AN INTRODUCTION TO SOME NEW PAPERS ON PROTEIN COMPLEX PREDICTION, RNA BLOCK ALIGNMENT, DE NOVO PEPTIDE SEQUENCING, METAGENOMIC BINNING, AND OTHER RESULTS

Abstract

This issue of the Journal of Bioinformatics and Computational Biology presents new results and discussions of several problems in the analysis of data from biological experiments. The eight papers of this issue are briefly summarized below. Protein interactions constitute a major aspect of all cellular processes. Consequently, analysis on the interactome is expected to produce several types of useful information such as protein function, 1 protein complexes, 2,3 and functional modules. 4 In this issue, Sohaee and Forst 5 describes a novel graph-theoretic clustering approach to detect functional modules from the interactome. The approach is based on the hypothesis that proteins within a complex should be within a short one to two hops from each other. This simple idea has achieved good recall and precision compared to existing methods. Gene expression profiling by microarrays for diagnostic and prognostic purposes has generated much excitement and research for the last 10 years. 6,7 In particular, prediction models based on various machine learning techniques have been applied to gene expression profiling data. In this issue, Khondoker et al. 8 present a largescale simulation study to investigate the relationship between the optimal number of genes for this type of prediction problems and various biological and technical factors, such as choice of classification algorithm, size of training samples, feature correlation, biological and technical variations, and minimum fold change. Multiple alignment of RNA structures has been studied for a while. 9 An RNA block is a structure-annotated multiple RNA sequence alignments. In this issue, Patel et al. 10 present BlockMatch, which is probably the first known method for aligning two RNA blocks. BlockMatch produces a high-quality alignment by considering the characteristics of all the sequences in the blocks along with their consensus structures during the alignment process. Experimental results of the authors show that phylogenetic trees produced by BlockMatch are more accurate than other widely used tools. De novo peptide sequencing is an important proteomic technique for identifying novel peptide sequences. 11 Generally, this has been done based on only one MS/MS spectrum. In this issue, He and Ma 12 propose using multiple spectra of the same v