Hong Kong University of Science and Technology

Infrared spectroscopy foundation model pretrained on 60 million simulated spectra, then adapted to real FTIR measurements of molecules and mixtures.

Breast-specialized multimodal pathology foundation model for core needle biopsy diagnosis, with conformal risk control gating report release.

Breast cancer histopathology foundation model distilled from three general-purpose PFMs, over 30x smaller with comparable balanced accuracy and AUC.

Molecular glue degrader design model that jointly generates the glue molecule and the E3 ligase-target ternary complex from unbound monomers.

Protein function captioning model fusing sequence, Foldseek structure tokens, and text through a BLIP-2 Q-Former for open-ended free-text annotation.

NMR foundation model trained on 158 million simulated 1H and 13C spectra, transferring simulation-learned representations to real experimental data.

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

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

Transcription factor binding site prediction model that refines a DNABERT-2 backbone with contrastive learning across diverse TFBS types.

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

B-cell epitope predictor pairing CNN and Transformer branches over protein language model embeddings to score linear and conformational epitopes.

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

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

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

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

Protein function annotation model predicting Gene Ontology terms with direct preference optimization layered on a frozen ESM-C sequence encoder.

Tri-modal pathology foundation model aligning whole-slide images, transcriptomes, and diagnostic reports, and running on any subset of the three.

Sparse-view CBCT reconstruction foundation model pretrained on 8,407 CT volumes, recovering full 3D anatomy from as few as six X-ray projections.

Universal foundation model that jointly generates diagnostic text and segments the corresponding targets across ten biomedical imaging modalities.

Multimodal 3D genome foundation model pairing Hi-C contact maps with epigenomic tracks, pretrained on over one million paired samples.

Generative foundation model that co-generates sequence and 3D coordinates for proteins, small molecules, and crystals under functional objectives.

Structure-based 3D molecule generation that denoises a ligand at atom level and motif level at once, conditioned on the target protein's pocket.

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

Medical vision-language model that unifies image comprehension and generation in one autoregressive transformer via heterogeneous LoRA adapters.

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

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

Protein cleavage site prediction that generalizes to proteolytic enzymes unseen in training by encoding active-site chemistry alongside sequence.

Biomedical vision-language model aligning image regions to UMLS clinical concepts, for zero-shot diagnosis across 10 imaging modalities.

DNA foundation model encoding the genome as probabilistic allele frequencies from the 1000 Genomes Project. 86M parameters, single-nucleotide tokens.

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

Chromatin loop caller for Hi-C, Micro-C, DNA SPRITE, and single-cell contact maps, pairing axial attention with a U-Net to work at very low coverage.

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

Whole-slide pathology assistant that states the morphological findings behind each diagnosis, trained on 180k VQA pairs from 9,850 gigapixel slides.

ECG foundation model that learns discrete rhythm tokens from noisy real-world recordings for arrhythmia classification and anomaly detection.

De novo enzyme design conditioned on the reaction to be catalysed: substrate and product SMILES in, catalytic pocket, enzyme, and docked complex out.

Chemical language model for molecular property prediction, pretrained on PubChem SMILES by masking whole functional groups rather than random tokens.

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

Multimodal protein model pairing a sequence encoder with a Gene Ontology branch, trained in recursive cycles that clean their own noisy labels.

Enzyme catalytic pocket design conditioned on a reaction: substrate and product in, pocket backbone, sequence, and EC class out.

Generalist medical vision-language foundation model with 40B parameters, spanning radiology, pathology, dermatology, retinography, and endoscopy.

Self-supervised pretraining framework for 3D medical image encoders that learns anatomy by predicting where a sub-volume sits within a CT scan.

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