Weird distribution of insert size

[zacshi]

I got a strange distribution of insert size as analyzed by Qualimap on the BAM file. Not totally aware of the library preparation procedure yet, but this is a PE-sequencing on NextSeq500, libraries were made with NEBNext Ultra II kit and size selection beads. The median insert size was expected to be around 300bp. The problem is that we got not only a much smaller median insert size (190bp), but also a non-symmetrical distribution. What could be the reason? Something related to the trimming I used to remove the adapters? Thanks for any help.

[GenoMax]

Can you tell us what aligner you used for this? Is it possible that the aligner kept reads with reasonable distance (did you provide an expected fragment size) when mapping and discarded smaller inserts (what was the alignment %?). Otherwise you are seeing a distribution of insert sizes that were in your dataset.

[zacshi]

Thanks! The aligner used was bwa-mem, nothing set up to restrict the fragment size. Mapping rate was like 99.8%. I guess that I should re-do the alignment with the fastq files before trimming just for curiosity.

[kcchan]

Are you sure the 300bp was the insert size and not the total fragment size? The adapters add about 150bp to the total fragment size so this looks about right in that case. The asymmetric distribution on the left hand side is caused by the size selection beads removing the shorter fragments.

[GenoMax]

It must be the total frag size. Since the above plot is post alignment it will correspond nicely to 150 bp inserts.

[zacshi]

From the fastQC results, we saw plenty of adapter sequence, which could be attributed to the wrong (smaller) insert size selection. So I cut the adapter sequence and did the alignment. The shown weird distribution was from the alignment after trimming. Now I have just finished alignment with the raw fastq files without any trimming, the distribution of insert size looks more symmetrical, at least no sudden loss of small inserts. So it seems the shape was caused by trimming (and loss of reads without minimum length which I set at 80). Anyway, since our PE read length is 151, the library was simply not prepared with the right insert size (300). Thanks for all the comments. I appreciate it!