QUANTITATIVE STRUCTURE-ACTIVITY RELATIONSHIP (QSAR) MODELING OF BIOACTIVE COMPOUNDS FROM MANGIFERA INDICA FOR ANTI-DIABETIC DRUG DEVELOPMENT

Abstract

This study evaluates how quantitative structure–activity relationship (QSAR) modeling can accelerate anti-diabetic drug discovery from Mangifera indica (mango) phytochemicals. Diabetes mellitus remains a global health burden, and natural products represent an abundant but under-optimized resource for therapeutic leads. To address this gap, we conducted a comprehensive screening of major scientific databases and ultimately analyzed 113 peer-reviewed studies that reported computable molecular structures, harmonizable bioactivity endpoints, and reproducible modeling workflows. The review focused on clinically relevant antidiabetic targets including α-glucosidase, α-amylase, dipeptidyl peptidase-4 (DPP-4), and protein tyrosine phosphatase 1B (PTP1B). Across the included studies, curated datasets were normalized to pIC₅₀ and pKᵢ scales to enable meaningful comparisons, while feature engineering spanned physicochemical descriptors, topological indices, and molecular fingerprints such as ECFP and MACCS. Machine learning approaches ranged from penalized regression models to advanced ensemble algorithms (e.g., boosting, bagging, and kernel-based methods), with rigorous validation achieved through scaffold-aware data splits, external test sets, Y-randomization, and explicit applicability domain (AD) assessment.  Convergent lines of evidence—including QSAR predictivity within AD, mechanistically plausible docking, and ADME/toxicity filtering—consistently highlighted polyphenolic chemotypes, particularly xanthones such as mangiferin and its aglycone norathyriol, as promising inhibitors of carbohydrate-metabolizing enzymes. In contrast, scaffolds targeting signaling enzymes (DPP-4 and PTP1B) demanded early consideration of selectivity, pharmacokinetics, and off-target liabilities to improve translational viability. This synthesis also identifies recurring optimism traps, such as reliance on internal-only validation, assay heterogeneity across studies, and insufficient reporting of AD boundaries. To mitigate these challenges, we propose a reproducible translational framework: (i) employ AD-bounded QSAR as a first-line triage tool to prioritize scaffolds, (ii) integrate orthogonal structure-based approaches such as docking and molecular dynamics for rationalization of binding interactions, and (iii) adopt permeability-aware optimization and formulation strategies to address polarity-driven bioavailability challenges inherent to many mango-derived polyphenols.