Tweet
Hello,
I am analysing paired-end data (2*100bp, Illumina sequencing) from ChIPseq experiment performed on a transcription factor.
- Library preparation was done following the Illumina protocol. PCR amplification was performed after adapter ligation and then size selection was performed by gel migration. The selected size was 400 +/- 50 bp , so it corresponds to an insert size around 280 +/- 50 bp (when the adapter length of approximately 120bp was deduced).
- Fragment size was checked on a Agilent DNA High Sensitivity chip: for all samples, fragment size was confirmed to be the one selected on gel (one beautiful peak for each sample)
- Paired-end sequencing (2*100bp) was performed on a HiSeq 2000 Illumina.
- Fastq quality control was ok
- Alignment was performed with Bowtie2 and I kept only pairs uniquely mapped, with no more than one mismatch per read and I removed duplicates.
Then,
- I checked the insert size by using picard-tools CollectInsertSizeMetrics and the distribution of insert size I obtained was very weird (see attached file):
* The peak to the right corresponds to the insert size I was expected,
* but for me, there is no reason to have the second peak to the left (smaller inserts) because size selection on gel was done and checked before.
I tried to look at the genome distribution of small versus long inserts but I was unable to see any difference.
I noted that the insert size of the left peak corresponds to half the size of the right peak, I don't know if this could help.
I visually checked the sam file for calculation of insert size and this appeared ok.
So, what could be this peak?
Have anyone an idea about this or have anyone ever observed this on this kind of data?
How can I handle this ? I can just eliminate smallest inserts, but I really want to know how it is possible to have inserts smaller than the size expected.
Thank you for your help!!!
I am analysing paired-end data (2*100bp, Illumina sequencing) from ChIPseq experiment performed on a transcription factor.
- Library preparation was done following the Illumina protocol. PCR amplification was performed after adapter ligation and then size selection was performed by gel migration. The selected size was 400 +/- 50 bp , so it corresponds to an insert size around 280 +/- 50 bp (when the adapter length of approximately 120bp was deduced).
- Fragment size was checked on a Agilent DNA High Sensitivity chip: for all samples, fragment size was confirmed to be the one selected on gel (one beautiful peak for each sample)
- Paired-end sequencing (2*100bp) was performed on a HiSeq 2000 Illumina.
- Fastq quality control was ok
- Alignment was performed with Bowtie2 and I kept only pairs uniquely mapped, with no more than one mismatch per read and I removed duplicates.
Then,
- I checked the insert size by using picard-tools CollectInsertSizeMetrics and the distribution of insert size I obtained was very weird (see attached file):
* The peak to the right corresponds to the insert size I was expected,
* but for me, there is no reason to have the second peak to the left (smaller inserts) because size selection on gel was done and checked before.
I tried to look at the genome distribution of small versus long inserts but I was unable to see any difference.
I noted that the insert size of the left peak corresponds to half the size of the right peak, I don't know if this could help.
I visually checked the sam file for calculation of insert size and this appeared ok.
So, what could be this peak?
Have anyone an idea about this or have anyone ever observed this on this kind of data?
How can I handle this ? I can just eliminate smallest inserts, but I really want to know how it is possible to have inserts smaller than the size expected.
Thank you for your help!!!