In-dels Illumina vs. PacBio

slumpy42 · 10-12-2015, 07:31 AM

Hello,

I have a PacBio-generated closed complete bacteria genome. I am aligning Illumina reads of the same strain to it using breseq, a pipeline which uses bowtie2 for read mapping.

The alignment finds 100s of insertions, all of them are in homo-polymer runs (AAA...) of nucleotides.

2 questions:

1) Is it known that Illumina produces extra nucleotides in repetitive regions, or that PacBio is too short? Or do they both have In-del problems?

2) How should I deal with this? Just ignore all of them?

Thanks,

Brian Bushnell · 10-12-2015, 10:40 AM

Illumina does not usually have problems with indels, homopolymer or not. PacBio has a high indel rate in the raw reads and homopolymers are harder to correct, particularly if you don't have enough coverage. You might run quiver to attempt to correct these errors. There are also other programs for correcting assemblies via alignment (ideally using Illumina reads as they generally lack indel errors), but I don't know which is best.

dgscofield · 10-14-2015, 11:31 AM

We've noticed single-base deletions in PacBio assemblies, via alignment of Illumina reads to otherwise very nice Pacbio assemblies. These occur about every 10^3 to 10^4 bases, and we've also found them to be connected with even quite short homopolymers.

I wrote a little two-stage script to correct these given the BAM of illumina alignments, you can find out more at its repository (below). As Brian writes I'm sure there are other solutions as well.

https://github.com/douglasgscofield/PacBio-utilities

flxlex · 10-16-2015, 05:38 AM

Try Pilon from the Broad: http://www.broadinstitute.org/software/pilon/

slumpy42 · 10-17-2015, 07:48 AM

Thanks for all your replies and suggestions.

I am going through the PacBio training webinars and they mention that deletions are the most common form of error in the circular consensus reads, but don't give a rate.

I have a 3.2 Mb genome and roughly 5,000 single base insertions. All look like they are in homopolymers.

@Biran_Bushnell : I assembled with HGAP.3, includes quiver as the final step. Should I run it again with different settings?

@dgscofield : Thanks, I'll try your script and Pilon and send you the before and after stats, if that would be helpful for you.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Error correction on PacBio with illumina and 454	lucio89	Pacific Biosciences	0	10-03-2014 05:13 PM
Cross-platform Assembly (PacBio & Illumina)	bonifera	Bioinformatics	7	05-17-2013 09:20 AM
De novo assembly of PacBio with short Illumina data	boetsie	Pacific Biosciences	1	10-06-2012 01:35 PM
hybrid assembly for PacBio and Illumina data	laelaps	Bioinformatics	1	05-01-2012 06:35 AM
Hybrid assembly of PacBio and Illumina	allo	Bioinformatics	3	05-01-2012 06:27 AM

10-12-2015, 07:31 AM	#1
slumpy42 Junior Member Location: NYC USA Join Date: Apr 2014 Posts: 3	In-dels Illumina vs. PacBio Hello, I have a PacBio-generated closed complete bacteria genome. I am aligning Illumina reads of the same strain to it using breseq, a pipeline which uses bowtie2 for read mapping. The alignment finds 100s of insertions, all of them are in homo-polymer runs (AAA...) of nucleotides. 2 questions: 1) Is it known that Illumina produces extra nucleotides in repetitive regions, or that PacBio is too short? Or do they both have In-del problems? 2) How should I deal with this? Just ignore all of them? Thanks,

10-14-2015, 11:31 AM	#3
dgscofield Member Location: Uppsala, Sweden Join Date: Nov 2010 Posts: 27	We've noticed single-base deletions in PacBio assemblies, via alignment of Illumina reads to otherwise very nice Pacbio assemblies. These occur about every 10^3 to 10^4 bases, and we've also found them to be connected with even quite short homopolymers. I wrote a little two-stage script to correct these given the BAM of illumina alignments, you can find out more at its repository (below). As Brian writes I'm sure there are other solutions as well. https://github.com/douglasgscofield/PacBio-utilities Last edited by dgscofield; 10-14-2015 at 11:34 AM.

10-12-2015, 10:40 AM	#2
Brian Bushnell Super Moderator Location: Walnut Creek, CA Join Date: Jan 2014 Posts: 2,707	Illumina does not usually have problems with indels, homopolymer or not. PacBio has a high indel rate in the raw reads and homopolymers are harder to correct, particularly if you don't have enough coverage. You might run quiver to attempt to correct these errors. There are also other programs for correcting assemblies via alignment (ideally using Illumina reads as they generally lack indel errors), but I don't know which is best.

10-16-2015, 05:38 AM	#4
flxlex Moderator Location: Oslo, Norway Join Date: Nov 2008 Posts: 415	Try Pilon from the Broad: http://www.broadinstitute.org/software/pilon/

10-17-2015, 07:48 AM	#5
slumpy42 Junior Member Location: NYC USA Join Date: Apr 2014 Posts: 3	Thanks for all your replies and suggestions. I am going through the PacBio training webinars and they mention that deletions are the most common form of error in the circular consensus reads, but don't give a rate. I have a 3.2 Mb genome and roughly 5,000 single base insertions. All look like they are in homopolymers. @Biran_Bushnell : I assembled with HGAP.3, includes quiver as the final step. Should I run it again with different settings? @dgscofield : Thanks, I'll try your script and Pilon and send you the before and after stats, if that would be helpful for you.