Support

Posted by cwhelan on Tue 12 of June, 2012 01:15 MYT

gap extension penalty

Posted by cwhelan on Tue 12 of June, 2012 01:15 MYT

This is sort of a general question, but I was wondering if you could comment on your decision to change the default gap extension penalty (-x) from 15 to 6 in version 2.07.14. I'm never quite sure of the best way to discover how to set parameters like that myself, so I thought I might be able to learn from your reasoning. In a typical human resequencing project, what does changing from -x15 to -x6 mean in terms of the average size of indels in read alignments at default thresholds?

Thanks,

Chris

Posted by zayed on Tue 12 of June, 2012 01:55 MYT

Posted by zayed on Tue 12 of June, 2012 01:55 MYT

Hi Chris

Thanks for the great question.

The decision came through some comparative analysis with other aligners, most notably Bowtie and Stampy. We found in some short indel benchmarking that setting the gap extension too high was excluding valid results we should have been reporting that were picked up by other tools and were known to exist in the 1000genomes calls.
The effect of such a change will relax the cost to extend a gap during the dynamic programming part of novoalign.

Posted by colin on Tue 12 of June, 2012 06:08 MYT

Posted by colin on Tue 12 of June, 2012 06:08 MYT

HI Chris,

The length of indels Novoalign can detect will depend on gap penalties and the read lengths. For 100bp paired end reads the maximum at gap extend of 6 will around 40bp and at 15 will be around 16. In practice you'd probably only reliablt call indels about 60% of this size.

Colin