Effect Factor-Guided QHMCAR: A Chemometric Framework for Constituent-to-Formulation Modeling of Antioxidant-Related Analytical Responses

Standardized interpretation of analytical responses in multi-herb formulations remains difficult: constituent measurements, composition ratios, and extraction conditions are rarely integrated quantitatively. Effect Factor (EF) provides a dimensionless, endpoint-aware index for matched-condition referencing, but constituent-level EF descriptors do not directly yield formulation-level expectations. Using an EF-indexed resource comprising 271 herbal materials and 22 multi-herb formulations with paired water and 30% ethanol extracts, we developed an EF-guided quantitative herbal medicine composition–activity relationship (QHMCAR) framework. Additive theoretical formulation EFs for 2,2′-azino-bis(3-ethylbenzothiazoline-6-sulfonic acid) (ABTS), 2,2-diphenyl-1-picrylhydrazyl (DPPH), total phenolic content (TPC), and total flavonoid content (TFC) were derived from constituent EFs, composition ratios, and solvent-matched extraction yields. Analyses included predictor screening, least-squares calibration, additive-deviation analysis, diagnostics, fixed-form refitting, nested leave-one-prescription-identity-out (LOPO), and solvent sensitivity. Retained full-data models showed apparent R2 of 0.748–0.830; fully nested grouped LOPO yielded Q2 of 0.543–0.809. Against a grouped-LOPO untransformed linear baseline, transformation screening modestly improved DPPH prediction, matched ABTS, and yielded lower Q2 for TPC and TFC. Positive additive deviations were endpoint-dependent, most pronounced for ABTS; solvent-sensitivity analysis identified limited additional variation except for an ABTS-specific intercept shift. QHMCAR enables transparent formulation-level expectation, calibration, and internal validation without implying pharmacological synergy, therapeutic efficacy, or external predictive validity.