[Bioperl-l] All-vs-all BLAST

[Josh Lauricha]

On Thu 07/22/04 11:32, Andreas Kahari wrote:
> On Thu, Jul 22, 2004 at 11:09:45AM +0100, Peter van Heusden wrote:
> Or MCL, which I think Jason has provided some blast-related
> input to:
> 
>     http://micans.org/mcl/

In my experiance MCL gave fairly horrible results when used with
proteins that may have "fused" domains. But then, all the ones we tried
were flawed. MCL just claimed to deal with multiple domains. Anyhow,
this was last summer so a) it might have changed or b) I'm remembering
another program, even c) I'm just plain wrong ;)

Even so, I think MCL still requires an All vs All blast.

Another option (which is installed with blast) is blastclust. This
does single linkage, so it's not that great if you have multiple
domains in your proteins, but none are. The problem is, with
multiple domains you end up clustering like:
----------- --------------------- --------------
------
------
-------
--------
             ------------
             --------------
              -----------------
            ---------------
                                        ---------
                                     --------
                                        ---------

where that should be three clusters, it gets grouped into one.

Can stackPACK(?) deal with that gracefully?

-- 

------------------------------------------------------
| Josh Lauricha            | Ford, you're turning    |
| laurichj at bioinfo.ucr.edu | into a penguin. Stop    |
| Bioinformatics, UCR      | it                      |
|----------------------------------------------------|
| OpenPG:                                            |
|  4E7D 0FC0 DB6C E91D 4D7B C7F3 9BE9 8740 E4DC 6184 |
|----------------------------------------------------|