Protein pathogenicity prediction
PATHOS (PATHOgenicity Scoring) is a deep learning-based tool designed to predict the pathogenicity of protein variants directly from sequence data. The tool analyzes mutations and predicts their pathogenic potential using advanced machine learning models. PATHOS provides pathogenicity scores with a threshold of 0.63, where scores above this threshold indicate likely pathogenic variants. The tool processes multiple variants simultaneously, providing reliable pathogenicity assessments for clinical and research applications.
Database Coverage
Our database contains precomputed
predictions for
~17,734 proteins from the human proteome, with over
~140 million mutations from SwissProt.
For proteins not (yet) in our database or large batch jobs, use the
standalone PATHOS package
on GitHub to run predictions locally.
If you use PATHOS, please cite:
Radjasandirane R., Cretin G., Diharce J., de Brevern A.G., & Gelly J.C. (2026). PATHOS: Predicting Variant Pathogenicity by Combining Protein Language Models and Biological Features. Artificial Intelligence in the Life Sciences, 100165. doi:10.1016/j.ailsci.2026.100165