Beyond Expression Profile Clustering
Author | : |
Publisher | : |
Total Pages | : 0 |
Release | : 2016 |
ISBN-10 | : OCLC:945087475 |
ISBN-13 | : |
Rating | : 4/5 (75 Downloads) |
Download or read book Beyond Expression Profile Clustering written by and published by . This book was released on 2016 with total page 0 pages. Available in PDF, EPUB and Kindle. Book excerpt: We present methods for comparative exploration of high-dimensional biological time-series datasets. In this work we focus on the analysis of gene expression data measured by either microarry or RNAseq. However, the methods we develop are applicable to a range of other high-dimensional datasets, including protein and metabolite datasets, as they become increasingly available. The datasets we work with consist of expression profiles for thousands or tens of thousands of gene. Each of these expression profiles is a time series of expression values for a specific gene. Expression profiles for the genes are sampled under two or more related conditions such as chemical/vehicle exposure or gene suppression/overexpression. Identifying similarities and differences across expression profiles can provide a better understanding of the relationships among the experimental conditions in the dataset. In pursuit of this, we present three novel methods for the analysis of multi-conditional time-series datasets. First, we introduce an approach called Multi-View Clustering. Given a dataset with expression profiles measured under two related conditions, Multi-View Clustering clusters genes on both expression profiles and alignments, where the alignment for a gene represents the correspondences between its expression profiles across conditions. Second, we describe Stable Alignment Clustering, an improved alignment clustering approach which harnesses a common method of assessing cluster stability to counteract instability in alignment clusters. Third, we present ConditionalMotif Finding a method for detecting patterns in time-series datasets where the genes are measured under more than two conditions. By grouping genes and conditions into subsets based on alignment similarity across conditions, this method characterizes the experimental conditions in the dataset by a set of motifs: gene/condition clusters that describe common patterns of change in the dataset.