François Charih

This work introduces a transformer-based model for lysine methylation site prediction, achieving state-of-the-art accuracy through multitask learning and validation with mass spectrometry.

This study evaluates sequence-based protein-protein interaction predictors for therapeutic peptide engineering, finding similarity-based methods more effective than deep learning models for known targets.

The paper identifies two distinct astroglial phenotypes across mouse cortical development and reveals significant sex differences in developmental gene expression patterns.