how to get uniquely aligned reads from bowtie

[anurupa]

hi
i have rum bowtie 2 using —very-sensitive option and got result like this

14559177 reads; of these:
  14559177 (100.00%) were unpaired; of these:
    4317487 (29.65%) aligned 0 times
    7816644 (53.69%) aligned exactly 1 time
    2425046 (16.66%) aligned >1 times
70.35% overall alignment rate

i just want to get the uniquely aligned reads alone i,e., 53.69% of reads (78126644). but my output file (sam format) seems to have the those which are aligned>1 times how to get rid of them?

in previous versions of bowtie i use to use -m option for that

[danielr]

Filter based on having the same read name (1st column) at multiple rows, or based on the mapping quality (5th column).

[lindylou]

filter on the 5th column, only retain reads with a q score of 255. This is the score given to all reads that are uniquely mappable. If you have bam files you can use

samtools view -q 255 your.bam > your_filtered.sam

[MGineste]

hi
i have rum bowtie 2 using —very-sensitive option

i just want to get the uniquely aligned reads alone i,e., 53.69% of reads (78126644). but my output file (sam format) seems to have the those which are aligned>1 times how to get rid of them?

in previous versions of bowtie i use to use -m option for that

Bonjour anurupa,

I'm encountering the same trouble following switching from bowtie1 to bowtie2. Did you find out a satisfying solution ?

I'm not sure that a filtering based on MAPQ using [samtools view -q] gives the desired result, as a condition for "uniqueness" for a given read is that its MAPQ is high on the best mapped position AND "much" higher than in any other mapped position.

Mathieu

[MGineste]

Bowtie2 must somehow define what is a unique alignment (may it be related to the MAPQ or not). This definition remains unclear to me.

Mathieu

[MGineste]

SAM files created by bowtie contains a 'XS:' tag for reporting secondary alignments for a given read.

You can check that the number of lines containing the 'XS:' tag corresponds to the number of reads showing >1 alignment(s) :

grep 'XS:' your_alignment_file.sam | wc -l

A working solution consists in removing the lines containing the ':XS' tag using sed :

sed '/XS:/d' your_alignment_file.sam > your_alignment_file_1alignmentonly.sam

(Thanks go to Christelle, our group's everyday-very-valuable bioinformatician...).

[anurupa]

you are correct MGineste. I have filtered my reads by

 grep -v XS:i: my file>output

[chefarov]

Removing the lines with "XS:" will leave the corresponding headers behind, which could create a problem to SAM/BAM viewers. That happened in my case.