Identify differentially covered genes only between two samples

Question

I have a question about finding differentially covered regions (coverage represents methylation level which goes from 0 to several thousands). I'm using enrichment based method which can be summarized with coverage per gene:

data <- matrix(sample(80), 20)
# Genes in rows
rownames(data) <- letters[1:20]
colnames(data) <- c("group_A_tr1", "group_A_tr2", "group_B_tr1", "group_B_tr2")

In data matrix each row represents a gene and each column represents a sample. There are two sample groups (A and B) with two technical replicates per each group. Problem is that we do not have any biological replicates.

My goal is to identify genes that are differentially methylated between two groups. I know that limma, edgeR, DESeq2 can be used in analysis like this, however I don't have enough samples. Basically I'll need to compare only two columns (after averaging technical replicates).

What method would be appropriate to work with data like this? Is it possible to treat technical replicates as biological ones?

Answer 1

Not really an answer but an extended comment... and most likely something you don't like to hear

I guess by technical replicates, it means taking the same biological sample and making 2 methylation libraries. If this is used as replicates in deseq2 or edger, the variance you are estimating is the technical variation that comes with preparing the library, bisulphite conversion etc..

Often the biological variance will be much larger. That is if you take 2 independent samples of group A, their variation in methylation is much larger than the technical variation you see from handling the samples. If the technical replicates are treated as biological replicates and differential analysis is performed, it will be seriously misleading.

I think both edgeR and DESeq2 doesn't estimate dispersion with 1 v 1 (which is right).

I would suggest maybe the following:

1. Run a fisher.test at every gene (methylated / unmethylated), get the odds ratio and p-value, rank this, but explain that the p-value is only for ranking, and will underestimate the variation.

2. A bit more refined, look for other studies that might be close to yours, get an estimate of the dispersion in their study and use a negative binomial distribution to again rank genes that show the biggest difference

Answer 2

You may run edgeR for methylation analysis without replicates (https://f1000research.com/articles/6-2055). But I also recommend you to have a look at the R DSS package. It has a smoothing step which allows to modelize the biological variability by taking information from neighboring sites.