maq with no mismatches

[staylor]

Hello,

Is it possible to run maq to allow no mismatches in the alignment? I tried using -n but that doesn't seem to allow 0 as a value. I get the error:

[ma_match] please use 1, 2 or 3 with option -n. Abort!

Thanks for any help,

Steve

[Torst]

I agree it is odd that maq fails with -n 0 but I guess that is the special case of an exact match which doesn't really need special software like maq.

Standard substring matching will do this quite fast for you, you just have to remember to check the revcom strand as well.

eg. substr() in C, substr() in Perl and so on.

[zee]

You could get good performance with novoalign with FASTQ reads as input:

novolaign -f <reads.fastq | reads.fasta > -d <db> -t 10 > output.

This would work on very long reads e.g. 50bp, 75bp ?

You could then use the novo2maq (in www.novocraft.com distribution) to do convert to MAQ's .map format and then carry on as per usual with your analysis.

[staylor]

I will certainly install novoalign, looks interesting.

In answer to Torst's response I need something that takes into account paired ends which would be a bit more involved if I wrote my own simple matcher...

I'd still be interested to know in maq what would be sensible cut off values to filter the existing output to get an idea of ones that are unlikely to have mapped. perhaps based on mapping quality (col 7 of mapview) and/or number of 0-mismatch hits of the first 24bp (col 12 of mapview).

Any ideas?

[zee]

I usually filter out alignments with MAQ mapping quality scores <= 10. Other criteria do come into play e.g. with paired-ends you might remove unpaired reads, incorrect orientation,etc.
Novoalign does do paired-end in case you were wondering about this.

[bioinfosm]

zee, probably not the best thread for this question.. what support would novoalign have for paired-end reads as regards Structural variant detection.

I am having a hard time moving ahead with the MAQ SV results..

[aadey]

I was able to run Maq with 0 mis, however I believe this is only in the first 24 bases of the read. The way I have found works best for sorting it all out is just by filtering your output file from maq mapview by the total mismatches column to assure the stringency level you are going for.

[zee]

Bioinfosm,

Maq has an SV detector although some of my collaborators have complained that it falls short in some ways.
Remember that with `maq.pl sv` there is some SV detection.

With novoalign, the tail end of the output file contains information on the fragment length distribution based on aligned pairs. I wrote a simple perl program to reformat the alignment output into BED format where each different call (long insert, delete, incorrect orientation,etc) was split into a different BED track.
I can share this script.

[bioinfosm]

Thanks Zee!