Análisis de datos de RNA-Seq comparación de métodos para el estudio de expresión génica diferencial

Abstract

RNA-Seq is a next generation sequencing (NGS) procedure of the DNA for discovering, profiling and quantifying RNA transcripts, so this technology allows for measure gene expression. The analysis of RNA-Seq data, which comprise discrete counts of reads mapped to genes or transcripts, is made up of different parts: quality control checks on raw sequence data,mapping reads to a reference genome, counts of reads and the detection of differentially expressed genes across different biological conditions. The Bioconductor project provides tools for identifying differential expression for RNA-Seq data by means of the use of R statistical programming language. We focus our attention on four Bioconductor packages: EdgeR, NOISeq, DESeq2 and Limma. Now let’s see the main features of these packages. - EdgeR: This package is based on the negative binomial distribution. The default method for the normalization of data is the trimmed mean of M-values (TMM). For filter out transcripts with very low counts, EdgeR establishes a minimum of counts per millions (CPM). Finally, it uses the exact test for the binomial negative to determinate differentially expressed genes. - NOISeq: It is based on non-parametric approaches for the differential expression analysis of RNA-Seq data. In this case, the normalization default technique implemented is RPKM (reads per kilobase per million). The method to filter out transcripts with low counts is CPM (counts per millions). The differential expression analysis between two experimental conditions is based on the probability of each transcript or gene of being differently expressed: it is obtained by comparing the differential expression statistics, M-D values, of that transcript or gene against distribution of changes in expression values when comparing replicates within the same condition, noise distribution. - DESeq2: It is based on negative binomial generalized linear models. The technique used for the normalization is normalisation based on the estimation of the effective library size. DESEq2 filters out the transcripts which very low counts by means of the mean of normalized counts for each gene. Finally, it uses the Wald test for determining differentially expressed genes. - Limma: This Bioconductor package was designed originally for the analysis of Microarray data, but it has been adapted for the analysis of RNA-Seq data. It is based on the use of linear models. The normalization default technique is the trimmed mean of M-values (TMM). To filter 3 out transcripts with very low counts, Limma uses the minimum of counts per millions (CPM). This package uses the t-test for determining differentially expressed genes. We analyze RNA-Seq data of eight patients with asthma. There are four untreated patients and four patients with dexamethasone, a potent glucocorticoid. We want to identify the number of differentially expressed transcripts between these two experimental conditions: untreateddexamethasone. It can be done by using the four Bioconductor packages seen previously. The obtained results are: 853 differentially expressed transcripts with EdgeR, 415 transcripts with NOISeq, 1212 transcripts with DESeq2, and 719 with Limma. These four packages agree on 277 differentially expressed transcripts between untreated patients and treated (dexamethasone) patients.