How to speed up BFAST run time?

[JackieBadger]

no i dont. email the bfast mailing list for help

[AmitL]

Sorry, I read the supplements but I still don't understand.

Do you use bfast regularly?
How did you index your reference?

[JackieBadger]

I dont work on human genome. You can get all the info from the paper to duplicate their run.
From what I read, for 10 indexes of the human genomes they used
a key size of 22
Hash width of 14
K cals (k) of 8
M = 1280

[AmitL]

Do you work with human genome?
If you do, can you share the commands you use when you run it and how long it takes?

Thanks

[JackieBadger]

The publication explains

[AmitL]

Thanks for the help, but I don't completely understand.
How can multiple binary keys speed up the search?

[JackieBadger]

ps.
multiple indexs are run separately, but in total will require more memory

[JackieBadger]

Take a look at the supplementary material which goes with the original BFAST publication. It explains all of the parameters in detail. It is quiet complex, and requires some exploration of parameters which will suit your genome of interest. Lucky for you the authors provide 10 binary keys for which you can use to execute 10 indexes within a human genome. If you follow the methods from the publication you should get decent recults in minimal time (they optimized their protocol against a human genome).

[AmitL]

I used a single index.

How do you create multiple indexes?
And what do you do with them? Do you then run multiple alignments in parallel?

Amit

[JackieBadger]

With shorter reads you would speed up your alignments by creating more than one index. With reads longer than 100bp you could get away with one index on a human sized genome. The authors use 10 indexes for the human genome in their example.

How many did you create?

[AmitL]

Hi guys,

I got warm recommendations on BFAST and decided to try it on my SOLID human sequencing data.
I got to the 'bfast match' stage, but it takes ages to run. Going from a single thread to multiple did not affect the running time significantly (what threading library is used here?).

I'll be grateful if someone can share their experience on how to speed this process up.

Thanks,
Amit