Enhancing PROTAC Ternary Complex Prediction with Ligand Information in AlphaFold 3

Accurate prediction of protein-ligand and protein-protein interactions is essential for computational drug discovery, yet remains a significant challenge, particularly for complexes involving large, flexible ligands. In this study, we assess the capabilities of AlphaFold 3 (AF3) for modeling ligand-mediated ternary complexes, focusing on PROteolysis-TArgeting Chimeras (PROTACs). PROTACs facilitate targeted protein degradation by recruiting E3 ubiquitin ligases to a protein of interest, offering a promising strategy for previously undruggable intracellular targets. However, their size, flexibility, and cooperative binding requirements pose significant challenges for accurate computational modeling. To address these challenges, we leverage AF3's inference code, which enables direct ligand incorporation, to predict 48 PROTAC-related complexes from the Protein Data Bank. We systematically evaluate AF3's predictive accuracy using RMSD, pTM, and DockQ scores, demonstrating that when ligand information is provided, AF3 achieves high structural accuracy, even for post-2021 structures that were not included in its training set. Additionally, we explore alternative input strategies---comparing molecular string representations (SMILES) versus explicit ligand atom positions---to refine ligand placement and improve interaction predictions. By analyzing the relationships between ligand positioning, protein-ligand interactions, and structural accuracy metrics, we provide insights into key factors influencing AF3's performance in modeling PROTAC-mediated ternary complexes. To ensure reproducibility, we publicly release our pipeline and results through an accompanying GitHub repository, providing a framework for future PROTAC structure prediction studies.