TimesVector: a vectorized clustering approach to the analysis of time series transcriptome data from multiple phenotypes

• DOI: 10.1093/bioinformatics/btw780
• Publication Date: 2016-12-05T20:05:40Z
• Authors: Inuk Jung; Kyuri Jo; Hyejin Kang; Hongryul Ahn; Youngjae Yu; Sun Kim
• Keywords: Expression (computer science); Biclustering; Sample (material)

Abstract Motivation Identifying biologically meaningful gene expression patterns from time series data is important to understand the underlying biological mechanisms. To identify significantly perturbed sets between different phenotypes, analysis of transcriptome requires consideration and sample dimensions. Thus, such seeks search that exhibit similar or two more conditions, constituting three-dimensional data, i.e. gene-time-condition. Computational complexity for analyzing very high, compared already difficult NP-hard dimensional biclustering algorithms. Because this challenge, traditional clustering algorithms are designed capture co-expressed genes with pattern in conditions. Results We present a triclustering algorithm, TimesVector, specifically distinctively TimesVector identifies clusters distinctive three steps: (i) dimension reduction time-condition concatenated vectors, (ii) post-processing detecting distinct (iii) rescuing unclassified clusters. Using four generated by both microarray high throughput sequencing platforms, we demonstrated successfully detected quality. improved quality existing tools only differential across conditions successfully. Availability Implementation The software available at http://biohealth.snu.ac.kr/software/TimesVector/. Supplementary information Bioinformatics online.