deltaTE: Detection of Translationally Regulated Genes by Integrative Analysis of Ribo-seq and RNA-seq Data

Abstract

Ribosome profiling quantifies the genome-wide ribosome occupancy of transcripts. With the integration of matched RNA sequencing data, the translation efficiency (TE) of genes can be calculated to reveal translational regulation. This layer of gene-expression regulation is otherwise difficult to assess on a global scale and generally not well understood in the context of human disease. Current statistical methods to calculate differences in TE have low accuracy, cannot accommodate complex experimental designs or confounding factors, and do not categorize genes into buffered, intensified, or exclusively translationally regulated genes. This article outlines a method [referred to as deltaTE (ΔTE), standing for change in TE] to identify translationally regulated genes, which addresses the shortcomings of previous methods. In an extensive benchmarking analysis, ΔTE outperforms all methods tested. Furthermore, applying ΔTE on data from human primary cells allows detection of substantially more translationally regulated genes, providing a clearer understanding of translational regulation in pathogenic processes. In this article, we describe protocols for data preparation, normalization, analysis, and visualization, starting from raw sequencing files. © 2019 The Authors. Basic Protocol: One-step detection and classification of differential translation efficiency genes using DTEG.R Alternate Protocol: Step-wise detection and classification of differential translation efficiency genes using R Support Protocol: Workflow from raw data to read counts.

Keywords: RNA-seq; Ribo-seq; deltaTE; translation efficiency; translational regulation.

Figure 1

Transcriptional and translational regulation (A). Genome‐wide quantification of mRNA counts and ribosome‐protected mRNA fragments (RPFs) using RNA sequencing (RNA‐seq) and ribosome profiling (Ribo‐seq), respectively. Lines are not drawn to scale. In a hypothetical study with two conditions, control and treatment, (B) a gene with change in mRNA counts and RPFs at the same rate is a differentially transcribed gene (DTG) and, (C) a gene with change in RPFs independent of change in mRNA counts, which leads to a change in translation efficiency, is defined as a differential translation efficiency gene (DTEG). TE = translation efficiency = RPF/mRNA. (D‐E) Classification of genes based on fold changes of RPF, mRNA, and TE. (D) A gene could be either/both DTG and/or DTEG, and based on the direction of change would fall into one of the eight gene‐regulatory possibilities (sig: significant, n.s.: not significant). Translationally forwarded genes are DTGs that have a significant change in mRNA and RPF at the same rate, with no significant change in TE. Conversely, translationally exclusive genes are DTEGs that have a significant change in RPF, with no change in mRNA leading to a significant change in TE. Several genes are both DTGs and DTEGs, and their regulatory class is determined based on a combination of the relative direction of change between transcription and translation efficiency. Specifically, translationally buffered genes have a significant change in TE that counteracts the change in RNA; hence, buffering the effect of transcription. Translationally intensified genes have a significant change in TE that acts with the effect of transcription. In all cases, the change in RNA can be either positive or negative, and where buffering or intensifying takes place, the direction of change is taken into account. For example, a gene that exhibits an increase in transcription and an increase in translation efficiency is classified as intensified, while a gene that exhibits an increase in transcription but a decrease in translational efficiency is classified as buffered. (E) Simulated data showing fold changes for each gene in RNA‐seq and Ribo‐seq data. Translationally forwarded genes (in blue), exclusive genes (in red), buffered genes (in purple), and intensified genes (in purple) are highlighted.

Figure 2

Translational regulation in sample data using DTEG.R script. Principal component analysis of (A) Ribo‐seq and (B) RNA‐seq datasets. (C) Scatter plot of log fold change values across both sequencing methodologies. Differentially transcribed genes (DTGs) and differential translation efficiency genes (DTEGs) are marked. (D‐G) Gene profiles of exemplars in each regulation class, translationally forwarded (D), exclusive (E), buffered (F), and intensified (G).

Figure 3

Quality check of Ribo‐seq data using Ribo‐TISH. The tool RiboTISH provides several visualizations to investigate the data quality of Ribo‐seq. First, it includes the length distribution for the Ribo‐seq reads as a histogram. As the length of ribosome‐protected mRNA fragment (RPF) is expected to be around 29 base pairs, the length distribution of the sequenced reads is used as a quality measure. Second, the 3‐nucleotide periodicity of the RPFs mapped on all known protein‐coding genes is shown for each read length. As shown, in these data, we have a high (93%) percentage of reads in Frame 1 with the predominant read length (29 bp). This is shown using a histogram of read coverage in the three frames, a barplot of the number of RPFs in each position around the START codon and STOP codon, and lastly a density plot for read coverage on the coding sequence across all genes.

Figure 4

Benchmarking of published tools to detect differential translation efficiency genes (DTEGs). Simulation datasets (A) derived from Oertlin et al. (2019), (B) derived from Xiao et al. (2016), and (C) generated using the Polyester package to introduce batch effects were used. All three simulations show that ΔTE outperforms all other published methods. Comparisons are made using all the DTEGs as the true set. Since Anota2Seq has two different functions for obtaining exclusive and buffered genes, the results are combined prior to comparison. Riborex is omitted in simulated datasets without batch effects (A, B), since it is equivalent to the ΔTE approach in these cases. The ratio method is based on quantifying the ratio of DESeq2 fold changes for mRNA counts and RPF. The overlap method identifies DTEGs as genes which have either significantly changing mRNA counts or RPFs but not both. (D) Analysis on published data showed inability of previous tools to reliably identify DTEGs.