A Hidden Markov Model method, capable of predicting and discriminating beta-barrel outer membrane proteins

Area of Science:

• Bioinformatics
• Structural Biology
• Genomics

Background:

• Integral membrane proteins are crucial, with beta-barrel proteins posing prediction challenges.
• Existing methods struggle with predicting beta-barrel transmembrane segments accurately.
• Outer membrane proteins in gram-negative bacteria are a key focus due to their structural complexity.

Purpose of the Study:

• To develop a robust method for predicting transmembrane beta-strands in outer membrane proteins.
• To discriminate beta-barrel outer membrane proteins from water-soluble proteins in large datasets.
• To enable the screening of entire proteomes for novel outer membrane protein discovery.

Main Methods:

• Utilized a Hidden Markov Model (HMM) trained in a discriminative manner.
• Employed a jackknife procedure and self-consistency tests for rigorous evaluation.
• Developed a web interface for non-commercial users to access the predictor.

Main Results:

• Achieved high per-residue accuracy (88.1%) and correlation coefficient (0.824) in self-consistency tests.
• Successfully predicted topologies for 9 out of 14 proteins in jackknife tests.
• Demonstrated high discrimination rates for outer membrane (88.8%) and water-soluble (89.2%) proteins.

Conclusions:

• The developed HMM-based strategy effectively screens proteomes for outer membrane proteins.
• The method is suitable for discovering novel beta-barrel outer membrane proteins.
• The freely available web interface provides a unique resource for beta-barrel outer membrane protein topology prediction.