Lipid discovery enabled by sequence statistics and machine learning

Area of Science:

• Microbiology
• Biochemistry
• Bioinformatics

Background:

• Bacterial membranes are crucial for cell function and are shaped by evolutionary pressures.
• The MprF enzyme modifies membrane lipids through covalent attachment of amino acids.

Purpose of the Study:

• To investigate the substrate specificity of the MprF enzyme.
• To identify novel MprF products and the organisms that synthesize them.
• To model evolutionary constraints on MprF sequences.

Main Methods:

• Comparative sequence analysis of MprF proteins across different bacterial species.
• Application of machine learning (restricted Boltzmann machine) to predict MprF substrate specificity.
• Identification and characterization of novel lipid products.

Main Results:

• Streptococcal MprF enzymes were found to synthesize lysyl-glucosyl-diacylglycerol (Lys-Glc-DAG).
• A novel MprF substrate, diglucosyl-diacylglycerol (Glc2-DAG), and its product were identified in Enterococcus.
• An expanded range of bacteria utilizing MprF for glycolipid modification was discovered.

Conclusions:

• MprF exhibits diverse substrate specificities across bacterial species.
• Machine learning approaches can effectively predict enzyme function and uncover novel biochemical pathways.
• The study expands the known repertoire of bacterial cationic lipids and the enzymes responsible for their synthesis.