YDS-Ternoplex: Surpassing AlphaFold 3-Type Models for Molecular Glue-Mediated Ternary Complex Prediction

Molecular glues represent an innovative class of drugs that enable previously impossible protein-protein interactions, but their rational design remains challenging, a problem that accurate ternary complex modeling can significantly address. Here we present YDS-Ternoplex, a novel computational approach that enhances AlphaFold 3-type models by incorporating enhanced sampling inductive bias during inference to accurately predict molecular glue-mediated ternary complex structures. We demonstrate YDS-Ternoplex’s capabilities across five diverse test cases, including both E3 ligase-based systems (VHL:CDO1 and CRBN complexes with mTOR-FRB, NEK7, and VAV1-SH3c) and non-E3 ligase complexes (FKBP12:mTOR-FRB). The model achieves remarkable accuracy with RMSD values as low as 1.303 Å compared to experimental structures and successfully predicts novel protein-protein interfaces not present in training data. Notably, in the FKBP12:mTOR-FRB case, YDS-Ternoplex correctly predicts a novel interface configuration instead of defaulting to known interactions present in training data, demonstrating strong generalization capabilities. Our results suggest that strategic enhancement of the inference process through inductive bias can significantly improve ternary complex prediction accuracy, potentially accelerating the development of molecular glue therapeutics for previously undruggable targets.