[Biodevelopers] Re: BLAST asymmetrical

[Martin Gollery]

Smith/Waterman is not global, Chris- it's local. It's Needleman-Wunsch
that is global.

On 1/20/07, Christopher Dwan <cdwan at bioteam.net> wrote:
>
> > I want all homologies.
> >
> >> Or even Smith-Waterman which will take a while to run.
> >
> > Do you know of a program that can calculate SW on a pair of genomes?
>
> This may be a semantic confusion on my part, but here's my answer to
> that specific question:
>
> If you really want the single best *global* alignment between two
> multi megabase sequences, yes SW is the way to go, and yes, it will
> take a really long time.  On the other hand, I've never met anyone
> who really, seriously cares about monolithic, global alignments of
> chromosomes.  Go down that road, and the next question will be "why
> can't we just run clustalw on whole chromosomes?"  Yes, of course you
> could ... but it'll be really slow and not very useful.
>
> Note:  This is not an invitation to the accelerator people in the
> audience to offer me a *faster* clustalw or SW.  I'm trying to steer
> people toward *better* uses of the tools.  You might as well work on
> multi-gigabyte cut-and-paste buffers so that I can stuff whole
> genomes into the NCBI web interface.
>
> On the other hand, if you want the best gene sized (a few kilobase)
> matches from within that pair of megabase sequences, it's a different
> question.  You're going to wind up chopping each sequence into
> overlapping chunks and running an all against all search of some
> sort.  The chunk size will be determined by how large you think the
> introns and exons in your genes are.  An even more clever approach
> might involve doing preliminary gene calls with a gene finding
> program like Glimmer, and then starting the all against all search
> from those hits.
>
> Chromosome vs. chromosome BLAST answers the question "is there a
> decent hit to any part of this chromosome in that other one".  The
> answer, broadly speaking, will be "yes, there is a statistically
> significant match there."
>
> If you want homologous genes, you're going to have to do a bit more
> work than just running a single program to get The Answers.
>
> -Chris Dwan
> _______________________________________________
> Biodevelopers mailing list
> Biodevelopers at bioinformatics.org
> https://bioinformatics.org/mailman/listinfo/biodevelopers
>


-- 
-- 
Martin Gollery
Associate Director
Center For Bioinformatics
University of Nevada at Reno
Dept. of Biochemistry / MS334
775-784-7042
-----------