DegradeQuery: AI Framework Predicts PROTAC Degradation from Context

A new prediction framework named DegradeQuery has been developed by researchers to anticipate protein degradation induced by PROTACs. These proteolysis-targeting chimeras function by linking a target protein to an E3 ubiquitin ligase, initiating degradation, which is influenced by both the degrader molecule and its biological environment. Although numerous structured records of molecules, targets, and E3 ligases exist in public databases, only a limited number have degradation measurements, leaving many chemical-biological interactions untapped in supervised learning. DegradeQuery transforms records lacking labels into a pretraining signal. Its counterfactual tuple pretraining approach compares existing tuples with alternatives by altering the target, E3 ligase, or both, allowing the model to grasp contextual links without assigning pseudo-labels. The refined representation is then adjusted to predict degradation based on the complete molecule-target-E3 tuple. This innovation, outlined in a paper on arXiv (arXiv:2608.10595), has the potential to greatly improve drug discovery efficiency by enhancing the accuracy of degradation outcome predictions, thereby minimizing the necessity for extensive experimental validation.