Please Help - SNP Calling in relation to Consensus Sequence Quality

[lg36]

Dear All,

I've posted on this before and attracted no interest at all. So I'd be very grateful to get your opinions. I'm relatively new to NGS sequence analysis so apologies for any naive assumptions.

As I understand (perhaps misunderstand) if a SNP is deemed to be present (after all the applied SNP filters, scores etc) then (most of the time) it will appear in the consensus sequence???

Is it therefore right to think if the consensus sequence is of a 'higher quality', the SNP's too will be of a 'higher quality'??

I'm defining quality of the consensus by the ability to reflect the presence or absence of certain oligonucleotides that we know to be there from PCR studies in the lab.

What I'm doing at the moment is trying to set the parameters correctly (paired end width, etc) in order to reach a consensus sequence that is correct according to what we know in the lab. Then I was planning on calling the SNPs based on the mapped reads we used for the best consensus sequence. Is there any validity at all in this approach, or are the two things SNP calling and consensus sequence production too 'unlinked' to make it invalid????

Many thanks in advance, apologies for any misconceptions

lg36

[pag]

Most base callers either call "n" at SNP location or pick one or the other (combination of sharpest peak and highest peak as scaled to average intensity for that base). Very few use IUPAC ambiguity codes to convey the potential difference. Quality for those locations, assuming an unbiased amplification level should be approximately 1/2 of the surrounding bases (I believe it's linearly scaled at least). So, if scripts don't already exist, you should be able to compare quality from one base to the neighboring two bases and see if the qual is significantly lower. Then you look at the trace itself (if available) to see if this is a potential SNP.

It's more difficult if it's a "chimera" or extended polymorphism or if there's a "frameshift."

I'd be interested to know if there are any chimera/SNP deconvolver programs or scripts out there for use that I can supply the reference sequence and it can let me know the alternate read. I was staring at at two overlapping reads that were significantly out-of-phase with each other yesterday and made about 50 base calls manually over an hour or two. Hardly the model of efficiency

[lg36]

Thanks pag for your help. So, is it true to say if I can replicate what we know to be the reality of the consensus sequence (from our lab findings), then the SNPs that are output from the same mapped reads will be more reliable?