Synergistic Geroprotectors Mapping through Systems Machine Learning and Graph Neural Networks.

Area of Science:

• Geroscience
• Computational Biology
• Pharmacology

Background:

• Aging hallmarks are shared across multiple age-related disorders (ARDs).
• Current geroprotective interventions often lack mechanistic breadth, targeting isolated pathways.
• A systems-level strategy is needed to modulate interconnected aging biology regulators.

Purpose of the Study:

• To design a systems-level strategy targeting four key aging regulators: Nrf2/Keap1, mTORC1, AMPK, and SIRT1.
• To identify natural compounds capable of executing this integrated geroprotective strategy using machine learning.
• To predict synergistic combinations of natural compounds for comprehensive geroprotection.

Main Methods:

• Developed a machine learning model to classify geroprotectors targeting Nrf2/Keap1, mTORC1, AMPK, and SIRT1.
• Performed drug-likeness screening and molecular docking to assess compound properties and target binding.
• Utilized a graph neural network model to predict synergistic combinations of natural compounds for ARDs.

Main Results:

• Identified a subset of natural compounds with favorable pharmacokinetics and strong binding affinities to all four geroprotective targets.
• Predicted five high-confidence synergistic pairs of natural compounds, including Baicalein, Pectolinarigenin, Phloretin, and Demethoxycurcumin.
• Demonstrated the potential for a coordinated geroprotective response across diverse ARDs.

Conclusions:

• A systems-level approach targeting multiple aging hallmarks can yield sustained geroprotective effects.
• Computational methods can effectively identify natural compounds and combinations for integrated geroprotection.
• Predicted natural compound combinations offer a promising strategy for synergistic and comprehensive treatment of multiple ARDs.