Computational characterization of parallel dimeric and trimeric coiled-coils using effective amino acid indices

Area of Science:

• Protein structure and dynamics
• Bioinformatics and computational biology
• Biophysics

Background:

• Coiled-coils are common protein-protein interaction motifs formed by winding alpha-helices.
• Understanding coiled-coil oligomerization is crucial for designing novel protein structures.
• Existing prediction methods have limitations in accuracy and scope.

Purpose of the Study:

• To develop a novel, accurate predictor for coiled-coil oligomerization states.
• To leverage machine learning and amino acid indices for improved prediction.
• To provide insights into the sequence-to-structure relationship of coiled-coils.

Main Methods:

• Developed RFCoil, a predictor using random forest (RF) algorithm.
• Utilized non-redundant amino acid indices for feature extraction.
• Benchmarked performance using 10-fold cross-validation and an independent test set.

Main Results:

• RFCoil achieved an AUC of 0.849 on cross-validation and 0.855 on independent testing.
• RFCoil outperformed four existing coiled-coil prediction tools.
• Identified predominant rules governing coiled-coil oligomer formation from the RF model.

Conclusions:

• RFCoil offers a significant improvement in predicting coiled-coil oligomerization states.
• The extracted rules provide valuable insights into coiled-coil structure-function relationships.
• RFCoil facilitates the design of protein-based materials and therapeutics.