A site-resolved post-translational modification framework for drug-response analysis
Post-translational modifications (PTMs) regulate protein activity and treatment-associated signalling, providing functional information that can connect drug-response estimates with candidate pharmacodynamic measurements. Assessing this information requires a framework that determines its contribution beyond fixed molecular features while retaining the identity of the modified sites underlying model predictions. We developed the PTM Biological Dynamics Layer (PTM-BDL), which integrates typed PTM states with protein sequence, structure-derived and drug chemical features to predict log-transformed half-maximal inhibitory concentration and resistance while providing site-level interpretation through integrated gradients. Retrospective evaluation used source-derived and propagated PTM profiles across biologically distinct BCR-ABL, HDAC/EP300 and EGFR/ERBB2 settings, yielding internal areas under the receiver operating characteristic curve (AUROCs) between 0.785 and 0.946. Retraining demonstrated a consistent incremental contribution from PTM inputs in K562/CML, with a mean AUROC increase of 0.0268 and positive gains in all five folds, alongside more modest incremental gains in HeLa/HDAC and EGFR/ERBB2. Within K562/CML, CRKL Y207 and STAT5A Y694 retained the leading attribution ranks for their proteins across three seeds, while greater PTM attribution in dasatinib than in cytotoxic drug contexts identified a focused setting for pharmacodynamic investigation. These findings support context-dependent use of PTM information and connect predictive assessment with prioritisation of individual sites, providing a basis for targeted pharmacodynamic measurements and independent experimental validation.
Authors
- Moustafa Zein (ORCID: https://orcid.org/0000-0002-0253-8985)
- Aboul Ella Hassanien (ORCID: https://orcid.org/0009-0002-4833-3122)
Institutions
- Cairo University (EG)
Publication Details
- Journal
- bioRxiv (Cold Spring Harbor Laboratory)
- Published
- 2026-11-20
- DOI
- https://doi.org/10.64898/2026.11.19.767872
- Primary Topic
- Computational Drug Discovery Methods
- Type
- preprint