Quiver

[Gators]

I've downloaded this and started using it. I have to say, I am either doing something wrong or it's not nearly as useful as I was hoping. Hopefully I'm doing something wrong here.

This is what I have tried.

Took 4 SMRT cells of long read data. Isolated out reads that are >5000. Imported these >5K reads into SMRT Portal. Ran RS_Mapping_QVs to get the new cmp.h5 file.

Finally, ran Quiver and output the reads as fastq.

Was left with about 12X coverage of my organism of interest. Tried Celera to assemble, and it essentially failed...most of the reads aren't even used.

I've taken the same data, used PacBioToCA to correct with Illumina, and assembled using about the same coverage and got less than 100 contigs...

Ideas?

[jbingham]

Quiver is a consensus and variant calling algorithm rather than a direct replacement for pacBioToCA. The main difference between aligning to 5 kb reads and taking consensus vs running pacBioToCA is in the trimming and filtering. Then there's the coverage difference.

There are settings to P_ErrorCorrection that can use single pass reads to do what you're attempting. They're not public yet though you can ask your local FAS.

Even with the right settings, you may have a coverage issue depending on your genome size. How large is your genome?

Quiver is for now best used as post-processing: take your assembled contigs, align all the PacBio data against it using RS_Mapping_QVs, and then run Quiver on it to get a final consensus that can be Q50 with adequate coverage. Quiver adds a good 10-15 phred points to the final assembly accuracy.

[gconcepcion]

What is Q50 ?

Q50 implies that there is on average 1 error in a 100,000bp stretch when aligning a consensus sequences to a reference.

See the wikipedia entry on phred score here:

https://en.wikipedia.org/wiki/Phred_quality_score