Abstract
Signal peptides recognition by bioinformatics approaches is particularly important for the efficient secretion and production of specific proteins. We concentrate on developing an integrated fuzzy Fisher clustering (IFFC) and designing a novel classifier based on IFFC for predicting secretory proteins. IFFC provides a powerful optimal discriminant vector calculated by fuzzy intra-cluster scatter matrix and fuzzy inter-cluster scatter matrix. Because the training samples and test samples are processed together in IFFC, it is convenient for users to employ their own specific samples of high reliability as training data if necessary. The cross-validation results on some benchmark datasets indicate that the fuzzy Fisher classifier is quite promising for signal peptide prediction.
Keywords: Fuzzy Fisher clustering, fuzzy scatter matrix, optimal discriminant vector, signal peptides, secretory protein recognition, Secretory proteins, Fisher linear discriminant analysis, Hidden Markov method (HMM), Non-secretory proteins, Fuzzy Fisher Criterion (FFC), cale-wavelet energy, clustering algorithm, Continuous wavelet transformation, protein sequences, jackknife test, 5-fold cross-validation, cellular automaton, fuzzy support vector machine network, Terahertz frequency, Mahalanobis distanceFuzzy Fisher clustering, fuzzy scatter matrix, optimal discriminant vector, signal peptides, secretory protein recognition, Secretory proteins, Fisher linear discriminant analysis, Hidden Markov method (HMM), Non-secretory proteins, Fuzzy Fisher Criterion (FFC), cale-wavelet energy, clustering algorithm, Continuous wavelet transformation, protein sequences, jackknife test, 5-fold cross-validation, cellular automaton, fuzzy support vector machine network, Terahertz frequency, Mahalanobis distance