Comparison of alignment tools

[genbio64]

Dear Seqanswers,
Recently my lab purchased a license for CLC genome workbench to use for our next-gen sequencing needs. We have been using Bowtie/SAM Tools but everyone else liked the ease of use of this software. Does anyone have experience using both of these pieces of software and can you tell me how accurate they may be? Obviously CLC wins in terms of presentation hands down, but from what I have read Bowtie is not only faster, but also more accurate. Any feedback will be greatly appreciated.

Cheers,

[drio]

What kind of experiments are you planning to run? What kind of sequencer do you have? What type of sequence are you going to generarte? I don't have experience with CLC or bowtie, but I suggest you take _also_ a look to bwa and bfast.

[ashrafi_h]

I tried BWA to align long reads to reference sequence and at the same time I used CLC. CLC seems to map a lot more reads (>75%) for reads are mapped vs. bwa (30%). I acknowledge that bwa has not been written to map the long reads but it has the long read indexing that allows you to do this to some extend. However, I am not sure more is better either. Anybody has any input?

[nilshomer]

BWA long should work "bwa bwasw".

[ashrafi_h]

bwa bwasw vs CLC

"bwa bwasw" Works! humm, Yes and no. The same 454 data used for BWA and CLC. CLC mapped a lot more reads than BWA (29K for BWA vs. 1.2M for CLC). But that was just with one data set. Other data sets were Ok but still CLC mapped more. I am not sure what was wrong with those 454 reads or BWA that did not work well.

[nilshomer]

Have you considered novoalign, which is another commercial offering? BLAT and ssaha2 also work for longer reads. Sounds like you're doing good due diligence.

[lh3]

bwasw would not work well when your reads are short or the error rate is high. For typical 454 reads produced nowadays, most mappers can map the vast majority. Also, for 454 alignment where you only want to find local hits, it is actually fairly easy to find one significant hit, but the question whether we should trust it or not.