novoalign - allow for more mismatches

[orionzhou]

Hi,

I want to use novoalign to map reads coming from relatively divergent genomes - allowing up to 8 mismatches for 90bp paired-end reads. Here is the command I used:

novoalign -d <genome> -f <1.1.fq.gz> <1.2.fq.gz> -k 200 -r E 10 -o SAM > <1.sam>

However, this only aligns reads with up to 2/3 mismatches, and produces much less alignments than BWA (using "-n 8" option). Anyone having any idea which parameters I should change or add?

Thanks,

[sparks]

Hi,

Did you mean -t 200 not -k 200? Or -k -t 200?

-t 200 will allow around 6-7 mismatches in each pair.

For divergent genome I suggest not using -t, the default will allow 8 or more mismatches per read.

I'd also add -x 4 to reduce the gap extend penalty, We use -x4 or -x6 for almost all Novoalign runs. The default is a bit too high. The gap open might also be reduced if you expect divergence to include more gaps. -g 20 might be a suitable value. Setting these too low can cause false positive alignments so you need to take care. I'd be tempted to run a few tests and maybe even analyse the frequency of short indels in the alignments and then set gap penalties to suit.

And consider using quality calibration, -k option. This could be interesting with divergent genomes because it will bring mismatch penalties down in line with divergence rate (% of mismatches) and should allow more mismatches. I'd test on 50K reads to see what difference it made.


Colin