomes that conditions DNA tokens on graph node attributes and local adjacency as well as masked modelling.

Twelve-lead ECG foundation model aligned to clinician-overread report text, adapting to new cardiac tasks from as few as ten labels.

Protein inverse folding that fuses sequence and structure priors from two frozen language models per residue. Reaches 63.11% recovery on CATH 4.2.

Protein language model scoring whether a kinase missense variant alters catalytic activity, fine-tuned from ProtBert on 2,553 curated alterations.

Pretrained atomic representation of non-canonical amino acids, scoring binding affinity, selectivity and cell permeability of therapeutic peptides.

Cardiac MR video foundation model whose frozen encoder estimates ejection fraction, myocardial strain and cardiomyopathy from cine, LGE and mapping.

Physiological time-series classification model bringing in-context learning to ECG, EEG, and ICU signals with temporal and channel-aware encodings.

Low-dose CT reconstruction by flow matching along the residual path to the paired standard-dose image, reaching diffusion-level quality in one step.

Pocket-conditioned 3D ligand generator that steers flow matching with LLM-written chemical priors, for de novo design and scaffold hopping.

Fragment-protein interaction model trained on cellular chemoproteomics, pairing ESM-2 residue embeddings with bilinear attention over ligand atoms.

Histopathology foundation model for whole-slide cancer diagnosis, covering 19 common cancer types and 205 clinical diagnostic tasks.

Lightweight histopathology model of 2 million parameters that runs on an ordinary hospital PC alongside a cloud model for cancer diagnosis.

Unsupervised gene finder that annotates any eukaryotic genome from a raw FASTA file, using a differentiable hidden Markov layer inside the network.

Flow-matching generative prior over high-quality PET images, used as a manifold constraint in an ADMM solver for low-dose PET reconstruction.

Structure scoring model for antibody-antigen and protein-ligand complexes, built on one frozen geometry-aware backbone reused across scoring tasks.

Volumetric cell segmentation for 3D fluorescence microscopy, pretrained on 5.1 TB of unlabeled volumes and generalizing across species and imaging.

Biomolecular complex structure prediction without multiple sequence alignment, for protein-protein, protein-ligand and protein-nucleic acid binding.

Tissue reconstruction model placing dissociated single cells into spatial coordinates by predicting pairwise distances in a learned embedding space.

Diffusion model that generates an ordered protein folding trajectory from sequence, predicting intermediates and transition states, not one fold.

Conformational ensemble sampler that anneals AlphaFold3's pairwise conditioning signal along the dynamical domains of a Gaussian Network Model.

Modality-agnostic transformer pretrained by masked abundance reconstruction on 48,837 proteomics profiles reprocessed from 1,397 PRIDE projects.

Multi-task cellular foundation model predicting drug sensitivity, perturbation expression and drug-protein binding from one pretrained checkpoint.

Protein structure prediction from multiple sequence alignments, trained across MSA depths so one model spans deep alignments and orphan proteins.