Work

Toward Distributed Robustness in Resistance-Aware Lead Optimization

Year: 2026
Type: article
Abstract: Drug resistance often enters lead optimization after a compound has been chosen. We define distributed robustness as redistributing predicted binding support away from predefined mutable contacts while retaining wild-type feasibility and panel coverage. ResistAgent separates site-level and observed combination-level liabilities, diagnoses their structural effects, and compares a matched-budget robust objective with an average-affinity comparator. Across 10 EGFR pairs, the median reduction in hotspot dependence, ΔDep (naive − robust; positive values favor the robust objective), was +0.042; across 9 valid HIV-RT pairs, it was +0.026. Panel passing was preserved but not improved: both objectives passed in every valid pair. ABL1 yielded 2/4 pairs without a corresponding robustness gain. The objective changes the dependence organization and candidate selection. Provider comparisons, ranking benchmarks, and explicit-water contact descriptors are reported alongside the primary endpoint.
Language: English
Subfield: Drug Discovery
Open Access status: green
OpenAlex ID: W7165542069
FWCI:
Cites: 47
Cited by:
Related to: 8
Parts: 5