Mem-mEN: Predicting Multi-Functional Types of Membrane Proteins by Interpretable Elastic Nets.

Area of Science:

• Proteomics
• Bioinformatics
• Computational Biology

Background:

• Membrane proteins are crucial for biological processes.
• Accurate prediction of membrane protein function is essential for their characterization.
• Current prediction tools lack interpretability and have limited performance.

Purpose of the Study:

• To develop an efficient and interpretable predictor for membrane protein functional types.
• To address the limitations of existing membrane protein prediction methods.
• To enable prediction of both single and multi-label functional types.

Main Methods:

• Utilizes Gene Ontology (GO) information retrieved via homologous accession numbers.
• Employs a multi-label elastic net (EN) classifier for functional type prediction.
• The predictor, Mem-mEN, generates sparse and interpretable solutions.

Main Results:

• Mem-mEN significantly outperforms state-of-the-art membrane protein predictors.
• Identified 338 essential GO terms out of over 7,900 for functional type determination.
• The predictor provides explanations for its functional type predictions.

Conclusions:

• Mem-mEN offers a superior and interpretable approach to membrane protein functional prediction.
• The identified essential GO terms enhance understanding of membrane protein roles.
• The developed tool aids in both prediction and explanation of membrane protein functions.