SenolyticSynergy: An Attention-Based Network for Discovering Novel Senolytic Combinations via Human Aging Genomics

Abstract

Senolytics, a category of drugs targeting aging processes, have garnered significant attention since their emergence in 2015. Unlike traditional drug development approaches that rely on randomized screening, research on aging-related pharmaceuticals has employed mechanism-based strategies, resulting in the discovery of the pioneering combination therapy of dasatinib (D) and quercetin (Q). Although preliminary studies with senolytic drug combinations have shown promising outcomes, the predictive capabilities of the research in this field remain limited by the extensive experimental data requirements. In this study, we employed differential gene expression analysis and machine learning techniques to investigate the combinatorial effects of senolytic drugs. We identified 1624 core aging-related genes and used this dataset to retrain a multimodal attention mechanism model, creating a specialized framework, SenolyticSynergy, for predicting effective senolytic drug combinations. We then utilized 63 established senolytic compounds as starting points for combination testing, developing a comprehensive dataset of 1953 potential drug combinations for aging interventions. Following rigorous filtration, we identified 190 high-confidence drug combinations and predicted their synergistic scores. Among these combinations, ten demonstrated exceptionally high synergistic scores, exceeding 8. The combination of temsirolimus and nitazoxanide ranked first and may be the most promising candidate. The analysis of the literature data and computational studies of molecular structures using 3D modeling validated the accuracy of these predictions. This framework paves the way for large-scale research into anti-aging drug combinations, advancing research capabilities in this field.