Virtual screening strategy for anti-DPP-IV natural flavonoid derivatives based on machine learning

Area of Science:

• Medicinal Chemistry
• Computational Chemistry
• Pharmacology

Background:

• Flavonoids exhibit antidiabetic effects by inhibiting dipeptidyl peptidase-IV (DPP-IV) activity.
• Structural variations in flavonoids significantly impact their DPP-IV inhibitory potency, complicating experimental discovery.
• An efficient screening pipeline is crucial for identifying novel, potent anti-DPP-IV flavonoid derivatives from natural products.

Purpose of the Study:

• To develop and apply a fusion strategy combining a Quantitative Structure-Activity Relationship (QSAR) model for discovering potent anti-DPP-IV flavonoid derivatives.
• To streamline the identification of natural product-derived flavonoids with significant DPP-IV inhibitory activity.

Main Methods:

• Construction of a high-quality QSAR model using a genetic algorithm (GA) with seven molecular property parameters, validated by leave-one-out cross-validation.
• Enrichment of 1,668 flavonoid derivatives from the Natural Product中國 database (NPCD) based on molecular fingerprint similarity (> 0.8).
• Screening of enriched derivatives using the QSAR model combined with the Quantitative Estimate of Druglikeness (QED) score, followed by ADMET analysis and molecular dynamics (MD) simulations.

Main Results:

• A robust QSAR model ($$ = 0.816, MAE_test = 0.14) was developed.
• A total of 33 flavonoid derivatives with predicted IC50 < 6.5 μM were identified.
• Three potent flavonoid derivatives were selected based on ADMET analysis and their DPP-IV inhibitory potential was confirmed via 100 ns MD simulation and MM/PB(GB)SA.

Conclusions:

• The developed QSAR-based fusion strategy is effective for discovering potent anti-DPP-IV flavonoid derivatives.
• The identified flavonoid derivatives show promise as novel therapeutic agents for diabetes.
• Computational approaches significantly accelerate the drug discovery process for natural product-derived compounds.