This is a test version of Biostars. For the public version, visit https://www.biostars.org.
Artificial reads - remove multiple mapped reads against reference genome, and only keep reads that completely match without any mismatches - samtools

Hello,

I have 7.5million 25mer artificial reads, I mapped them against a reference genome using bowtie2.

The following flag was used bowtie2 -k 2 --very-sensitive

7558491 reads; of these:
  7558491 (100.00%) were unpaired; of these:
    1399350 (18.51%) aligned 0 times
    5200484 (68.80%) aligned exactly 1 time
    958657 (12.68%) aligned >1 times
81.49% overall alignment rate

The are the results, I want to remove all the unmapped and multiple aligned reads. I understand, I have to take into account mapping quality, but as these are artificial reads, would I still use:

samtools view -F 4 -q 2 test.bam | wc -l

samtools
• 2,118 views • link
written by Jalil Sharif • Jalil Sharif

samtools does not know or care that the reads were artificial. It is going to carry out the operation you are asking it to do.

only keep reads that completely match without any mismatches

Hopefully you controlled for that already in your bowtie command.

• 0 views • link
written by GenoMax GenoMax

That is obvious, but does not address the fact, of what the most appropriate way is to remove reads that map to multiple locations.

• 0 views • link
written by Jalil Sharif • Jalil Sharif

Mapping qualities are not handled the same way by different aligners. So with bowtie2 you could use a MAPQ filter of >=40 to get reads which had only 1 convincing alignment as noted in the linked blog post.

• 0 views • link
written by GenoMax GenoMax

I ran samtools view -F 4 -q 42 test.bam | wc -l and I get 6159141, why is there a discrepancy, as originally only 5200484 aligned?

When I did the calculations 6159141 - 5200484 = 958657, so I am still retaining the 958657 multi-mapped reads.

• 0 views • link
written by Jalil Sharif • Jalil Sharif

0 answers

No answers yet.

Log in to answer this question.

More posts like this