All Competitors

Protein backbone generator running a diffusion transformer over SaProt structural tokens, with an IPA token cache to speed up de novo design.

Protein language model that predicts per-residue local energetic frustration directly from sequence, covering whole proteomes and disordered regions.

Cryptic protein binding site prediction from sequence, backed by a database of 5,151 cryptic sites mined from 6 million apo-holo PDB alignments.

Protein-protein docking scorer that ranks interface poses from image-encoded patches, swapping PIsToN's Vision Transformer for Vision Mamba.

Multimodal tokenizer for antibody CDR loops, encoding backbone dihedrals and sequence as discrete tokens that plug into antibody language models.

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.

Blind flexible protein-ligand docking model trained by two-player self-play, predicting bound ligand and pocket poses in 0.32 seconds per complex.

Protein language model adding long-range contact supervision to ESM2 via LoRA, improving all eight protein-level tasks with no structural input.

Text-to-text biological language model spanning molecules, proteins, and text, adding IUPAC names and multi-task instruction tuning to BioT5.

Full-atom flow matching model that generates a ligand and the induced-fit holo pocket together, starting from an apo binding site.

Scaling-law study of protein language models identifying compute-optimal training for causal and masked objectives on 939 million protein sequences.

Protein language models pretrained on Rosetta biophysics simulations rather than evolutionary data, then finetuned on small experimental assays.

Missense pathogenicity prediction that folds wild-type and mutant sequences with ESMFold and encodes each structure as a graph autoencoder embedding.

Multiscale graph neural network for fixed-backbone protein binder sequence design with a contrastive decoding algorithm to improve target selectivity.

Protein-ligand binding site prediction that ranks pocket residues and pocket center coordinates, staying accurate on AlphaFold-predicted structures.

Virtual drug screening from per-atom protein and ligand embeddings retrieved by nearest neighbors. 30.4 EF1% on DUD-E at ~14 s per million molecules.

3D convolutional network that predicts subcellular fluorescence labels from transmitted-light microscopy, enabling label-free imaging of living cells.

Cross-modal protein encoder that aligns ESM-2 sequence embeddings with ProteinMPNN structure embeddings in a shared space for cross-modal retrieval.

Contrastive geometric model unifying structure- and ligand-based drug design for zero-shot virtual screening, target fishing, and pocket selection.

Molecular dynamics emulator generating time-coarsened trajectories for small molecules, peptides, and proteins from one shared atomic representation.

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

Full-atom protein model accuracy estimation, regressing per-atom lDDT with an SE(3)-transformer over a heavy-atom graph of the modeled structure.

Cross-domain molecular foundation model encoding small molecules, protein pockets, and their complexes in 2D and 3D on one Transformer backbone.