Chinese Academy of Sciences

Long-context RNA foundation model reading whole mRNA transcripts at single-nucleotide resolution, pretrained at a native 10,240 nt context.

Single-cell foundation model that predicts latent representations of graph-connected gene blocks instead of reconstructing individual gene counts.

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

Vision-guided model that builds virtual 3D organoid surrogates from brightfield microscopy to predict chemical perturbation responses without omics.

Diffusion-based generative RNA model for de novo sequence design, conditioned on function, RNA family, structure, or binding proteins.

Direction-aware foundation model trained on bulk RNA-seq differential-expression profiles to simulate coordinated gene dynamics in viral infection.

Autoregressive language model trained on 37 million intrinsically disordered region sequences, generating IDRs given flanking folded domains.

Autoregressive model for therapeutic mRNA design that jointly generates 5' UTR, CDS, and 3' UTR, pretrained on 30 million full-length natural mRNAs.

Gastrointestinal histopathology foundation model pretrained on 353 million multi-scale patches from 210,000 H&E whole-slide images of GI tissue.

Protein-language diffusion model generating all-atom conformational ensembles for intrinsically disordered proteins and disordered regions.

Generative RNA foundation model trained on 114 million full-length sequences for de novo design of tRNAs, aptamers, CRISPR guide RNAs, and mRNAs.

Structure-aware generative DNA language model pretrained on influenza genomes that forecasts future antigenic variants across regions and subtypes.

Conversational single-cell and spatial multi-omics brain foundation model, with zero-shot cell annotation and disease prediction across species.

Deep learning framework that predicts DNA methylation from genomic sequence across 39 human tissues, with an scRNA-seq variant for unseen cell types.

Enzyme backbone design model that adds substrate and catalytic-site control to a pretrained SE(3) flow-matching generator via a lightweight adapter.

Scientific multitask language model whose byte-level genome modeling beats Evo 7B on DNA perplexity and mutation-effect prediction at 1.5B params.

SE(3)-equivariant chemical language model for pocket-based 3D molecule generation, used to design an HPK1 inhibitor with in vivo anti-tumor efficacy.

Generative antibody and nanobody design model that co-designs CDR sequences and antigen-bound structures for de novo design and affinity maturation.

Antibody-antigen binding free energy predictor fusing ESM-2 sequence embeddings with persistent homology and interface geometry via cross-attention.

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

Molecular structure elucidation model that reads IR, Raman, UV-Vis, NMR, and mass spectra as text and generates SMILES end to end.

Influenza genomic language model adapting DNABERT-2 to ~900,000 viral genomes, identifying subtypes, segments, and pathogenicity from sequence.

Enzyme Commission number prediction that pools ESM Cambrian embeddings across unlabeled sequence homologs, scoring 0.788 F1 on full 4-digit EC.

Cross-species brain spatial transcriptomics foundation model pretrained on 133M cells from human, macaque, marmoset, and mouse whole brains.

Single-cell perturbation response prediction using dual conditional diffusion bridges that link unpaired control and perturbed populations.

Multimodal drug-response model coupling cell and molecule foundation models, pretrained on 1.8M perturbation RNA-seq profiles over 22,000 compounds.

EEG foundation model pairing a decoupled time-frequency tokenizer with a multi-scale state-space encoder for generalization under distribution shift.

Cyclic peptide design conditioned on target protein structure, generating all four cyclization types via all-atom, all-bond harmonic SDE modeling.

Pocket-based molecular docking with an SE(3)-equivariant diffusion transformer. Places 77.65% of top-1 poses within 2 Å RMSD on PoseBusters.

Single-cell multi-omics foundation model with a Mamba backbone, pretrained on 2.7 million paired scRNA-seq and scATAC-seq profiles.

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

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

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

Secondary metabolite structure prediction from microbial biosynthetic gene clusters, generating SMILES strings from Pfam functional-domain tokens.

Translates whole-brain imaging phenotypes between humans and mice through a shared latent space built from transcriptomics and connectivity.

Peptide binder design by mimicking the binding interface of a known receptor or antibody, generating all-atom peptides through latent diffusion.

Grounded multimodal language model for endoscopic surgery, supporting visual dialogue, region-based question answering, and bounding-box grounding.

Biomedical multimodal LLM that answers questions about medical images and returns pixel-level segmentation masks, using a mixture-of-experts design.

Ultrasound foundation model pretrained by federated learning across 16 institutions, transferring to disease diagnosis and lesion segmentation.

RNA language model built from bidirectional Mamba2 blocks with a flash-attention head, pretrained on 100 million sequences up to 2,048 nucleotides.

Knowledge-informed cross-species foundation model pre-trained on 101 million human and mouse single-cell transcriptomes to decipher gene regulation.

Spatial transcriptomics foundation model using cross-attention over niche ligand genes, pretrained on 4.1M deconvolved human Visium samples.