Hi all,

I have a question about edgeR Bisulfite sequencing and differential methylation analysis. I'm following the main manual 'differential analysis of sequence read count'. I have found some inconsistencies in library size preparation.

In the manual, the library sizes for each sample should be the average of the total read counts for the methylated and unmethylated libraries. But the code shows:

TotalLibSize <- y$samples$lib.size[Methylation=="Me"] + y$samples$lib.size[Methylation=="Un"]

In the next chapter in the same step and note 'as before' the code is:

TotalLibSizepr <- 0.5*ypr$samples$lib.size[Methylation=="Me"] + 0.5*ypr$samples$lib.size[Methylation=="Un"]

Could You explain to me the difference between both approaches? In other papers, there are also two different approaches.

Regards

There is actually no difference between the two approaches in terms of differential results, both settings leading to the same p-values. The only change is to the logCPM column of the topTags table. In our paper (https://f1000research.com/articles/6-2055/v2) we note that

the two library sizes for each sample should be equal. Otherwise, the library size values are arbitrary and any settings would lead to the same P-value.

I agree though that the inconsistency in the User's Guide is confusing. I will revise the Guide so that we use the average throughout.