6
$\begingroup$

PacBio is selling ~10x PacBio SEQUEL long reads as an upgrade to Illumina data for SV discovery.

In a clinical setting, the main requirements are proper sensitivity and specificity but also the processing of cohorts, at least families. This requires a genotyping step, such that it can be identified whether a given variant is shared by two or more individuals or whether it is not.

What are the tools of the trade for this task?

As having 50-60x PacBio reads is not an option from a economic point of view, one has to make do with 10x coverage.

$\endgroup$

2 Answers 2

2
$\begingroup$

There is an evaluation of PB Honey and Sniffles algorithms for low coverage PacBio datasets in this preprint and another evaluation is shown on this poster. Both reports agree that optimal is (surprisingly) combination of PB Honey and Sniffles.

Author of Sniffles have benchmarked Sniffles against PB Honey, where he shown that Sniffles performs significantly better. Take a look on this presentation (slide 15).

Another option is SMRT-SV, but I am not aware of any benchmarking.

$\endgroup$
4
  • $\begingroup$ SMRT-SV uses assembly-quality coverage of PB reads and sadly of low use. PB Honey and Sniffles are essentially useless for clinical germline studies as neither supports multi-sample calling and when comparing calls from both tools on patients where we also had Illumina X Ten genomes showed promise in PB data but severe problems with existing tools. $\endgroup$
    – Manuel
    Jun 1, 2017 at 19:57
  • $\begingroup$ I can imagine that if the SV calls would take into account all the family data in one go, the results would be more reliable. However is it really a problem to do variant calls individually and then compare / merge them? I would guess that it would be still better than Illumina... $\endgroup$ Jun 2, 2017 at 15:28
  • $\begingroup$ Has anyone successfully used either of these programs on MinION generated data? $\endgroup$ Jun 4, 2017 at 12:11
  • $\begingroup$ @MatthewBashton for this you should open a new question... $\endgroup$ Jun 4, 2017 at 12:41
1
$\begingroup$

I'm aware of the following (very few and suboptimal) options:

  • Sniffles -- sadly not very reliable in my experience, also no genotyping step or multi-sample support
  • PB Honey -- no genotyping step or multi-sample support
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.