Bowtie 2 parameters

[chedonat]

Hi All,

I am using bowtie 2 to align some illumina sample data on an influenza genome which has less than 14k nt. The samples have more than 10 millions of paired reads. Each reads having 100 bp.

I initially launch the default bowtie 2 commands, without specifying supplement parameters. The percentage of paired reads which aligned correctly was between 35 and 50%. And the percentage of the overall alignment (including reads which align in a single way) was between 38 and 55 %.

After some readings on the manual, I changed the parameters to the following command:

bowtie2 -L 10 -N 1 -i S,1,0.20 --fr –x …….

Which means that a seed length of 10, One mismatches is allowed in a seed, the seed interval is 3 (1+0.2*10).

The alignment percentage increase between 40 to 75% for paired alignment and between 68 to 90% for overall alignment.

The alignment's results are quite better, but the matter is that with a deep look onto the alignment. I noticed that there is some reads which aligned with more than 10 mismatches. We can even found some with 14, 18, 21 mismatches.

That makes me doubt of my parameters and the quality of my alignment.

I am a newer in the Bioinformatics and I would like to have, please, your point of vue on that issue.

many thanks

[peromhc]

Does the lower mapping rate make sense in light of the biology of the virus-- very rapid evolution, etc??

Also, does whatever the research goal require that more than 38-55% of reads map. Nobody likes to throw away data, but in genomics/bioinformatics, one has to accept this to a degree..

Lastly, have you tried looking at the reads that are not mapped? Are they lower quality reads? Have you tried to assemble the unmapped reads? Is contamination possible?

[vbiaudet]

I think it's normal to have some reads with more than 10 mismatch with your parameters the maximum of mismatch is about 30 mismatch (100/3)
because you can have until 1 mimatch by seed of 10. You must try with N=0, the position of a read will be more accurate.
VB