Riborex: fast and flexible identification of differential translation from Ribo-seq data

Abstract

Motivation: Global analysis of translation regulation has recently been enabled by the development of Ribosome Profiling, or Ribo-seq, technology. This approach provides maps of ribosome activity for each expressed gene in a given biological sample. Measurements of translation efficiency are generated when Ribo-seq data is analyzed in combination with matched RNA-seq gene expression profiles. Existing computational methods for identifying genes with differential translation across samples are based on sound principles, but require users to choose between accuracy and speed.

Results: We present Riborex, a computational tool for mapping genome-wide differences in translation efficiency. Riborex shares a similar mathematical structure with existing methods, but has a simplified implementation. Riborex directly leverages established RNA-seq analysis frameworks for all parameter estimation, providing users with a choice among robust engines for these computations. The result is a method that is dramatically faster than available methods without sacrificing accuracy.

Availability and implementation: https://github.com/smithlabcode/riborex.

Contact: andrewds@usc.edu.

Supplementary information: Supplementary data are available at Bioinformatics online.

Fig. 1

Accuracy identifying differentially translated genes in simulated data (see supp. info.) comparing Riborex (using DESeq2, edgeR and Voom (Law et al., 2014)), Xtail, RiboDiff and Babel. Values are averages of 100 simulations and error bars indicate standard deviation of F scores. (A) and (B) Results based on rat heart data (Schafer et al., 2015) with 1% and 10% implanted true differentially translated genes, respectively, and simulated 4-fold change in translation efficiency. (C) and (D) Results from simulations based on mouse HuR data (Diaz-Muñoz et al., 2015)