how critical is the filtering of potential PCR duplicates?

[julien]

Hello,

I don't really have a good appreciation for how critical it is to remove potential PCR duplicates from the alignments (either via samtools rmdup or picard). It would appear (based on posts in this forum and pipelines that I have found online) that many ??? do not worry about this issue, am I wrong ? It did not seem to affect my data very much but I would like to know how important an issue this is.

Thank you !!!!

[nilshomer]

Hello,

I don't really have a good appreciation for how critical it is to remove potential PCR duplicates from the alignments (either via samtools rmdup or picard). It would appear (based on posts in this forum and pipelines that I have found online) that many ??? do not worry about this issue, am I wrong ? It did not seem to affect my data very much but I would like to know how important an issue this is.

Thank you !!!!

Most whole-genome human resequencing papers try to detect PCR duplicates and I find that it is extremely important to remove them (or flag them), especially when dealing with low-complexity libraries. The latter can happen with small amounts of input DNA or re-using the same library. When searching for variants, it is assumed that each read is conditionally independent given the underlying DNA sequence (i.e. not the same DNA fragment read twice).

[dnusol]

Hi nilshomer, sorry for getting back to this thread, but do you know if Picard is able to remove duplicates from single-reads? Samtools manual does not recommend it but I am not sure if it is a matter of implementation or a real technical issue on the single read part which I don't get.

best

[nilshomer]

Hi nilshomer, sorry for getting back to this thread, but do you know if Picard is able to remove duplicates from single-reads? Samtools manual does not recommend it but I am not sure if it is a matter of implementation or a real technical issue on the single read part which I don't get.

best

If you have high coverage (say 50x for 50bp reads), then it becomes very likely that multiple single end reads will start at the same coordinate. You will then wont be able to tell coverage from PCR duplicates. For PE or LMP reads, both ends can be used to determine duplication, since the insert size has its own distribution.