University of Science and Technology of China

Olfactory receptor-odorant interaction prediction from sequence and SMILES, pairing protein and chemical language models through cross-attention.

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

RNA modification profiling from nanopore direct RNA-seq signal; self-supervised pretraining resolves 11 modification types and extends to new ones.

Lung pathology foundation model adapted from Virchow2 on whole-slide images, validated across 32 tasks spanning the lung diagnostic workflow.

Pathology foundation model that infers spatial transcriptomics and proteomics directly from routine H&E whole-slide images, with no spatial assay.

Pan-cancer pretrained diffusion model imputing genome-wide expression from sparse spatial transcriptomics panels, zero-shot and reference-free.

Family of autoregressive genomic foundation models that reconcile k-mer tokenization with single-nucleotide resolution at contexts up to 98k bp.

Enzyme design model that jointly generates enzyme sequences and substrate-binding pockets, conditioned on functional priors and substrate structure.

Structure-based drug design model that generates ligands for a protein pocket, pairing a diffusion structure encoder with preference optimization.

Discrete graph diffusion model for multi-property molecular generation, composing per-property score guidance over arbitrary condition subsets.

RNA foundation model pretrained jointly on sequences and secondary structures for structure prediction, homology and splice site classification.

Text-prompted pathology image segmentation across 160 tissue, cell, and nuclei categories, replacing point and box inputs with natural language.

Peptide-HLA immunogenicity prediction with a BiLSTM ensemble, inside a pipeline that finds microbial epitopes mimicking tumor neoantigens.

Relative protein-ligand binding affinity prediction from docked complexes, matching Schrodinger FEP+ ranking accuracy zero-shot on the FEP benchmark.

Single-cell foundation model that fuses scRNA-seq profiles with text, pairing a cell encoder with an LLM for cell annotation and clustering.

Histopathology foundation model pretrained on 300K whole-slide images across 20 tissue types and validated on 112 clinical-grade downstream tasks.

3D molecule generation that writes a valid 1D SELFIES string with a pretrained language model, then predicts its conformer with a diffusion module.

Long-context generative genomic foundation model with a 98k-nucleotide window, trained on 386 billion bases of eukaryotic DNA for sequence design.

Antibody CDR sequence-structure co-design by flow matching, starting from an informative structural prior rather than from Gaussian noise.

Protein-text foundation model aligning sequences with function descriptions through segment-wise objectives for static and dynamic functional sites.

Multi-omics instruction-tuned LLM that reads DNA, RNA, protein, and multi-molecule sequences and answers natural-language questions about them.

Self-supervised histopathology encoder that models tissue entities as a graph and pretrains by latent graph diffusion over masked subgraphs.

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

Protein inverse folding as a generative Markov bridge, refining a structure-derived sequence prior with a frozen protein language model.

Dual-language transformer pretrained on paired protein and mRNA coding sequences, scoring protein and mRNA properties and generating optimized CDS.

3D CT vision-language model that drafts radiology reports, answers questions about volumes, and screens for disease from a masked-autoencoder encoder.

Protein-ligand docking framework that picks the binding pocket by contrastive alignment, then refines the pose with bi-level iterative refinement.

Structure-based drug design diffusion model that re-extracts the essential binding subcomplex from a pocket at every step of 3D ligand generation.

Vision-language pretraining for 3D CT volumes, aligning scans with their radiology reports for zero-shot classification, retrieval, and segmentation.

Vision-language pre-training framework for universal brain MRI diagnosis, learning from imaging-report pairs to cover more than ten brain diseases.

Protein language model ranking stabilizing mutations with no assay data, trained jointly to predict the growth temperature of each sequence's host.

3D CT localization foundation model that pairs MedLAM with SAM to segment any anatomical structure at a fixed, dataset-independent annotation cost.

Equivariant heterogeneous graph network that rescores docked protein-ligand poses for virtual screening and ranks structural analogs by activity.

Protein language model reading each residue alongside an unsupervised local-fragment token, so one encoder serves residue- and chain-level tasks.