Geometric Deep Learning Revolutionizes Multi-Target Drug Design

This comprehensive review explores the transformative potential of Geometric Deep Learning (GDL) in the fields of polypharmacology and multi-target drug design, signaling a shift away from the conventional "one drug, one target" paradigm. The traditional approach often falls short in treating complex multifactorial diseases due to issues like compensatory signaling pathways and drug resistance. GDL offers a synergistic strategy by leveraging non-Euclidean molecular data to capture intricate three-dimensional structural interdependencies. The review systematically surveys various GDL architectures, from invariant graph neural networks to SE(3)-equivariant diffusion models. It critically analyzes their applications across three key areas: characterizing shared binding pockets through geometric embeddings, predicting multi-target bioactivity via heterogeneous graph fusion, and the de novo generation of ligands capable of simultaneously targeting multiple proteins. Special emphasis is placed on emerging structure-conditioned generative algorithms that combine diffusion models with reinforcement learning to autonomously resolve complex geometric conflicts between competing binding sites. The paper also underscores the importance of multimodal omics integration and specialized geometric benchmarking for validating these advanced models, positioning GDL as a pivotal force in rational, structure-driven polypharmacological molecular engineering.