Predicting Transmembrane Topology of β-barrel Membrane Proteins with A Hidden Markov Model

Very few methods address the problem of predicting β-barrel membrane proteins directly from sequence because of very few highresolution structures for transmembrane β-barrel proteins have been determined thus far. In this paper, we developed a novel method based on Hidden Markov Model (HMM) to model and predict the transmembrane regions of β-barrel membrane proteins. This model is evolved from an existing method with two important improvements in architectures of strand and β-turns. The resulting method achieved a prediction accuracy of 88% on a two-state test, which is better than many existing methods based on HMM or non- HMM. Furthermore, this model achieved a high accuracy of 89.7% on four-state topology predictions. All these results indicate that our method is powerful for transmembrane topology prediction of β-barrel membrane proteins.