MPLID (Membrane Protein-Lipid Interaction Database): A Large-Scale Experimental Resource of Residue-Level...

Area of Science:

• Biochemistry
• Structural Biology
• Bioinformatics

Background:

• Membrane proteins are crucial drug targets, and their interactions with lipids are vital for function.
• Existing resources often infer membrane location rather than direct lipid contact.
• Understanding direct protein-lipid contacts is key for drug design and mechanistic studies.

Purpose of the Study:

• To create a residue-level dataset of experimentally verified protein-lipid contacts.
• To provide a resource for training machine learning models for lipid-binding site prediction.
• To differentiate direct lipid contacts from computationally defined membrane boundaries.

Main Methods:

• Curated dataset from experimentally resolved lipid molecules in Protein Data Bank structures.
• Utilized a 4.0 Å distance cutoff for defining lipid-contact residues.
• Proteins clustered at 30% sequence identity to prevent data leakage, with defined training, validation, and test splits.
• Analyzed amino acid composition at lipid-contact sites.

Main Results:

• The Membrane Protein-Lipid Interaction Database (MPLID) contains 80,439 lipid-contact annotations across 4,704 membrane proteins.
• Identified specific amino acid enrichments (e.g., tryptophan, arginine) and depletions (e.g., proline, isoleucine) at contact sites.
• Demonstrated distinct biological insights compared to existing membrane region prediction tools.

Conclusions:

• MPLID provides a large-scale, experimentally grounded resource for studying protein-lipid interactions.
• The dataset enables the development of accurate lipid-contact prediction models.
• MPLID supports structure-guided drug design and membrane protein engineering efforts.