P6C4 has higher error rate than P5C3?

[ymc]

Dear all,

I tried to compare the P6C4 and P5C3 chemistries.

P6C4 E coli K12 raw data from:

https://github.com/PacificBiosciences/DevNet/wiki/Datasets

P5C3 E coli K12 raw data from: (referred by the PBcR site)

http://www.ncbi.nlm.nih.gov/sra/SRX533603

P6C4 P5C3
Total 780M 374M
Coverage 168x 80x
avgReadLen 11074 8682
stdevReadLen 7107 5394
longestRead 41390 28647

I ran them through the PBcR-quiver pipeline and then use QUAST to evaluate the results:

PBcR-quiver(P6C4) PBcR-quiver(P5C3)
#misassemblies 5 3
#local_misassemblies 2 2
duplication_ratio 1.003 1.0
bases_not_covered 93 5756 (reference has 4641652bp)
#mismatches 4 2
#indels 27 3 (indels is the number of indels not number of indels in basepairs)
#genes 4513+3p 4508+2p (total 4516 genes, p means partial)

We can see that the coverage is a lot better most likely due to longer read length and higher throughput. But the errors also increased significantly even though the throughput doubled. What do you think was the cause? Different labs perhaps? What's your experience with the two chemistries?

[rhall]

Are the assemblies complete? i.e. circular genomes? The Quast metrics are not comparable unless both assemblies are complete, and the genomes have been correctly circularized.
P6C4 is generally higher accuracy, and higher consensus accuracy at lower coverage, but it is possible to overload a P6C4 cell degrading the data quality. P6C4 has been shown to perform better in all metrics than P5C3, witch is no longer available.

[ymc]

Thank you for your reply. What do you mean by a complete assembly?

I see that my assembly has only one contig and with no N's. I checked it with dotplot and didn't see duplication near the ends, so I didn't circularize it using the method suggested by PacBio. Is this a complete assembly?

[rhall]

If you see no duplication at the ends the assemblies did not result in circular genomes. Did both P5 and P6 generate single contigs? The discrepancy in the number of bases not covered makes me think that the P5 assembly is not complete.

[ymc]

If you see no duplication at the ends the assemblies did not result in circular genomes. Did both P5 and P6 generate single contigs? The discrepancy in the number of bases not covered makes me think that the P5 assembly is not complete.

The P5C3 assembly results in one single contig of 4636871bp and I didn't see duplications in dotplot.

The P6C4 assembly results in two contigs. One is 4656910bp and one is 36612bp. The smaller one has duplication at the ends but the long one is normal. Can the small one be a plasmid? I generated the above stat by taking out the smaller contig.