Experimental determination and data-driven prediction of homotypic transmembrane domain interfaces

Area of Science:

• Biochemistry
• Structural Biology
• Bioinformatics

Background:

• Single-pass membrane proteins assemble into complexes via transmembrane domain (TMD) interactions.
• The full extent of the TMD-TMD interactome is not yet understood.
• Predicting these interactions from primary sequence is a significant challenge.

Purpose of the Study:

• To systematically analyze the physical and evolutionary properties of homotypic TMD-TMD interfaces.
• To develop a predictive method for identifying TMD-TMD interaction sites.
• To advance the understanding of membrane protein complex assembly.

Main Methods:

• Generation of a dataset of 50 self-interacting TMDs, including nine experimentally identified human protein TMD interfaces.
• Analysis of residue conservation, co-evolution, polarity, and amino acid composition at interfaces.
• Training a machine learning algorithm (THOIPA) using identified interface features.

Main Results:

• Interfacial residues are more conserved, co-evolved, and polar than non-interfacial residues.
• Interface positions show a deficiency in β-branched residues and are located deep within the membrane's hydrophobic core.
• The GxxxG motif is overrepresented at interfaces, while other smallxxx(small) motifs are weakly represented.
• The THOIPA algorithm demonstrates high accuracy in predicting key interface residues.

Conclusions:

• Homotypic TMD-TMD interfaces possess distinct physical and evolutionary characteristics.
• These characteristics can be leveraged to predict interaction sites using computational methods.
• The THOIPA tool provides a valuable resource for studying membrane protein interactions and assembly.