I'd like to know whether it is technically possible to improve rare mutant detection frequency by increasing the depth of the sequencing to 500X for WES using illumina library preparation kits and Hiseq. Can the bioinformatics pipelines used for WES samples with 50X and 100X coverage handle data also with 500X coverage?
Thanks in advance.
Most analysis pipeline for WES and variant calling can handle 500x or even 1000x or more (provided that storage, CPU, network, etc, are available). Having said that, an important consideration to keep in mind is artifacts due to sequencing errors can be introduced. These might look like an SNV with low variant allele frequency, so special care is needed
We've been using the standard tools, BWA, GATK etc to look at Fluidigm amplicon sequencing and deep targeted sequencing. A good run gets several thousand fold coverage and the standard tools work just fine. As our colleagues state above, you have to be careful of PCR artifacts but you can mostly filter them out.
What I'm still not totally clear on is the actual resolution of deep sequencing like this. At a certain point, the background error rate starts to swallow up the low prevalence mutations. We can certainly see mutations at 1% prevalence but going below that the evidence becomes a bit shaky.
The second issue is how to validate a mutation you detected at 0.1%, for example. Sanger sequencing won't do the job, so you'll have to plan ahead.
I want to reiterate what Scott Newman already mentioned.
I was involved in a similar study where the goal was to detect single nucleotide mutations in plasmids of a bacterial population (in the setting of an industrial fermentation process where it was necessary to verify that no mutations occured during a certain period of time).
The bottom line is quite simple; you can only detect mutations at frequencies which are higher than the error rate of the sequencer.
Let's assume an error rate of 1% (i.e. in average every 100th nucleotide gets called wrong) and a coverage of 500x at the loci of your interest. Assuming your wildtype sequence at this location is "A", you can expect 495 reads saying "A" and the rest being mistakes ("G","C","T" or indel) [this of course being simplistic, because errors won't be Independent and equal]. A possible rare mutation occuring at a frequency of 1% will also have an expectation value of around 5 reads at this position. If you further increase your coverage, let's say to 5000x, nothing changes. You expect 4950 reads being "A", and 50 reads being either errors or your rare mutation, indistinguishable from each other.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Bioinformatics