Ignore CCS reads - a correct assumption?

[ritzriya]

Hi all,

As we know that Pacbio provides CLR as well as CCS reads for error correction of the long reads for any dataset. I want to know, for Variation Detection, is it okay to ignore the CCS reads completely and simply use the .bas.h5 file as given the raw read to find variants across a reference genome?

Any help regarding this would be great..

[llee]

The bas.h5 file contains both single pass CLR and multipass CCS reads. The PacBio pbh5 R API (https://github.com/PacificBiosciences/R-pbh5) or the bash5tools.py utility (https://github.com/PacificBiosciences/pbh5tools) can be used to access and extract both read types from the bas.h5 file. Their usefulness in variant detection depends on a variety of other factors, including frequency of variation, insert/amplicon length, CLR vs. CCS depth of coverage, SNP phasing requirements, and method of SNP detection. Assuming you are examining haploid/diploid SNP calls, here are two different scenarios that would argue for one or the other:

1. 5kb shear of genomic DNA. In this case, the insert sizes are generally too long for CCS read generation and variant detection with aligned CLR using GATK with base QV recalibration is recommended.
2. 500bp amplicons. The shorter insert sizes allow for sufficient CCS yield, and variant detection with aligned CCS using GATK with or without base calibration is recommended.

I would suggest you examine the readlengths and yield of your CLR vs. CCS reads and evaluate against your variant detection needs to choose the best read type for your application. It is also possible to call variants with both read types in parallel and evaluate the quantity and concordance of SNP calls.

[ritzriya]

Thank you llee. I will try calling SNPs from both the read types and compare the results.