Differential Expression Analyses between closely related species

Question

[whw84]

Hi all,

The project I am currently working on involves five closely related cricket species in the same genus. I have 8 (neutral) replicates of each species and the project was conceived with interspecific DE analyses in mind.

So far, I have made 5 de novo transcriptomes (one for each species) and estimated transcript abundance with kallisto. I notice now that most of the software for making transcript and gene expression matrices (for example $TRINITY_HOME/util/abundance_estimates_to_matrix.pl) require a single reference file. This makes sense if you are comparing different treatment conditions or populations within a species, but how do you proceed when you have multiple reference files? If I were to use a single reference file for all species, surely that would introduce significant bias towards the species used as the reference?

I hope I've provided enough information, but please let me know if you need more!

If anyone has any insight here, it would be greatly appreciated. In the mean time, I'll keep digging for answers.

Thank you.

Answer 1

[Mensur Dlakic]

This may not be exactly on point as I don't have enough information about your final goals. I am assuming you want to merge all the results in the end and compare expression profiles.

It doesn't matter if you reference files are separate, as long as they have identical annotations for identical genes. This is to say that you can perform each analysis with their own reference files, and then bring them into the same point of reference when DGE analysis is done. Assuming your gene XYZ is called the same way in all the species and you end up with 5 tables with gene names and their log2FC values, it is easy to join them afterwards.

Answer 2

[Lada]

Hi,

I don't know if you have resolved your problem but I am also a beginner in the interspecies transcriptomic and this is what I got so far:

1. I made de novo assemblies using Trinity for each species (I don't have reference genomes !) and I mapped my reads to estimate the expression level
2. Using a Transdecoder I generated the coding sequence datasets (PEP/FASTA).
3. I run Orthofinder to detect single-copy orthologues between each species pair (as previously said you have to be sure to compare same genes)

This is where I stopped because now I have an excel table of orthologues and I have to figure out how to assign my read counts to each gene for each species. For the DE analysis, I found this method that was published last year which accounts for some normalisations that are needed for the cross-species comparisons: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6441199/

I hope it helps and let me know how did it go for you.

Lada