Allele-specific expression in interspecific hybrids

Question

[Sandy]

Hi all,

I am very new to allele specific expression and I hope someone can help me.

I've been reading a lot in this forum about analyzing allele-specific expression, instead of giving me answers it made me raise a lot of questions instead. Before going to my question, let me give you an overview of my experiment.

So I have two grass species, Species A and Species B. Both species are in the same genus and both are produced seeds through selfing. We crossed Species A (maternal) with Species (paternal). Unfortunately, due to low hybridization success, we didn't get viable seeds from its reciprocal cross. The resulting hybrid, i.e. AB, along with its parental species A and B, we did RNA sequencing. Due to limited budget, we do not have DNA-Seq for any of them.

So I mapped the rnaseq reads of species A to species A reference genome and species B to species B reference genome using STAR and quantified them using featurecounts. Important to note that I used the exact accession numbers of species A and species B with the accession number that is used to create the reference genomes of both species. For AB hybrid, I mapped it to the concatenated reference genome of species A and species B. With this, I know exactly from which parent each gene came from. For this purpose, I want to use the single-copy ortholog between species for allele-specific expression instead of the heterozygote SNV sites.

So my questions are:

1. Is this a fair approach? Assuming that the single-copy ortholog between species paired up during meiosis which makes it in a way an allele of a particular gene or ortholog group (?).
2. When identifying which ortholog group exhibits allelic imbalance using a binomial test, should I do it using RAW counts or NORMALIZED counts? If the latter, what normalization method is appropriate?

Thank you very much for your expert opinions about this topic.

Best,
Sandy

Answer 1

[colindaven]

This is a very tricky topic. I'm not convinced that most analysts do this correctly.

I can't say much about cross species allele specific expression. Most ASE work has revolved around one species to my knowledge.

The ideal approach would be to have a haplotype resolved assembly - ideally complete - and then map short RNA reads to both (perfectly annotated) haplotypes. A more typical approach has been to add variation from both samples (SNPs) into the common reference genome - since SNPs cause allelic bias in read mapping, and then map read samples to those pseudoreference(s).

Tricky - yes. Doable - maybe.

I liked the approach of this tool very much when I did some work on this some time ago.

https://github.com/secastel/phaser

For your example I think you need to read the literature deeply to find comparable problems and follow their methods.

[Sandy]

I agree. Essentially, what we would like to know if the genome of species A or B is favorably expressed over the other. I will look into the link you suggested.

[colindaven]

Possibly the correct way to do this - though certainly not easy is:

• create a pangenome out of the two species using minigraph-cactus
• use vg giraffe or the vg rna mapper (mpmap)? to align rna reads to the pangenome
• write out the read counts continue from there.
• or project reads back to one/both of the references so you have access to gene annotation again.
• continue as usual

Pangenomics is a difficult and immature topic in itself though.

[Sandy]

This good be another approach. I will also consult my colleague here who is working on pangenomes for this Genus. But do you think the concatenated genome approach is a lackluster approach?