How to work out coverage?

[James]

Hi everyone,

I want to work out what coverage/depth of sequencing we have on our last RNA-seq run. We are working on a small genome, so Ideally we would multiplex to bring the cost down and do more experiments. Before working out if we can multiplex I need to work out how much coverage we have and then how much we need.

Cheers, J

[malachig]

The issue of coverage in RNA-seq data (transcriptomes) is arguably more complicated than it is for whole genome shotgun sequencing or exome sequencing.

In sequencing the genome, you want to be reasonably sure that you are covering the majority of the genome with sufficient depth to genotype and identify mutations at each position. Since each portion of the genome is present in approximately equal amounts (i.e. the target space is approximately uniform in representation) the amount of sequencing needed to achieve a particular level of coverage can be predicted.

In transcriptomes, the situation is not as simply (IMHO). Although the size of the genome of the organism may be important and having a small genome may help help you there are other important factors to consider. For example, the dynamic range of gene expression levels. This varies from species to species and tissue to tissue. It is also highly dependent on library construction method. Is the library polyA+, if not how are ribosomal sequences removed? Was a library normalization method applied? etc. In some transcriptome libraries I have seen, a small number of highly expressed genes consumed a huge percentage of all reads. This was particularly pronounced in libraries created with riboMinus RNA compared to polyA+ RNA. Another issue is what you hope to get out of the data? You can use RNA-seq libraries to profile gene level expression, differential expression, identify mutations, RNA editing, gene fusions, alternative splicing, etc. Do you need to be sure to profile genes with very low expression levels (say down to 1 copy per cell or even less...)?

One strategy is to think about which of these analysis options you are going to pursue. Then think about which would require the most depth to satisfy. Since you have some example data, see how well this data performs at increasing levels of depth for this task. That is, make saturation curves for the metric of interest.

Several papers have described the relationship between library depth and output for several different applications of RNA-seq. The supplementary materials of the following paper describes some of these: ALEXA-seq (see Supplementary Text and Supplementary Figure 4)