All Competitors

Whole-slide histopathology foundation model that fuses 10x, 20x, and 40x views by attending only between adjacent magnifications.

Pan-allele peptide-MHC binding prediction unifying MHC class I and II, trained on a diversity-balanced set of 214 class I and 98 class II alleles.

Protein-ligand binding affinity scoring model for virtual screening that generalizes to novel pockets and ligands under strict train-test splits.

EEG foundation model that corrects low-frequency bias by reconstructing band-standardized time-frequency targets. State of the art on 24 of 41 tasks.

Protein language model for microbial smORF-encoded small proteins, pairing multi-scale convolutions with transformer layers in a compact encoder.

Hybrid diffusion-and-physics docking model that predicts protein-ligand binding poses and generalizes out-of-distribution for virtual screening.

EEG foundation model coupling spatial and temporal transformer branches through a shared soft mixture-of-experts, adapted by tuning 5.1% of weights.

Breast cancer histopathology foundation model distilled from three general-purpose PFMs, over 30x smaller with comparable balanced accuracy and AUC.

Physics-guided distillation that transfers 3D molecular dynamics knowledge into SMILES language models, improving MoleculeNet property prediction.

Enzyme-substrate specificity prediction by end-to-end co-folding, with no predefined binding pocket. AUROC 0.766 on unseen enzymes and substrates.

Single-cell foundation model adding a gated cell-level contrastive objective to masked expression pretraining for transferable frozen cell embeddings.

Dual-target structure-based drug design that fuses two pocket-conditioned Bayesian flow distributions to generate 3D ligands binding both proteins.

Antibody-specific epitope prediction that replaces sequence-offset rotary attention with backbone local-frame 3D geometry. 0.410 MCC on AsEP.

EEG foundation model pretrained to predict structured latent states rather than masked waveforms, reaching 52.94% frozen macro balanced accuracy.

Genome-scale synthetic lethality prediction for any human gene pair from Gene Ontology annotations, no protein interaction network required.

Signal peptide design framework that generates, filters, and ranks cargo-specific secretion signals using evolution-constrained discrete diffusion.

Genomic foundation model on a Mamba, attention, and mixture-of-experts backbone, with 1M-token context for variant scoring and regulatory DNA design.

EEG foundation model for continuous monitoring, using windowed alternating attention to hold KV-cache memory constant on recordings up to 14 hours.

Cellular foundation models predicting how human cells respond to genetic and pharmacological perturbation, trained on petascale multi-omic data.

Structure-free virtual screening model co-embedding protein residues and small molecules from sequence and 2D chemistry, scoring a compound in 10 ms.

Lysine crotonylation site prediction for human non-histone proteins, fusing a frozen ProteinBERT feature branch with a sequence transformer.

A pathology tile encoder fine-tuned from Midnight-12k so its features track tissue morphology rather than the scanner and lab behind the slide.

A robustified fine-tune of Phikon-v2 for H&E tiles, retrained to strip scanner and laboratory signatures out of its feature space.