Assembly vs. SNPs

[fishinvest]

Hey Guys,

My lab is working toward the goal of comparing genes from new strains against existing strains in a whole battery of tests (like dN/dS) which will require us to get orthologous groups, align them, and run the downstream tests.

My PI has pointed me towards several SNP calling pipelines (e.g. BWA-SamTools-Picard-GATK) with the idea that once we call the SNPs we can compare with the reference genome and make a putative set of chromosomes for the new strain.

This seems the wrong way to make genes. Should I be using something like Velvet to ASSEMBLE instead of using a pipeline to call SNPs?

We are working mostly with Illumina

Thanks!

[swbarnes2]

velvet only works if you have pretty high coverage. And velvet will not tell you if you have a mix of alleles.

Unless the reference you are aligning to is quite divergent from your samples, calling SNPs will work fine.

[fishinvest]

No worries about indels,etc?

So we can use the SNP calling pipeline to infer genes? genomes? What about indels, reversals, etc?

My coverage is actually quite high so that shouldn't be an issue

Thanks!

[swbarnes2]

"Infer genes"? No, a variant caller can't do that. mpileup can catch small indels, having paired end data really helps there. Velvet can't do that either.

If you have more complex variations, mpileup will tell you that there is a discrepancy, it won't be able to tell you exactly what it is. You'd have to investigate yourself (using something like velvet could help)

Velvet would likely assemble the sequence correctly in that case, if there were enough reads.

Also, what are you assembling?

Assembling PCR amplicons of genes, sure. Assembling a whole plant genome? Not at all the same thing. Much, much harder, and your genome will be in a hundred thousand pieces.