Trinity combine assembly

Question

[kishorssf91]

Hi, Researchers typically perform differential expression analysis for the same species in different conditions with biological replicates. For doing this study trinity has a pipeline almost from top to bottom. Usually, it will make a combine assembly. But i am having few confusions. Is it reasonable to use simillar (not same) species as replicates from the same ecological niche and then do a DE analysis with simillar species from another ecological niche using trinity pipeline? Then how can i get particular gene or particular transcript for particular species as those are being assembled as common assembly? Do they assemble all the common gene as a single gene and varrying transcripts as different isoforms for a genes?

[Dunois]

You ending up with chimeric assemblies that wouldn't be comparable between the two conditions would be my guess.

I suppose what you're basically suggesting is metatranscriptomic differential expression analysis. Maybe take a look at this paper: https://academic.oup.com/bioinformatics/article/37/Supplement_1/i34/6319701. I think this has been limited to prokaryotes thus far, however.

[kishorssf91]

Thanks for your response. I am actually using 6 fish species from two habitat type from a same family. Three of them from marine water and another three of them from freshwater. And was checking for certain expressed features. I found trinity as a pipeline can do a DE study from top to bottom, hence i choose it. I dont have any replicate rather i considered hatitat as a type and 3 similar species as 3 replicate for one type. Can i go through with this hypothesis using trinity combine assembly pipeline? Or there is any other way to handle this situation. If i get some suggestions, i will be really grateful to you.

[Dunois]

Sorry for replying so late to this.

I don't think differential expression analysis would actually be appropriate here, especially if you're using six unique fish species (i.e., A, B, C, D, E, and F). I suppose the most you could do is assemble each transcriptome individually, and then follow up with an elimination strategy of sorts. You'd basically have to perform an all-vs-all search (e.g., using something like OrthoFinder), and retain only the transcripts/genes unique to each species and/or unique to species from the same habitat.

[kishorssf91]

Thank you so much. So far i have used the Orthofinder and tried to count the expression on level of orthogroup. But yet to get the desired result. Can i run on only annotated and same genes with similar length from 6 species with a manual normalization and read counts?

[Dunois]

That isn't going to work. Orthogroups are hierarchical, and you'll end up trying to quantify expression levels for entire families of genes. I suppose you used Trinity to construct the assembly? In that case, you'll also have redundant isoforms to deal with. Did you run OrthoFinder on transcriptomes that are non-redundant? (Trinity should have scripts that'll yield you assembly subsets with one "representative" isoform per group of related transcripts.)

OrthoFinder should have created a pairwise orthologs directory somewhere. I'd recommend you pull out all pairwise orthologs to the gene of interest (given you know its sequence in at least one of your six species), and then try the differential expression analysis (if you really insist on that) using one such transcript as the representative for each group of 3 fish species. I do think you'll see that the expression levels will be "different" but I think that'll be more because of experimental noise than actual differences in expression.

Is it reasonable to use simillar (not same) species as replicates from the same ecological niche and then do a DE analysis with simillar species from another ecological niche using trinity pipeline?

IMHO the strategy is flawed and you need to go back to the drawing board.