PCR-Free Libraries & Bioinformatic PCR Duplicate detection?

aeonsim · 10-22-2013, 09:52 AM

Does any one have any thoughts or experience with regards to the Bioinformatics pipeline for Illumina PCR-Free libraries?

Especially with regards to the need to run Picards MarkDuplicates or some other PCR duplicate detection tool?

As we've started a large sequencing project using the PCR-Free libraries and having run Markduplicates over the initial test lane of sequence it's reporting a PCR-duplicate rate of between 0.004-0.006. Some of which it claims are Optical Duplicates and the rest (2.5x the Optical numbers) it claims are Library duplicates. However considering the kit & protocol was one of the PCR-Free ones I'm wondering if these are actual duplicates or misidentification by Markduplicates (potentially reads from repetitive regions).

Any thoughts would be greatly appreciated.

atcghelix · 10-22-2013, 03:44 PM

What are you sequencing? And are you sequencing paired-ends?

As I understand it tools find "PCR duplicates" by comparing mapping coordinates of reads. However, if in your library prep you have actual fragments that shear in the same place then the resulting reads will also map to the same location. So they'll look like PCR duplicates in that they map to the same spot on the reference, but they actually represent good, unique reads.

The probability of this happening will increase with smaller references (smaller genomes, RNA-seq, targeted enrichment sequencing). The probability should decrease greatly by using paired-end reads as well, as the fragments in the library then have to shear at the same location at each end of the fragment.

I don't know if there's a standard protocol, but I'd be inclined to filter out the optical duplicates and leave in the "PCR" duplicates in your case.

aeonsim · 10-22-2013, 04:33 PM

Data is for WGS Bos Taurus, ~3GB Genome, using 101bp PE or 109bp PE PCR-Free Libraries.

atcghelix · 10-23-2013, 02:36 PM

If you pull out the reads that have been marked as duplicates, do they look like they are highly repetitive?

And just to make sure, the reads that are showing up as duplicates are reads that map to the Bos taurus genome, yes?

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
PCR-free libraries	matth431	Illumina/Solexa	4	08-02-2013 06:35 AM
PCR-free NGS Libraries from 5 ng of Input DNA	swiftbiosci	Vendor Forum	5	06-28-2013 10:57 PM
High PCR duplicate with library - possible cause?	JustinNelligan	Epigenetics	0	03-11-2013 03:27 AM
High GC content and PCR duplicate	ttnguyen	Illumina/Solexa	0	03-07-2011 11:35 AM
duplicate PCR Single and Paired End	m_elena_bioinfo	Bioinformatics	12	03-03-2011 06:48 AM

10-22-2013, 09:52 AM	#1
aeonsim Member Location: Belgium Join Date: Jun 2011 Posts: 45	PCR-Free Libraries & Bioinformatic PCR Duplicate detection? Does any one have any thoughts or experience with regards to the Bioinformatics pipeline for Illumina PCR-Free libraries? Especially with regards to the need to run Picards MarkDuplicates or some other PCR duplicate detection tool? As we've started a large sequencing project using the PCR-Free libraries and having run Markduplicates over the initial test lane of sequence it's reporting a PCR-duplicate rate of between 0.004-0.006. Some of which it claims are Optical Duplicates and the rest (2.5x the Optical numbers) it claims are Library duplicates. However considering the kit & protocol was one of the PCR-Free ones I'm wondering if these are actual duplicates or misidentification by Markduplicates (potentially reads from repetitive regions). Any thoughts would be greatly appreciated.

10-22-2013, 03:44 PM	#2
atcghelix Member Location: CA Join Date: Jul 2013 Posts: 74	What are you sequencing? And are you sequencing paired-ends? As I understand it tools find "PCR duplicates" by comparing mapping coordinates of reads. However, if in your library prep you have actual fragments that shear in the same place then the resulting reads will also map to the same location. So they'll look like PCR duplicates in that they map to the same spot on the reference, but they actually represent good, unique reads. The probability of this happening will increase with smaller references (smaller genomes, RNA-seq, targeted enrichment sequencing). The probability should decrease greatly by using paired-end reads as well, as the fragments in the library then have to shear at the same location at each end of the fragment. I don't know if there's a standard protocol, but I'd be inclined to filter out the optical duplicates and leave in the "PCR" duplicates in your case.

10-22-2013, 04:33 PM	#3
aeonsim Member Location: Belgium Join Date: Jun 2011 Posts: 45	Data is for WGS Bos Taurus, ~3GB Genome, using 101bp PE or 109bp PE PCR-Free Libraries.

10-23-2013, 02:36 PM	#4
atcghelix Member Location: CA Join Date: Jul 2013 Posts: 74	If you pull out the reads that have been marked as duplicates, do they look like they are highly repetitive? And just to make sure, the reads that are showing up as duplicates are reads that map to the Bos taurus genome, yes?