Université de Montréal

Protein inverse folding model aligning ProteinMPNN by multi-objective preference optimization to improve developability without losing fold fidelity.

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

Protein conformation ensemble generation aligned to force-field energies, calibrating an AlphaFold 3-style diffusion model against MD thermodynamics.

AMPLIFY post-trained to absorb 3D structure from protein graph networks, lifting trRosetta contact precision from 0.253 to 0.320 at P@L/5.

ESM-2 post-trained to absorb 3D structure from protein graph networks, raising contact precision on a withheld CASP16 set by 59%.

Protein structure tokenizer that encodes 3D backbones as discrete VQ-VAE tokens, fixing the codebook under-utilization that caps their vocabulary.

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

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

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

EEG foundation model that pairs a convolutional encoder with a GPT backbone, pretrained by masked-segment reconstruction for low-data BCI decoding.

Zero-shot protein conformation sampling that perturbs an input structure and anneals it back with a score network trained only on crystal structures.