duplicates

[inzaghi]

I have Illumina pair-ended seq data. I try to remove duplicates from each DNA fragment.

For example, I have two pairs reads. A1 A2 and B1 B2. After Bowtie alignment, I have data as follows:

pair A ----------- chr ----------- strand ----------- position ----------- mate

A1 ----------- chr3 ----------- plus ------------- 37 ----------- 1
A2 ----------- chr3 ------------ minus ----------- 137 ----------- 2


pair B ----------- chr ----------- strand ----------- position ----------- mate

B2 ----------- chr3 ----------- plus --------------- 37 ----------- 2
B1 ----------- chr3 ------------ minus ----------- 137 ----------- 1


The only difference between pair A and pair B is the mate. Are pair A and pair B are duplicated? Thanks!

[Heisman]

I don't quite understand your question. In general, any reads that come from the same original biological template strands are duplicates. In reality, if two reads come from the same biological template but one has incurred a PCR error, should they be considered duplicates? It would be harder to say for certain that they came from the same biological template.

Picard MarkDuplicates, for paired end reads, works by looking at the 5' alignment position of each mate; if these match a different paired read then it considers those two pairs to be duplicates, even if some of the bases are different due to errors.

[inzaghi]

I don't quite understand your question. In general, any reads that come from the same original biological template strands are duplicates. In reality, if two reads come from the same biological template but one has incurred a PCR error, should they be considered duplicates? It would be harder to say for certain that they came from the same biological template.

Picard MarkDuplicates, for paired end reads, works by looking at the 5' alignment position of each mate; if these match a different paired read then it considers those two pairs to be duplicates, even if some of the bases are different due to errors.

Hi Heisman, thanks for your reply

Picard MarkDuplicates, for paired end reads. Does that mean reads from mate 1 must match reads from mate 1 and reads from mate2 must match reads from mate2 ?

I mean if reads from mate 1 could match reads from mate 2.

[Heisman]

Hi Heisman, thanks for your reply

Picard MarkDuplicates, for paired end reads. Does that mean reads from mate 1 must match reads from mate 1 and reads from mate2 must match reads from mate2 ?

I mean if reads from mate 1 could match reads from mate 2.

Ok, I understand, I think. If you have a paired read with mates A and B, and a second paired read with mates A' and B', would it be a duplicate read if:

1. A matched A' and B matched B'

2. A matched B' and B matched A'

If the reads are mapped with the same orientation, then I believe there is really no meaning to separating scenario 1 and 2 above. So, in that sense, in both scenarios they would be duplicates.

[inzaghi]

Ok, I understand, I think. If you have a paired read with mates A and B, and a second paired read with mates A' and B', would it be a duplicate read if:

1. A matched A' and B matched B'

2. A matched B' and B matched A'

If the reads are mapped with the same orientation, then I believe there is really no meaning to separating scenario 1 and 2 above. So, in that sense, in both scenarios they would be duplicates.

Hi Heisman,

You explain very well

Thank you so much for your prompt reply!

[blanco]

Hi folks,
I used Picard to remove duplicates but it removed around 60% of the reads (samtolls rmdup removed even more). That seems quite a lot - I have heard around 30% is normal.

My library was performed with RiboMinus so there are likely some Ribosomal sequences in the library. Could that account for the massive amount of removed reads?

thanks,
blanco