SNP calling on low coverage NGS data

[JdeBruin]

Hi,

I am new to SNP calling and would appreciate some input. I want to try and call SNP's for a specific organisms. The hurdles I have is, I do not have a reference genome to map against and I have a very low coverage 4x per organisms.

Any suggestions would be appreciated.

[TiborNagy]

The problem is the following: If you do not have reference genome, you need a de-novo assembly. You can find heterozygous positions in de-novo (which can be called SNPs), but in you case only two reads support this hipothesis.
So, use this sequences to create a reference and sequence other samples to find SNPs or make a higher coverage sequencing from the initial sample.

[SNPsaurus]

4X read depth can be even more misleading, since at heterozygous loci you will sometimes get 2 reads per allele, but also 3 reads to one allele vs 1 read for the other, or even 4 reads to one allele and no reads for the other. On the other hand, any allele supported by a few reads is probably real; you just can't fully genotype the sample but you can identify SNPs.

Genotyping by sequencing was developed to solve this problem. By sequencing only 10,000-100,000 loci across a genome, the reads attain good depth with a small amount of sequencing. We do genotyping by sequencing, as do other companies and academic cores. You don't need a reference for the informatics, but rather the reads can just be compared from sample to sample using a core set of reads representing the loci as a reference.

[swbarnes2]

You are in trouble. 4x is not enough for de novo assembly, and it is not really enough for calling homozygous SNPs, let alone heterozygous ones.

If these are all the same species, I'd combine all the reads together, and do de novo assembly on that. Then align each sample's reads to that assembly, call SNPs on those alignments, and realize that you are going to miss most of the heterozygous SNPs, and many homozygous ones too.

[Naibin Duan]

4X coveragedata without reference is almost impossible to call

In my opinion,4X coverage reseq-data without any reference is almost impossible to call SNP.
Best.
Naibin