All Competitors

Sequence-only peptide toxicity prediction, pairing a fine-tuned ESM-2 backbone with a bidirectional LSTM and focal loss for rare toxic peptides.

Protein function prediction that assigns Gene Ontology terms from predicted 3D structure, ESM-2 embeddings, and cross-species network propagation.

Genomic language model for scRNA-seq cell-type annotation, reweighting rare classes so diseased cell types are not swamped by common ones.

265M-parameter genomic foundation model whose BioToken scheme folds variants and region annotations into the tokens for variant effect prediction.

Histology nuclei segmentation that adapts SAM to train on several datasets at once, aligning auxiliary domains without diluting the primary one.

Antimicrobial peptide platform whose GPT-style generator is conditioned on E. coli or S. aureus activity, then filtered for potency and hemolysis.

Multimodal medical imaging foundation model built for chromosome karyotype analysis, with 92.75% sensitivity for structural abnormality detection.

Structure-based drug design model pairing an autoregressive transformer for ligand graphs with a diffusion head for 3D binding-pose coordinates.

Single-cell RNA integration model using adversarial batch training to embed and label cells from a new study without supplying a batch ID.

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

De novo binder design across small molecules, peptides, and antibodies from one geometric latent diffusion model over graphs of molecular blocks.

Antibody sequence generation model that samples paired human VH/VL chains, covering inpainting, inverse folding, and CDR grafting in one network.

De novo peptide sequencing from mirror-protease mass spectra, reading paired complementary spectra to recover near-complete fragment ion coverage.

Open therapeutics foundation models from Google, built on Gemma-2, for drug-discovery property prediction and conversational reasoning.

Ames mutagenicity prediction conditioned on bacterial tester strain and S9 metabolic activation, holding sensitivity on chemically novel compounds.

Whole-heart segmentation foundation model for CT and MRI, pretrained self-supervised on unlabeled cardiac scans with an xLSTM-UNet backbone.

Protein sequence representation model pairing a sequence autoencoder with a denoising diffusion model over its latent space for frozen embeddings.

Transferable coarse-grained force field for molecular dynamics of proteins, RNA, and lipids, built on the MACE equivariant graph architecture.

Histopathology foundation model pretrained with DINOv2 on tiles chosen by unsupervised hierarchical clustering over 350 million whole-slide tiles.

Splice-site and variant-impact prediction from DNA sequence, with pretrained models for human, mouse, zebrafish, honey bee, and Arabidopsis.

Protein language model pretrained on over nine billion sequences, giving residue embeddings and zero-shot single-site mutant scores from its logits.

Generative diffusion foundation model for human gait dynamics that estimates ground reaction forces from partial motion-capture kinematics.

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