Consider dovetail pairs as concordant pairs?

[wisense]

Hi, all

Recently I have a paired-end data from Illumina/Solexa sequencing machine. My sequencing DNA fragments are about 100-300 bp, and my read length is 125 bp for one end of the pairs. My question is, if the dovetail pairs are general in my data? Do I need to consider these pairs as concordant pairs in the Bowtie2 alignment?

I have tried to specify "--dovetail" parameter in Bowtie2, and it has 10% more concordant pairs in the alignment compared to the results without "--dovetail".

Without "--dovetail":

30047395 reads; of these:
  30047395 (100.00%) were paired; of these:
    7364630 (24.51%) aligned concordantly 0 times
    18847070 (62.72%) aligned concordantly exactly 1 time
    3835695 (12.77%) aligned concordantly >1 times
75.49% overall alignment rate

With "--dovetail":

30047395 reads; of these:
  30047395 (100.00%) were paired; of these:
    4190935 (13.95%) aligned concordantly 0 times
    21610905 (71.92%) aligned concordantly exactly 1 time
    4245555 (14.13%) aligned concordantly >1 times
86.05% overall alignment rate

Hope to receive your suggestions, thanks.
wisense

[dcameron]

My question is, if the dovetail pairs are general in my data? Do I need to consider these pairs as concordant pairs in the Bowtie2 alignment?

Have you done any adapter trimming? Alignment of full-length reads is likely to result in soft clipped dovetailing reads. Since a significant portion of your library contains fragments shorter than the read length, dovetailing fragments are not unexpected. What you should do with them depends on your pipeline and analysis intent, but yes it is likely that the dovetailing reads originated from fragments shorter than the read length.

Be aware that the untrimmed reads can overalign due to microhomology between the adapter and the reference at the alignment position thus resulting in reads that appear to have the incorrect orientation (which for structural variant calling, could be a big issue).