CNRS

Modality-agnostic transformer pretrained by masked abundance reconstruction on 48,837 proteomics profiles reprocessed from 1,397 PRIDE projects.

Anti-phage defense gene predictor fusing protein language model embeddings with a contrastive genomic-context transformer over 64-gene windows.

Protein-protein interaction predictor fusing evolutionary and structural embeddings to screen bacterial and host-pathogen proteomes in minutes.

Genomic language model predicting drug-induced translational readthrough at premature stop codons, with an R2 of 0.94 across eight compounds.

End-to-end protein structure alignment from 3D coordinates, matching TM-align quality at LDDT 0.56 and beating Foldseek on SCOPe40 search.

Single-cell foundation model pre-trained on 50 million cells that infers cell-specific gene regulatory networks from transformer attention matrices.

Protein conformational motion prediction from a single structure, using an SE(3)-equivariant GNN trained on ensembles mined from the PDB.

Retinal encoding models that predict ganglion cell responses to visual stimuli, with pretrained checkpoints across four species and two modalities.

Protein sequence-structure encoder biasing attention with relative residue-frame transforms, for inverse folding and peptide binding prediction.

Protein-protein binding interface prediction from conformational ensembles, resolving interfaces in flexible and intrinsically disordered regions.

RNA sequence design model that generates protein-binding RNAs from a target structure alone, growing sequences outward from an anchored seed.

Liquid-liquid phase separation predictor that scores proteins and residues from sequence and designs phase-separating peptides by gradient descent.

Genomic language model reading bacterial gene neighborhoods as sentences of protein-family tokens to predict anti-phage defense function.

Protein language model fine-tuned to score any bacterial protein for anti-phage defense function, detecting homology too remote for HMM profiles.

Spike inference from calcium imaging traces, driven by a multistate GCaMP kinetic model that generates the synthetic data its decoders are trained on.

Phage protein function annotation from sequence, assigning hierarchical functional categories from frozen protein language model embeddings.

RNA backbone torsion and pseudo-torsion angle prediction from sequence alone, built by fine-tuning a DNABERT checkpoint on solved RNA structures.

Frugal conditional diffusion model that samples human SNP haplotypes in PCA space, producing artificial genomes for a chosen continental ancestry.

Protein motion prediction from sequence alone, mapping language model embeddings to continuous 3D displacement vectors with a lightweight CNN.

Graph attention foundation model for spatial transcriptomics that assigns spatial domains zero-shot across gene panels, tissues, and technologies.

A Wasserstein GAN that generates artificial human genomes in PCA space, synthesizing 65,535-SNP haplotypes for 26 worldwide populations.

De novo protein backbone generation by SE(3) diffusion over rigid-body frames. Samples designable monomers up to 500 residues without pretraining.

Nucleus instance segmentation for fluorescence microscopy and H&E histology, predicting a star-convex polygon per pixel to separate crowded nuclei.