ProAffiMuSeq: sequence-based method to predict the binding free energy change of protein-protein complexes upon...

Area of Science:

• Computational biology
• Biochemistry
• Genetics

Background:

• Protein-protein interactions are crucial for cellular functions.
• Mutations disrupting these interactions can lead to diseases.
• Existing computational methods for predicting binding affinity changes often require complex structural inputs and lack functional class information.

Purpose of the Study:

• To develop a novel computational method, ProAffiMuSeq, for predicting changes in protein-protein binding free energy (ΔΔG).
• To incorporate sequence-based features and functional class information into the prediction model.
• To overcome the limitations of structure-dependent methods for analyzing mutations in protein-protein complexes.

Main Methods:

• Developed ProAffiMuSeq, a sequence-based computational method.
• Utilized sequence features and functional class information of protein-protein complexes.
• Employed 10-fold cross-validation and independent test datasets for performance evaluation.
• Validated the method on an external dataset.

Main Results:

• Achieved an average correlation of 0.73 (MAE 0.86 kcal/mol) in 10-fold cross-validation.
• Obtained a correlation of 0.75 (MAE 0.94 kcal/mol) on the test dataset.
• Demonstrated comparable performance to structure-based methods on external validation.
• Showcased the capability for large-scale analysis of mutations in protein-protein complexes.

Conclusions:

• ProAffiMuSeq accurately predicts changes in binding free energy using sequence and functional class data.
• The method provides a valuable tool for studying disease-causing mutations in protein-protein interactions.
• ProAffiMuSeq enables large-scale analyses without the need for protein complex structural information.