ask for gapped (indel) alignment software

[polyhedron]

i have now single-end solexa data and later will be also working on PE.

i've tried soap2 for the -g 5 option for gapped alignment, however, on known indels, there's nothing properly mapped. although i've seen on seqanswers someone did gapped alignment successfully (but on soap rather than soap2?). and the developer says soap2 can do gapped PET alignment, but "gapped" means the interval between the ends rather than indels?

bowtie says it doesn't support gapped alignment yet.

may anyone tell me which free software can do gapped alignment (for indels) on both SE and PE data? many thx!

[francois.sabot]

I may say that BWA accepts indel...

[Michael.James.Clark]

Novoalign, BWA, BFAST

Oh, Novoalign is not exactly free. You can get a free trial from them, though, and it is not very expensive.

[polyhedron]

Thanks very much, François and Michael!

I've been testing bwa these days, and find bwa have considerable better mappability than soap2, and good with indels. The trade-off is that bwa is 3 or 4 times slower than soap2, and I haven't find so flexible output options as in soap2, like

-r INT How to report repeat hits, 0=none; 1=random one; 2=all, [1]

Therefore, I think when I've got really huge data, I would rather run soap2 first to get some basic idea and results from my reads, then just let bwa running for days to get more precise results.

[sci_guy]

SHRiMP is very robust to INDELs, but is very slow (as it performs Smith-Waterman alignments).

[lry198010]

-r INT How to report repeat hits, 0=none; 1=random one; 2=all, [1]

you can try options -n and -N in sampe or samse subrouting!

[drio]

Therefore, I think when I've got really huge data, I would rather run soap2 first to get some basic idea and results from my reads, then just let bwa running for days to get more precise results.

Days? BWA is very fast and accurate. How much data are you trying to align? On what environment? A lane of hiseq (75bp single ended) should take less than 1 day on a typical 8 core machine. You can further reduce the running times by splitting and computing the data into multiple machines.

[Michael.James.Clark]

Yeah, I'm with drio on this one. Something is wrong with your approach using BWA if it's taking days, even for a very large data set. You may not be running it in an optimal fashion for your computational resources if that's the case.

All three gapped aligners that I mentioned--Novoalign, BFAST, BWA--are quite fast with Illumina reads, so if it seems slow, there may be an incorrect setting or it may be getting run in a suboptimal way.

[olesk]

Depending on the size of your reference sequence and amount of sequence data you could test if my program R2R is your solution.
Find it at: http://milne.ruc.dk/R2R/

[lexa]

there is a new mapper available called STAMPY (http://www.well.ox.ac.uk/project-stampy). the paper seems promising but I haven't tested it myself.

[colindaven]

The best tools I've used for alignment of small indels are Stampy and Shrimp2. We've validated quite a number of these.
Unfortunately they are slow as well.

As a follow-up I've been using Dindel on Stampy aligned data for further testing. No wet-lab data on this yet though.

[ttnguyen]

Novoalign and BWA work quite well for this problem. BWA is faster than the free version of Novoalign (not support MPI), but Novoalign maybe more accurate.
It might be worth taking a look at these surveys - though they are not the most up to date:

http://lh3lh3.users.sourceforge.net/NGSalign.shtml

http://bib.oxfordjournals.org/content/11/5/473.full