Bioinformatics
Bioinformatics
Computational Biology and Chemistry
Bio-basis function neural network for prediction of protease cleavage sites in proteins
IEEE Transactions on Neural Networks
| Hi-index | 0.00 |
Transmembrane proteins are difficult to crystallize owing to the presence of lipid environment and the number of membrane protein structures deposited in Protein Data Bank is limited. Hence, computational techniques become essential and powerful tools to aid biologists for understanding the structure and function of membrane proteins. We propose an architecture for discriminating transmembrane α-helical proteins and transmembrane β-barrel proteins from genomic sequences, and then predict their transmembrane segments with Z-coordinate idea and RBF networks regression techniques. In the discrimination of transmembrane proteins, our approach has correctly predicted the transmembrane proteins with a cross-validated accuracy of more than 98% in a set of 5888 proteins, which contain 424 α-helical proteins, 203 β-barrel proteins, and 5261 globular proteins. Also, our method showed a TM-segment recall of 97.3% in a independent set of 41 α-helical proteins. The improvement of TM-segment recall is more than 9% when comparing with other modern α-helix transmembrane segment predictors.