KAIST

Biomolecular complex structure prediction without multiple sequence alignment, for protein-protein, protein-ligand and protein-nucleic acid binding.

Enzyme-substrate specificity prediction by end-to-end co-folding, with no predefined binding pocket. AUROC 0.766 on unseen enzymes and substrates.

Masked DNA language model with a learnable, adaptive tokenizer that produces context-dependent, variable-length segments instead of fixed k-mers.

Sparse autoencoder for blood-cell microscopy that decomposes hematology foundation model embeddings into expert-validated sub-cellular concepts.

Absolute protein folding stability prediction that estimates ΔG by jointly modeling the folded and unfolded ensembles as residue-pair distograms.

Single-nucleus RNA-seq foundation model built on bidirectional Mamba blocks, reading all 19,306 genes per cell without dimensionality reduction.

Functional MRI foundation model that learns brain dynamics as a stochastic optimal control problem, self-supervised on 41,072 UK Biobank subjects.

Fragment-based molecular generation by masked discrete diffusion over SAFE strings, covering de novo design, linker design, and lead optimization.

Structure-based drug design that generates molecules as synthetic pathways over 1.2M purchasable building blocks, so each design carries its route.