SELFormer-Guided Discovery of Xanthohumol and Cirsilineol as Multi-Target Natural Therapeutics for Type 2 Diabetes: Computational Prediction and Experimental Validation

Abstract

Type 2 diabetes mellitus(T2DM) requires multi-target therapeutic approaches addressing both insulin resistance and secretion deficits. Natural compounds are promising candidates with the potential to act on multiple targets, but systematic identification remains challenging. To establish a computational framework for identifying natural compounds with multi-target therapeutic potential against T2DM through integrated structure-activity analysis and experimental validation. We developed an SELFormer deep learning model to predict natural compound activities against six T2DM-related proteins(Glucagon-like peptide-1 receptor[GLP1R], Kinesin Family Member 11[KIF11] for insulin secretion; Insulin Receptor[INSR], Peroxisome Proliferator-Activated Receptor Gamma[PPARG], Fibroblast Growth Factor Receptor 1[FGFR1], Insulin-like Growth Factor 1 Receptor[IGF1R] for insulin resistance). Uniform Manifold Approximation and Projection(UMAP) for dimension reduction clustering characterized chemical space distributions and molecular docking validated multitarget binding. Selected compounds were experimentally validated using 3T3-L1 adipocytes and Mouse Insulinoma(MIN6) pancreatic β-cells. The SELFormer model achieved a robust performance(R² = 0.937, RMSE = 0.331 for training; R² = 0.918, RMSE = 0.353 for testing).Using DeepChem models, natural compounds were selected from the ligands generated by the SELFormer model. Among the natural compounds analyzed, 35 demonstrated high activity (pIC50 > 7), clustering into eight distinct chemical families with diverse physicochemical properties. Multi-target network analysis revealed curcumin, xanthohumol, hesperetin, (-)epicatechin, and cirsilineol as ligands targeting proteins related to T2DM. They had low binding energy to the target proteins (-7 to -10). Food source analysis identified strawberry, grape, and tea varieties as rich sources of bioactive compounds. In 3T3-L1 adipocytes, curcumin, xanthohumol, hesperetin, (-)-epicatechin, and cirsilineol significantly enhanced insulin-stimulated glucose uptake at a 10 μM to achieve an efficacy comparable to metformin.In MIN6 cells, xanthohumol and cirsilineol increased glucose-stimulated insulin secretion to levels similar to exendin-4, while curcumin, hesperetin, and (-)-epicatechin showed modest but significant increases in insulin secretion. In conclusion, this integrated framework successfully identified natural compounds with validated therapeutic potential in T2DM by targeting key dual pathways. The identification of these compounds with multiple targets provides a foundation for evidence-based dietary interventions and natural product therapeutics, establishing a systematic methodology for the discovery of drugs for complex metabolic disorders.