SNP calling and no reference genome

[atma_weapon]

Hello, We have sequenced (RNA sequencing) three different strain of the same tree species (living in three different geographic locations) with the intention of discovering SNPs between the three strains.
This species has no reference genome, the company that performed the sequencing performed a transcriptome assembly for each of the three sequencing, and also provided a SNP calling for each of them.
I have two questions, if you can help:

- What does it mean to perform a SNP calling inside the same individual? the same gene producing transcripts with different nucleotides in same position?

- How can I find SNPs between the three strains? should I perform a regular SNP calling of each sequencing against the assembled transcriptome of the others strains?

thank you very much

[dpryan]

Yes, within-individual SNP calling would look for heterozygous (or whatever the equivalent term is for higher ploidy plants is) variants. Of course, the problem is that any recent paralog will look like a SNP...but there's nothing that can be done about that. Since you paid a company to do this, they should address any questions you have there.

For the between-sample calling, you might look to the literature. This is a recent paper on reference-free SNP calling using De Bruijn graphs.

[atma_weapon]

thank you, I have noticed that the company assembled the 3 sample into a consensus reference transcriptome, and later they provided a SNP call for each sample comparing it against this reference, is it a correct approach? or should we do a between-sample calling as you suggested?

Yes, within-individual SNP calling would look for heterozygous (or whatever the equivalent term is for higher ploidy plants is) variants. Of course, the problem is that any recent paralog will look like a SNP...but there's nothing that can be done about that. Since you paid a company to do this, they should address any questions you have there.

For the between-sample calling, you might look to the literature. This is a recent paper on reference-free SNP calling using De Bruijn graphs.

[dpryan]

What they did sounds reasonable to me at least. I'd try to use those results first.

[akp1d12]

kSNP is a program that works without a reference