AI-driven discovery of synergistic combinations from Psoralea corylifolia fruit for treating osteoporosis

  • Phytomedicine. 2026 Jun 11:159:158427. doi: 10.1016/j.phymed.2026.158427.
Chun-Lu Liu  1 ,  Ming-Rui Li  1 ,  Jing Long  1 ,  Si-Tong Qian  1 ,  Cai Zhang  1 ,  Ping Li  1 ,  Yan Jiang  2 ,  Hui-Jun Li  3
Affiliations
  • 1. State Key Laboratory of Natural Medicines, China Pharmaceutical University, 639 Longmian Road, Nanjing, 211198, China.
  • 2. College of Chemical Engineering, Nanjing Forestry University, No. 159 Longpan Road, Nanjing 210037, China. Electronic address: [email protected].
  • 3. State Key Laboratory of Natural Medicines, China Pharmaceutical University, 639 Longmian Road, Nanjing, 211198, China. Electronic address: [email protected].
PMID: 42308991 DOI: 10.1016/j.phymed.2026.158427
Abstract

Background: Estrogen-deficient osteoporosis demands safe and effective therapies. Psoralea corylifolia fruit (PCF), a phytoestrogenic herbal medicine for osteoporosis, lacks clarified key active components and synergistic combinations, which limits its application.

Purpose: This study established an AI-driven stepwise strategy to identify anti-osteoporotic synergistic combination from PCF, quantify their synergism, and validate pharmacodynamics, thus providing a reference for complex disease treatment.

Methods: An AI-driven stepwise integrated strategy was proposed: Ⅰ) Artificial intelligence-assisted preliminary screening of PCF active components via dual estrogenic/anti-osteoporotic models; Ⅱ) molecular simulation (docking/dynamics) for secondary screening of candidate components; Ⅲ) multi-dimensional evaluation for precision targeting of lead components; Ⅳ) Chou-Talalay method used to determine synergism (combination index, CI<0.9), additivity (0.9≤CI≤1.1), or antagonism (CI>1.1). Ultimately, the pharmacodynamic activity of the screened synergistic combination was validated through systematic in vitro and in vivo pharmacodynamic evaluations, combined bioequivalence analysis based on 90% confidence interval.

Results: Machine learning (ML) screening identified 20 active components, narrowed to 6 via molecular simulations. Three leads viz neobavaisoflavone (NBIF), bavachin (BA) and isobavachalcone (IB) were finalized. The NBIF+BA synergistic combination with significant synergism (CI<0.9) showed superior efficacy to individual components, and equivalence to PCF extract (90% confidence interval: 78.12%-132.15%).

Conclusion: This AI-driven strategy successfully identified a synergistic combination of NBIF+BA with potent anti-osteoporotic effects from PCF, presenting a promising approach for the treatment of complex diseases.

Keywords
AI-driven stepwise screening; Estrogen receptor; Machine learning; Osteoporosis; Psoralea corylifolia fruit; Synergistic combination.
Products