Artificial intelligence empowers targeted protein degradation: Core technological innovations, multi‐scenario applications, and translational prospects
TL;DR
This review systematically explores AI applications in TPD development, covering the prediction and design of stable ternary complexes, rational optimization of linkers, high‐throughput screening for E3 ligase ligands, and accurate predictions of degradation efficiency and ADMET properties.
Abstract
Abstract Targeted protein degradation (TPD) has emerged as a transformative therapeutic strategy that offers unprecedented opportunities to eliminate traditionally “undruggable” proteins that have posed significant challenges in traditional drug development. Current TPD approaches, including proteolysis‐targeting chimeras (PROTACs), molecular glues, and lysosome‐targeting chimeras (LYTACs), encounter several limitations. These include the complexity of forming stable ternary complexes, suboptimal design of linkers, a limited repertoire of E3 ligases, and inadequate pharmacokinetic properties. Artificial intelligence (AI) has rapidly become essential in addressing these challenges, revolutionizing the TPD drug discovery process through data‐driven insights and predictive modeling. This review systematically explores AI applications in TPD development, covering the prediction and design of stable ternary complexes, rational optimization of linkers, high‐throughput screening for E3 ligase ligands, and accurate predictions of degradation efficiency and ADMET (Absorption, Distribution, Metabolism, Excretion, Toxicity) properties. Additionally, this review underscores AI's pioneering role in discovering molecular glues, from target identification to activity prediction, and discusses the AI‐driven optimization of emerging TPD modalities, such as LYTACs and PROTAC/IMiD bifunctional molecules. Despite significant progress, several critical challenges remain, such as the absence of standardized datasets, the static modeling of dynamic biological systems, and the opaque nature of advanced AI architectures. Future research should concentrate on integrating multi‐omics data to improve model training, developing dynamic and mechanistic AI frameworks, advancing explainable AI (XAI) to enhance mechanistic interpretability, and encouraging transdisciplinary collaboration to expedite clinical translation. By integrating AI with structural biology, pharmacology, and experimental validation, TPD technologies hold the potential to expand the druggable proteome and provide novel therapeutic solutions for cancer, neurological disorders, and other persistent diseases.