Phobius API

Phobius API predicts transmembrane protein topology and signal peptides to distinguish transmembrane helices from signal peptides for accurate protein annotation.

Key Features:

• Integrated Prediction Model: Combines prediction of transmembrane protein topology and signal peptide presence within a single framework that accounts for their interplay.
• Hidden Markov Model (HMM) Approach: Uses a hidden Markov model that models sequence regions of signal peptides and transmembrane proteins through interconnected states to differentiate similar hydrophobic sequences.
• Curated Training Dataset: Trained on a newly assembled and curated dataset to enhance predictive accuracy compared with TMHMM and SignalP.
• Reduced Error Rates: Significantly lowers false classifications, reducing the error rate for signal peptides from 26.1% to 3.9% and for transmembrane helices from 19.0% to 7.7%.
• Whole-Genome Annotation: Applied to the proteomes of Homo sapiens and Escherichia coli, demonstrating a substantial reduction in false classifications compared with TMHMM/SignalP and suitability for comprehensive genome annotation.

Scientific Applications:

• Protein Structure Prediction: Provides orientation and location of transmembrane regions to inform structure–function analyses of membrane proteins.
• Genomic Annotation: Reduces false positives in genome annotation for identifying membrane-spanning and secreted proteins.
• Biotechnological Applications: Informs design of recombinant proteins in drug discovery and synthetic biology by predicting signal peptides and localization.

Methodology:

Phobius uses a hidden Markov model that models signal peptide and transmembrane sequence regions through interconnected states and was trained on a newly assembled, curated dataset while integrating topology and signal peptide prediction in a single framework.