Sun Yat-sen University

Hyperbolic protein language model for alignment-free phylogenetic inference, turning ESM2-650M embeddings into distance matrices for tree placement.

Molecular reasoning model built on DeepSeek-7B, using chain-of-thought and reinforcement learning for property prediction, generation, and reactions.

Generative virtual-cell model predicting whole-transcriptome responses to unseen compounds and genetic perturbations, from cell lines to organoids.

Designs synthesizable PROTAC degraders from reaction templates and purchasable building blocks, with reinforcement learning tuning the generator.

Histopathology foundation model extracting general-purpose features from H&E patches by distilling the UNI, Phikon, and CONCH pathology encoders.

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

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

Blind protein-ligand docking model adding Ollivier-Ricci curvature descriptors and degree-aware message passing, predicting poses in 0.09 seconds.

Multi-sequence MRI foundation model pretrained on 336,476 volumetric scans, ranking first on 41 of 44 downstream clinical benchmarks.

Structure-based virtual screening model that jointly predicts protein-ligand complex structures and binding fitness from sequence and SMILES.

EEG foundation model with cross-scale spatiotemporal tokenization and sparse structured attention, evaluated on 11 decoding tasks across 16 datasets.

Genome-anchored histopathology embeddings that predict molecular biomarkers, subtypes, and survival from whole-slide images alone at inference.

Multimodal viral foundation model over nucleotide and protein sequence, built for virus discovery, function annotation, and antibody design.

Pocket-conditioned 3D diffusion model for scaffold decoration, guided by evolutionary residue conservation and a protein-ligand interaction prior.

Cell Painting image encoder that turns whole-slide multi-channel microscopy into morphological profiles in one pass, with no cell segmentation step.

Cervical cytology screening system pretrained on 127,471 whole-slide images from 48 centers, with test-time adaptation for new clinical sites.

Geometric foundation model matching enzymes to the reactions they catalyze, trained on 1.5 million structure-informed enzyme-reaction pairs.

Text-guided medical image synthesis across OCT, fundus, X-ray, CT and MRI. Synthetic data lifts downstream clinical tasks by 12-17%.

Region-aware bilingual medical multimodal LLM that handles image- and region-level vision-language tasks across eight imaging modalities.

Single-cell foundation model with 800M parameters trained on ~100 million human cells, for annotation, perturbation prediction, and gene analysis.

Ophthalmic imaging foundation model pretrained on 2.78M images across 11 modalities for diagnosis, prognosis, and visual question answering.

Unified DNA, RNA, and protein foundation model with 1.8B parameters, pretrained across 169,861 species to learn the central dogma from sequence.

Medical vision-language pretraining unifying fusion-encoder and dual-encoder designs, handling image-only, text-only, and paired inputs in one model.

Self-supervised medical vision-and-language pretraining via multi-modal masked autoencoders that reconstruct masked image patches and text tokens.

Medical vision-language pretraining framework that injects structured medical knowledge into radiology image-text learning for VQA and retrieval.