MemPype: a pipeline for the annotation of eukaryotic membrane proteins

Area of Science:

• Bioinformatics
• Computational Biology
• Molecular Biology

Background:

• Accurate prediction of membrane protein characteristics is crucial for understanding cellular functions.
• Existing tools often lack comprehensive analysis for signal peptides, GPI anchors, and subcellular localization.
• Eukaryotic membrane protein annotation requires integrated approaches for reliable classification.

Purpose of the Study:

• To develop and present MemPype, an integrated Python pipeline for eukaryotic membrane protein annotation.
• To enhance the prediction accuracy of signal peptides, GPI anchors, and subcellular localization.
• To provide a user-friendly web server for computational analysis of membrane proteins.

Main Methods:

• Integration of established methods: SPEP (signal peptides), PredGPI (GPI anchors), ENSEMBL (all-alpha membrane topology).
• Inclusion of MemLoci for discriminating subcellular localization: 'cell membrane', 'internal membranes', 'organelle membranes'.
• Utilizing homology-based inheritance and computational predictions, including homology searches and filtering for specific protein features.

Main Results:

• MemLoci achieved 70% accuracy and 0.50 GCC on a homology-unbiased validation set, outperforming other predictors.
• The pipeline accurately predicts signal peptides, GPI anchors (0.8% FPR), and all-alpha membrane topology (<1% FPR).
• Homology searches (≥50% identity, ≥50% coverage) aid in identifying membrane-associated proteins and localization tags.

Conclusions:

• MemPype offers a robust and accurate pipeline for comprehensive eukaryotic membrane protein annotation.
• The integrated approach, particularly MemLoci, significantly improves subcellular localization prediction.
• The available web server facilitates accessibility and application of these advanced computational tools in biological research.