Extending the R Library PROPER to Enable Power Calculations for Isoform-Level Analysis with EBSeq.

Amadou Gaye

doi:10.3389/fgene.2016.00225

Abstract

RNA-Sequencing (RNA-Seq) has become a routine technology for investigating gene expression differences in comparative transcriptomic studies. Differential expression (DE) analysis of the isoforms of genes is just emerging now that expression (read counts) can be estimated with higher accuracy at the isoform level. Estimating the statistical power that can be achieved with a specific number of repeats is a key step in RNA-Seq analysis. The R library proper was developed to provide realistic empirical power analysis. However, proper uses differential expression methods more suited for power calculation of gene-level expression data. We propose extensions to this tool that would allow for power analysis which takes into account the specificities of isoforms expression. This was achieved by enabling the use of EBSeq, a DE approach well-tailored for isoform-level expression, as an additional analysis method within PROPER. The new extensions and exemplar code for their usage are freely available online at: https://github.com/agaye/proper_extension

Highlights

PROspective Power Evaluation (PROPER) (Wu et al, 2015) is an R library developed for power calculations in RNA-seq differential expression (DE) analysis
Since for isoform analysis, unlike gene-level analysis, isoform names are required in addition to the gene name, we inserted a check to ensure this requirement is met to prevent the function from crashing if input tables similar to those for edgeR, DEseq, and DSS are provided while EBSeq is specified as DE analysis method
We extended the package PROPER to allow for transcript isoform analysis and have a more comprehensive tool

Summary

INTRODUCTION

PROspective Power Evaluation (PROPER) (Wu et al, 2015) is an R library developed for power calculations in RNA-seq differential expression (DE) analysis. In addition to the usual parameters included in power estimation such as sample size, effect size, and within-group variance, PROPER takes into account other keys characteristics of count data that influence power including the distribution of the mean expression level, the sequencing depth and the threshold for filtering out molecules. EBSeq uses an empirical Bayesian approach to model a number of features observed in RNA-seq This tool is more suited for isoform level inference because it accommodates isoform expression estimation uncertainty by modeling the differential variability observed in distinct groups of isoforms (Leng et al, 2013) and takes into account colinearity between isoforms originating from the same. In EBSeq, a posterior probability of being differentially expressed (PPDE) is computed as measure of statistical significance

MATERIALS AND METHODS

Findings

CONCLUSIONS