AI-aided drug development for protein degraders: Design, lead identification, and optimization

In recent years, Targeted Protein Degradation (TPD) has evolved as an innovative pharmaceutical paradigm with the potential to overcome limitations of classical protein activity-modulating drugs. This is mainly because TPD engages the cell's natural protein degradation machinery (i.e., ubiquitin proteasome system) to deplete the target protein. Hence, TPD reaches beyond protein functions and its inhibition, e.g., protein scaffolding effects. Moreover, TPD-driven molecular degraders offer better target selectivity and differential resistance profiles. Prominent among classes of molecular degraders is PROTAC (PROteolysis TArgeting Chimera), which is a heterobifunctional molecule consisting of three building blocks (i.e., E3 binder-linker-warhead). The enforced proximity of a target protein and a E3 ligase through a warhead and a E3 binder, respectively, ubiquitinates the protein, thereby signaling its proteolysis. This modality has opened up avenues in targeting 80 % of proteins which were previously deemed ‘undruggable’. In fact, a number of ML/AI based methods have already been developed for de novo design of PROTACs and prediction of activities and degradation capacities. Lately, the design of efficient building block libraries and classification models for identification of new building blocks are being increasingly popular, especially linker as its nature and length is crucial in PROTAC functionality and developability. In this book chapter, we provide a comprehensive summary of existing ML/AI methods in the PROTAC universe by covering topics related to de novo design and its importance, in silico methods for understanding important features, optimization of pharmacokinetic properties and their possible implications as a revolutionary drug class.