A Generic Approach to Biological Sequence Segmentation Problems: Application to Protein Secondary Structure Prediction

• Published in: Pattern Recognition in Computational Molecular Biology pp. 114 - 128
• Main Authors: Guermeur, Yann, Lauer, Fabien
• Format: Book Chapter
• Language: English
• Published: Hoboken, NJ, USA John Wiley & Sons, Inc 19.11.2015; Wiley
• Subjects: Bioinformatics; biological sequence segmentation; Computer Science; dynamic programming algorithm; hidden semi‐Markov model; Machine Learning; MSVMpred; multiclass support vector machines; neural networks; pattern recognition; protein secondary structure prediction; multiclass support vector machines; dynamic programming algorithm; neural networks; Hidden semi-Markov model; biological sequence segmentation; Pattern recognition; Protein secondary structure prediction; MSVMpred
• ISBN: 9781118893685; 1118893689
• DOI: 10.1002/9781119078845.ch7

This chapter reformulates the biological problems of interest as pattern recognition problems. It presents the bottom part of the hierarchy: MSVMpred. MSVMpred is a three‐layer cascade of classifiers. The base classifiers use different sets of descriptors all of which include the content of an analysis window. The base classifiers are of two types, namely multiclass support vector machines (M‐SVMs) with a dedicated kernel and neural networks (NNs). The chapter introduces the whole hybrid architecture and focuses on the features of the upper part of the hierarchy, that is, the specification and implementation of the generative model. Its dedication for protein secondary structure prediction is exposed in the chapter. For a given biological sequence, the final prediction is obtained by means of the dynamic programming algorithm computing the single best sequence of states, that is, the variant of the Viterbi algorithm dedicated to the hidden semi‐Markov model (HSMM) implemented.