Benchmarking the Builders: A Comparative Analysis of PRosettaC and AlphaFold3 for Predicting PROTAC Ternary Complexes

Area of Science:

• Biochemistry and Structural Biology
• Computational Chemistry
• Drug Discovery

Background:

• Targeted protein degradation using Proteolysis-Targeting Chimeras (PROTACs) is a rapidly advancing therapeutic modality.
• Accurate prediction of ternary complex structures, comprising the target protein, E3 ligase, and PROTAC, is crucial for rational degrader design but remains challenging.
• Existing computational tools require rigorous benchmarking against experimental data.

Purpose of the Study:

• To systematically evaluate the performance of AlphaFold3 and PRosettaC in predicting ternary complex structures.
• To assess the impact of accessory proteins and linker sampling on prediction accuracy.
• To introduce a dynamic evaluation framework incorporating molecular dynamics simulations for more realistic benchmarking.

Main Methods:

• Benchmarking AlphaFold3 and PRosettaC against 36 experimentally resolved ternary complexes.
• Utilizing DockQ score to quantify the structural fidelity of predicted complexes.
• Employing molecular dynamics simulations to analyze transient conformational states and dynamic compatibility.

Main Results:

• AlphaFold3's accuracy can be overestimated due to the inclusion of non-specific accessory proteins.
• PRosettaC provides geometrically accurate models in specific cases but struggles with insufficient linker sampling.
• Dynamic analysis revealed that some PRosettaC models achieve high accuracy in transient states not captured by static evaluation.

Conclusions:

• Static benchmarking may not fully capture the performance of computational tools for PROTAC ternary complex prediction.
• Incorporating protein flexibility through molecular dynamics is essential for a more comprehensive evaluation.
• This study provides a refined framework for assessing in silico tools, aiding in the rational development of PROTACs.