Using kernelized partial canonical correlation analysis to study directly coupled side chains and allostery in small...

Area of Science:

• Structural biology
• Computational biophysics
• Bioinformatics

Background:

• Understanding protein dynamics and communication pathways is crucial.
• Existing methods struggle with multivariate angular data representing side-chain conformations.
• Novel approaches are needed to infer direct couplings from complex protein data.

Purpose of the Study:

• To develop a novel method for inferring direct structural dependencies between protein side chains.
• To address the limitations of existing methods in handling multivariate angular data.
• To identify functional regions and interactions in allosteric proteins.

Main Methods:

• Developed kernelized partial canonical correlation analysis (KPCCA), an extension of CCA.
• Applied KPCCA to infer direct couplings between side chains, distinguishing them from indirect ones.
• Utilized conformational data and fluctuations from inactive structures of small G proteins.

Main Results:

• KPCCA successfully identified allosterically important residues in Ras and related protein families.
• The method highlighted densely connected regions in the inferred interaction graph.
• Results showed good agreement with empirical findings and revealed common allosteric characteristics across small G protein sub-families.

Conclusions:

• KPCCA is an effective tool for inferring direct couplings in protein side chains.
• The method aids in identifying functional regions and communication pathways in allosteric proteins.
• KPCCA demonstrates utility in comparative analysis of protein families like Ras, Rho, and Rab.