Alexandre G. de Brevern

PATHOS predicts variant pathogenicity by integrating protein language models and biological features, achieving high accuracy on clinical and ClinVar datasets.

This study uses deep learning to design anti-VEGF nanobodies with improved affinity and stability through structure-guided mutational modeling and a sequence-based melting temperature predictor.

This study evaluates 65 variant effect predictors, finding AlphaMissense performs best and variants are classified as easy, moderate, or hard to predict based on structural and functional information.

The paper benchmarks PEPOP version 2.0 methods for predicting peptides mimicking discontinuous epitopes, finding most outperform SUPERFICIAL and chance.

PEPOP provides 35 peptide design methods to generate sequences mimicking discontinuous epitopes for antibody interactions, improving experimental applications.

The paper introduces kPred, a knowledge-based method for predicting protein backbone conformations using structural prototypes, achieving mean Q16 accuracies between 40.8% and 66.3%.

PBxplore analyzes protein structure and deformability using Protein Blocks, providing outputs like frequencies and entropy from molecular dynamics simulations.