ShrinkBayes: a versatile R-package for analysis of count-based sequencing data in complex study designs.

Henk MW Verheul,Mark A van de Wiel,Maarten Neerincx,Daoud Sie,Tineke E Buffart

doi:10.1186/1471-2105-15-116

Henk MW Verheul, Mark A van de Wiel + Show 3 more

Open Access

https://doi.org/10.1186/1471-2105-15-116

Copy DOI

Abstract

BackgroundComplex designs are common in (observational) clinical studies. Sequencing data for such studies are produced more and more often, implying challenges for the analysis, such as excess of zeros, presence of random effects and multi-parameter inference. Moreover, when sample sizes are small, inference is likely to be too liberal when, in a Bayesian setting, applying a non-appropriate prior or to lack power when not carefully borrowing information across features.ResultsWe show on microRNA sequencing data from a clinical cancer study how our software ShrinkBayes tackles the aforementioned challenges. In addition, we illustrate its comparatively good performance on multi-parameter inference for groups using a data-based simulation. Finally, in the small sample size setting, we demonstrate its high power and improved FDR estimation by use of Gaussian mixture priors that include a point mass.ConclusionShrinkBayes is a versatile software package for the analysis of count-based sequencing data, which is particularly useful for studies with small sample sizes or complex designs.

Highlights

Complex designs are common in clinical studies
Since the prior may be crucial for inference in a multiple testing setting, we extended the class of admissible priors to nonparametric and parametric mixture priors [2]
Priors To study which of the priors performs best in terms of FDR estimation and power, we compared them on simulated data sets, including those in [1]

Summary

Introduction

Complex designs are common in (observational) clinical studies Sequencing data for such studies are produced more and more often, implying challenges for the analysis, such as excess of zeros, presence of random effects and multi-parameter inference. Following the surge of count-based sequencing data, a plethora of software packages for differential expression analysis of such data has emerged [1]. Many of these methods are limited in use due to restrictions on the study design, the model and inference like a) 2- or Kgroup comparisons only; b) no random effects; c) no explicit solution for excess of zeros and d) no multiparameter inference. We extend the class of admitted priors with mixtures of a multivariate point mass and a Gaussian product density to allow for powerful multi-parameter inference

Results

Discussion

Conclusion