Why Illumina reads may have uneven coverage?

ojy · 12-06-2011, 07:01 PM

Why Illumina reads may have uneven coverage of the genome? Does it depend on GC content of the genome, and if so, then why and how exactly?
I would appreciate any explanations or references (preferably so that a non-biology background person can understand)

stefanoberri · 12-07-2011, 01:41 AM

Hi.

We observed that read coverage is affected by GC content. Each sample, however, have a different bias and has to be corrected independently. We don't know exaclty why, and we suspect it is somehow linked to DNA quality and extraction method.

We briefly mention CG bias in a recent paper. Read supplementary material for a few examples. Other articles (cited) mention GC content bias.

pmiguel · 12-07-2011, 07:03 AM

The evidence from full genome sequencing experiments is that the majority of bias comes from enrichment PCR. If you search, you can find references to this.

We have seen this to be the case first hand. A standard TruSeq shotgun library of Deinococcus radiodurans (a ~70% GC genome) gave us very uneven coverage. Mapped back to the reference, the reads would vary from 100x coverage down to zero. However running the same library directly, without intervening enrichment PCR, resulted in quite even coverage.

The down-side was that cluster densities we got from the non-amplified library were much lower than expected from qPCR titration. Subsequent documentation from Illumina leads me to believe that the TruSeq adapters have some modification that interferes with them binding to the flow cell. (Possibly LNAs? I don't know if LNAs are recalcitrant to NaOH denaturation.)

--
Phillip

KevinLam · 12-12-2011, 12:47 AM

out of curiousity, your illumina reads are from GAIIx? or HiSeq? AFAIK, differences in the sample prep resulted in a different GC coverage profile even within the same companies' platform.
anecdotal evidence only, no figures or anything.

TonyBrooks · 12-12-2011, 04:01 AM

Hi Phillip
I think Illumina have seen this problem with non-PCR'd libraries. They seem to think the problem arises with the Y adapter not dissociating properly in the NaOH denaturation step. Their suggested solution is to perform a one cycle PCR to completely linearise the libraries before qPCR and flowcell loading.
Also, the Kapa library amplification module is supposedly better at generating even coverage (when compared to the Phusion Taq in the Illumina kits), although we haven't tried it yet.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Experience of Illumina TruSeq Exome coverage?	leesi	Illumina/Solexa	2	03-06-2013 12:45 AM
Illumina Whole Exome coverage	kellogg	Illumina/Solexa	5	07-14-2011 10:18 AM
Variant calling for high-coverage Illumina data	dgmacarthur	Bioinformatics	19	04-08-2011 07:34 AM
MIRA output for Illumina mapping giving 100% coverage!	Kasycas	Bioinformatics	1	09-10-2010 04:45 AM
In Sequence: Illumina GA’s Deep Coverage Shown to Be Useful for Profiling, Discovery	Newsbot!	Illumina/Solexa	0	02-26-2008 03:20 PM

12-06-2011, 07:01 PM	#1
ojy Junior Member Location: California Join Date: Jul 2011 Posts: 5	Why Illumina reads may have uneven coverage? Why Illumina reads may have uneven coverage of the genome? Does it depend on GC content of the genome, and if so, then why and how exactly? I would appreciate any explanations or references (preferably so that a non-biology background person can understand)

12-12-2011, 12:47 AM	#4
KevinLam Senior Member Location: SEA Join Date: Nov 2009 Posts: 203	out of curiousity, your illumina reads are from GAIIx? or HiSeq? AFAIK, differences in the sample prep resulted in a different GC coverage profile even within the same companies' platform. anecdotal evidence only, no figures or anything. __________________ http://kevin-gattaca.blogspot.com/

12-07-2011, 01:41 AM	#2
stefanoberri Member Location: Cambridge area, UK Join Date: Jan 2010 Posts: 35	Hi. We observed that read coverage is affected by GC content. Each sample, however, have a different bias and has to be corrected independently. We don't know exaclty why, and we suspect it is somehow linked to DNA quality and extraction method. We briefly mention CG bias in a recent paper. Read supplementary material for a few examples. Other articles (cited) mention GC content bias.

12-07-2011, 07:03 AM	#3
pmiguel Senior Member Location: Purdue University, West Lafayette, Indiana Join Date: Aug 2008 Posts: 2,317	The evidence from full genome sequencing experiments is that the majority of bias comes from enrichment PCR. If you search, you can find references to this. We have seen this to be the case first hand. A standard TruSeq shotgun library of Deinococcus radiodurans (a ~70% GC genome) gave us very uneven coverage. Mapped back to the reference, the reads would vary from 100x coverage down to zero. However running the same library directly, without intervening enrichment PCR, resulted in quite even coverage. The down-side was that cluster densities we got from the non-amplified library were much lower than expected from qPCR titration. Subsequent documentation from Illumina leads me to believe that the TruSeq adapters have some modification that interferes with them binding to the flow cell. (Possibly LNAs? I don't know if LNAs are recalcitrant to NaOH denaturation.) -- Phillip

12-12-2011, 04:01 AM	#5
TonyBrooks Senior Member Location: London Join Date: Jun 2009 Posts: 298	Hi Phillip I think Illumina have seen this problem with non-PCR'd libraries. They seem to think the problem arises with the Y adapter not dissociating properly in the NaOH denaturation step. Their suggested solution is to perform a one cycle PCR to completely linearise the libraries before qPCR and flowcell loading. Also, the Kapa library amplification module is supposedly better at generating even coverage (when compared to the Phusion Taq in the Illumina kits), although we haven't tried it yet.