Support for PacBio reads

[harrismia]

Hi Jody,

   It would be great if TB Profiler had support for PacBio in addition to Illumina and MinION.  We are starting to do more PacBio sequencing and have a number of isolates with both PacBio and Illumina reads.  It would be great to be able to use TB Profiler on this data. 

 TB Profiler is a fantastic tool.  Thanks for all of your efforts!

Michael

[jodyphelan]

Hi @harrismia ,

Sure I can look into this. Is there any publicly available data that I could use to optimise the variant calling pipeline?

[harrismia]

Hi @jodyphelan ,

Thanks a lot! The following BioProjects have some publicly available data: PRJNA719670, PRJNA480888, PRJNA436997 and PRJNA421446. PRJEB8783, PRJEB31443, PRJEB27802 and PRJNA598991 .

[jodyphelan]

Thanks, I've had a look at these data and performed a small analysis to optimise the variant calling parameters - namely the frequency at which variants are called. As pacbio is fairly noisy compared to illumina a minimum altrnate fraction is required to minimise the false positive variants. Using Illumina as the gold standard I looked at pairs of Illumina and pacbio to characterise the performance at several cutoffs.

ROC curve

By looking at the ROC curve it looks like 0.6 is a decent tradeoff between False positive and True positive rate.

I'll add in the code to enable --platform pacbio over the next few days and make 0.6 the default cutoff