Mem-ADSVM

Mem-ADSVM predicts membrane proteins and assigns their multifunctional membrane types using a two-layer SVM-based multi-label classification approach for proteomic functional annotation.

Key Features:

• Two-Layer Prediction Architecture: A sequential architecture with Layer I for membrane/non-membrane identification and Layer II for multi-label functional type assignment.
• Layer I (Binary SVM): A binary support vector machine classifies proteins as membrane or non-membrane based on extracted GO information.
• Layer II (Multi-Label Multi-Class SVM with Adaptive-Decision Scheme): A multi-label multi-class SVM equipped with an adaptive-decision scheme assigns one or more functional membrane types to proteins identified as membrane.
• Gene Ontology (GO) Information Utilization: The method retrieves GO information by searching a compact GO-term database using the protein's homologous accession number to generate input features for classification.

Scientific Applications:

• Proteomics research: Identification and multi-label categorization of membrane proteins to support studies of protein function and membrane-associated processes.
• Membrane protein functional annotation: Assigning one or more functional types to membrane proteins to enable analysis of multifunctionality in proteomic datasets.

Methodology:

Retrieve GO information from a compact GO-term database via homologous accession number, apply a binary SVM to classify membrane versus non-membrane proteins, and use a multi-label multi-class SVM with an adaptive-decision scheme to assign one or more membrane functional types.