high percentage of **total_single_sided_mapped**

[wbszhu]

Hi，

In the results (for proximity ligations in the genome with the help of linker), I found a high percentage of total_single_sided_mapped ~30%, what could be the reason for this?

LuZhang

[Phlya]

Hi, did you do anything special to deal with the linker sequence?

[wbszhu]

I identified the linker by using the tool ChIA-PET2 to extract the real reads from the raw data and remove the linker part before going to pairtools to sort the pairs

[wbszhu]

I can't quite answer the question about the experimental handling of the linker. Because an unpublished technique is involved.

[agalitsyna]

Then, if the resulting read length is shorter than ~30 bps, bwa mem might struggle to map these reads. It's dedicated for long reads alignments only.

[wbszhu]

Thanks，Do you think the resulting read length is shorter than ~30 bps should be kept?

[agalitsyna]

I usually use bwa aln instead of bwa mem for short reads mapping. It's syntax is more complicated, but it saved me a whole dataset once, when after trimming all the reads were 20-30 bps.
Quick assessment: for human genome you may expect unique hit of the read longer than 15 bps, below that value it may not help much to keep it, because short reads will have a large chance to be multi-mapped.

Let me know how it goes! I may share my version of distiller that uses bwa aln if that helps.

[wbszhu]

Thanks for your sharing, I will have a try.

[wbszhu]

I split the data into 16-70bp and greater than 70bp, using bwa aln and bwa mem respectively, and found that there are ~30% of total_single_sided_mapped in the 16-70bp data. Also, ~20% of the total_single_sided_mapped reads still exist in the greater than 70bp data.

[agalitsyna]

Hi!
Your procedure makes me think about the cases when the adapter sequence was clipped from one side of the read but not another (so that one side is short and another is normal). Would these cases fall into 16-70bp or >70bp category by your design?
It's also important that both R1 and R2 of the same read go into the same file for bwa mapping.

Other possibilities worthy checking:

• fastqc after trimming will show if there are any undercut adapter sequences, or some other sequences are repeatedly shared between the reads
• blast after extracting unmapped ones from .bam files would by help, it will provide an insight why some pieces of the reads are reported as unmapped by bwa:
  samtools view -b -f 4 HiC_057_439_embryo_rep1.lane1.XA.dm3-embryo.bam | samtools fasta --