FAVES: A Computational Framework for Evaluating Molecular Compliance Across Regulatory Jurisdictions

Abstract

We present FAVES (Flagging, Assessment, Verification, Evaluation, Screening), a computational framework for evaluating molecular compliance across eight regulatory jurisdictions — DEA, FDA, EPA, CWC, EU REACH, BTWC, Australia Schedule, and OPCW. The framework implements a three-tier detection system: whitelist matching against known safe compounds, direct name and identifier resolution against schedule databases, and scaffold analysis using substructure matching against controlled substance core structures.

Results

On a curated benchmark of 102 compounds (51 controlled, 51 clean), FAVES achieves 100% sensitivity (all controlled substances detected) and 100% specificity (no false positives on clean compounds). The framework processes each compound in under 100ms and is designed to run inline within computational chemistry pipelines rather than as a post-processing step.

Significance

Compliance screening in drug discovery is typically performed at the end of the pipeline — after months of work — when a regulatory flag means starting over. FAVES moves compliance to the beginning and every junction of the pipeline, eliminating late-stage surprises. The framework is integrated into NovoMCP as the check_compliance tool, screening every molecule at every compute stage automatically.