Technion – Israel Institute of Technology

Generative transformer for phylogenetic inference that transduces sets of unaligned molecular sequences directly into Newick-format trees.

Multimodal protein representation model that iteratively fuses a sequence language model with a 3D structure encoder through a shared learnable token.

Protein language model conditioned on ensembles of computed conformations, giving state-aware embeddings for interaction, localization, and function.

T-cell clonal expansion detection from scRNA-seq alone, without paired TCR sequencing. Trained on 2.6M pan-cancer T cells, reaching 0.85-0.96 AUROC.

Ligand-aware protein language model that cross-attends SaProt embeddings to ligand SMILES, beating SaProt across six downstream benchmarks.

Attention architecture fusing docking scores with protein language model embeddings, so an enzyme gets a different representation per substrate.

Generates free-text descriptions of protein function, catalytic activity, subcellular localization, and domains from sequence alone.

Histopathology image synthesis steered by a hand-drawn coarse tissue mask, refined into a fine-grained semantic mask before H&E or IHC rendering.

Histopathology semantic masks generated de novo from noise, then rendered as photorealistic H&E or PD-L1 patches by a paired image-translation GAN.