Calculating gene expression dominance coefficient via DESeq2
1
0
Entering edit mode
@ebf5ca55
Last seen 11 days ago
Sweden

I am interested in calculating dominance coefficients of gene expression. For this we have RNAseq data consisting of two homozygous lines (i.e. Genotype AA and BB, respectively) and of the heterozygous crosses between them (i.e. Genotype AB and BA, respectively). We have data for both sexes in each group and I am using DESeq2 to analyse the data using the following design design = ~Sex+Genotype. For now I will ignore the Sex component as it is not relevant for my question.

In a first step I am identifying genes that show different expression in AA versus BB and I can easily do so using the suggested DE steps (thanks to the great DESeq2 manual/tutorial :)).

In a second step I am interested whether there are any heterozygotes (i.e. AB) that show non-additive gene expression, in other words are there cases where AB gene expression is different from the mean gene expression between AA and BB. I can calculated this via the dominance coefficient as 1-(AA-AB)/(AA-BB) which would be 1 if A gene expression is dominant over B, 0.5 if they are additive and 0 if A is recessive to B. This overall seems to work well and I get a nice distribution around 0.5 meaning that the majority of genes show additive gene expression.

Is there a way to use DESeq2 to test if a gene shows significant dominance or recessivity? I.e. a way to do DE between AB and the mid-gene expression between AA and BB?

My solution so far is to manually create all homozygote mid-gene expressions (that I call AABB) where I take the mean between any AA and BB pair. I.e. I have 3 ´AA´ samples and 2 BB samples, I would create 6 AA1BB1, AA1BB2, AA2,BB1, AA2BB2, AA3BB1, AA3BB2. I do the same thing for the heterozygotes where I similarly pair up AB and BA to get ABBA and then do the DE contrast AABB versus ABBA. This sort of works as it seems to gives me reasonable results (as far as I can judge) but I am not happy with my solution because I am concerned that this is not the best way or potentially a really bad way of doing this as it may give DESeq2 false power due to potential pseudo replication?

Any help would be greatly appreciated

RNASeq DESeq2 • 181 views
0
Entering edit mode
@mikelove
Last seen 1 day ago
United States

One note is that DESeq2 and other Bioc RNA-seq tools are actually analyzing log of RNA abundance, whereas I believe you want to compare absolute abundance, right? You want to know if AB (scaled count) is half of AA and BB (scaled count), as the allelic dosage affects abundance not log abundance.

I wouldn't really ignore this detail either, it's kind of non-ignorable for this kind of test.

The TreCASE software offers tests of dominance, see here:

https://rdrr.io/cran/rxSeq/man/trecase.A.out.html