Residue-Conditioned In-Pocket Design for Paralogue Selectivity
Octara is a reasoning-guided framework for sequential, residue-conditioned molecular design within protein binding pockets. The architecture integrates literature-derived scaffold selection, protein structure analysis, target–anti-target alignment, and chemically defined molecular modifications. A structural containment operator preserves previously incorporated molecular features throughout iterative optimization. Across 18 attempted receptor-pair campaigns, 17 produced final molecular leads, with the selected starting scaffold preserved in all 17. Comparison against 44 marketed and reference compounds spanning 12 receptor pairs demonstrated complete retention of annotated primary recognition motifs in 19 comparisons, against a 12.6% unrelated-active decoy rate. The generated candidates exhibited different medicinal chemistry strategies, including recognition-framework preservation, residue-directed electrophilic modifications, and alternative molecular architectures. These included a Cys388-directed chloroacetamide in ABL1 and a macrocyclic peptidomimetic candidate for KRAS G12C. The study demonstrates a computational framework for translating structural evidence and medicinal chemistry knowledge into traceable molecular design decisions across multiple protein targets. All reported candidates are computational designs; no compounds were synthesized or experimentally assayed.
Authors
- Ahmed amine Sadouk (ORCID: https://orcid.org/0009-0009-3258-6601)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-09-20
- DOI
- https://doi.org/10.5281/zenodo.22879768
- Primary Topic
- Computational Drug Discovery Methods
- Type
- preprint