Improved identification of outer membrane beta barrel proteins using primary sequence, predicted secondary structure, and evolutionary information

Abstract

Membrane proteins (MPs) are difficult to identify in genomes and to crystallize, making it hard to determine their tertiary structures. MPs could be categorized into α-helical (AMP) and outer membrane proteins which mostly include beta barrel folds (OMBBs). The AMPs are relatively easy to predict from a protein sequence because they usually include several long membrane-spanning hydrophobic α-helices. The OMBBs play important roles in cell biology, they are targeted by multiple drugs, and they are more challenging to identify as they have shorter membrane-spanning regions which lack a folding pattern, that is, as consistent as in the case of the AMPs. Hence, accurate in silico methods for prediction of OMBBs from their primary sequences are needed. We present an accurate sequence-based predictor of OMBBs, called OMBBpred, which utilizes a Support Vector Machine classifier and a custom-designed set of 34 novel numerical descriptors derived from predicted secondary structures, hydrophobicity, and evolutionary information. Our method outperforms modern existing OMBB predictors and achieves accuracy of above 98% when tested on two existing benchmark datasets and 96% on a new large dataset. OMBBpred reduces the error rates of the second best method, depending on the dataset used, by between 13 and 65%, and generates predictions with high specificity of above 96%. Our solution is a useful tool for high-throughput discovery of the OMBBs on a genome scale and can be found at http://biomine.ece. ualberta.ca/OMBBpred/OMBBpred.htm.

TprC/D (Tp0117/131), a Trimeric, Pore-Forming Rare Outer Membrane Protein of Treponema pallidum, Has a Bipartite Domain Structure

“…Because the T. pallidum genome does not encode orthologs for known OMPs of proteobacteria other than TP0326/BamA (22), in silico identification of ␤-barrels rests upon structural predictions. Whereas the ␣-helical transmembrane domains of IM proteins are readily identifiable, such is not the case for the membrane-spanning elements (i.e., amphipathic ␤-strands) of OMPs (41,56,63).…”

PredRSA: a gradient boosted regression trees approach for predicting protein solvent accessibility

“…Each element in the PSSM represents the probability of each residue position in the multiple sequence alignment. Plenty of previous studies have shown that multiple sequence alignments in the form of PSSM can substantially improve overall prediction performance [ 35 – 38 ]. In this article, the PSSM profile for each protein sequence is generated with default parameters (3 iterations and 0.001 of E-value cutoff) against the non-redundant (nr) dataset obtained from the NCBI.…”

Prediction of ketoacyl synthase family using reduced amino acid alphabets

“…Classification protocol SVM is a very powerful and popular method for supervised pattern recognition and has been widely used in the realm of bioinformatics [3,6,29,37,38,42,[44][45][46]. To handle a multiclass problem, the one-versus-one (OVO) and oneversus-rest (OVR) approaches are generally applied to extend traditional SVM.…”