HDXRank: A Deep Learning Framework for Ranking Protein Complex Predictions with Hydrogen-Deuterium Exchange Data

Area of Science:

• Structural biology
• Computational biology
• Biophysics

Background:

• Accurate protein-protein complex structure modeling is crucial for understanding biological processes.
• Hydrogen-deuterium exchange (HDX) experiments offer insights into protein binding interfaces.
• Integrating HDX data into modeling can enhance prediction accuracy.

Purpose of the Study:

• To develop HDXRank, a graph neural network (GNN)-based framework for ranking protein complex structures using HDX data.
• To evaluate HDXRank's ability to improve the accuracy of protein complex modeling.

Main Methods:

• Developed HDXRank, a GNN framework for candidate structure ranking based on HDX data alignment.
• Trained HDXRank on a curated HDX dataset to capture local structural features.
• Integrated HDXRank with existing modeling tools (rigid docking, AlphaFold).

Main Results:

• HDXRank effectively ranks protein complex models by aligning with HDX experimental data.
• The framework successfully prioritizes functionally relevant models.
• HDXRank improved prediction quality across various protein targets.

Conclusions:

• HDXRank offers a versatile approach to enhance protein complex modeling by incorporating HDX data.
• This framework transforms HDX profile alignment into a valuable model quality metric.
• HDXRank shows significant potential for advancing the study of molecular recognition.