Thanks, Bob, for your reply and also for pointing to the correct
chapters in the lastz-readme! Having said that, it's great to hear
the wrapper is scheduled to be extended/improved.

> It's not clear whether the wrapper would be better if it allowed
> each scoring option to be set as a separate field, or if they were
> incorporated in a file, or if (somehow) they were one of your
> galaxy history items. Do you have any thoughts on that?

I'm a bit curious. On the one hand, it'd be truely convenient to
have all arguments specified just by submitting a "lastz-config"-
history-item to the wrapper. On the other hand, I assume most
galaxy users are sticking to YASRA-modes and only those with special
needs would go for custom-configs. I, for example, will have to
include lastz-mapping in a workflow and will save my desired
lastz-config within the according workflow item. In this regard, it
wouldn't be too nice, if I had to have a config-file handy (anyone
else willing to execute this workflow would need the config file as
well, right?!).

Thanks again!
Uwe

-----Urspr√ľngliche Nachricht-----
Von: Bob Harris [mailto:rshar...@bx.psu.edu] 
Gesendet: Montag, 6. Juni 2011 18:06
An: Appelt, Uwe
Cc: galaxy-user@lists.bx.psu.edu
Betreff: Re: [galaxy-user] LASTZ arguments in "Full parameter list"-mode

Howdy, Uwe,

(I don't speak for the galaxy group, but I watch the list for lastz- 
related messages.)

> Is it possible to specify more arguments in "Full parameter list"- 
> mode, particularly those that're implicitely described in YASRA-mode  
> (match/mismatch rewards and step-size (Z/step))?

As you indicate, the galaxy wrapper for lastz only provides access to  
certain combinations of arguments.  At present, the only ways I know  
that you would be able to access other lastz parameters would be  
either to (1) write your own custom galaxy wrapper or (2) install  
lastz and run it from the command line.

Kelly Vincent is planning to update the galaxy wrapper some time this  
summer, so this is something she's likely to address with that update.

> However, heading to "Full parameter list"-mode leads to excessive  
> CPU-use, because (presumably) the step-size (default=1?) is left  
> unset.

You are correct that the default step is 1.  (defaults are shown at  
the bottom of each command section in the lastz readme file which can  
be found at www.bx.psu.edu/~rsharris/lastz or at the Miller Lab  
website)  Depending on your data, a step of 1 may be overly sensitive  
causing lastz to spend a more time than necessary.

> Also, it'd be great to be able to alter not only gap penalties, but  
> match/mismatch rewards as well. (Obviously a mismatch would cause  
> the same problem).


Ideally I think the wrapper interface should allow you to point at a  
lastz scoring file, which can contain all the scoring parameters.   
This is an oversight-- the thinking when the wrapper was written was  
that the yasra settings would be sufficient (and work well) and would  
simplify the user's choices, making it more likely that the user would  
choose settings appropriate for their data.

I've discussed this with Kelly some, this morning.  It's not clear  
whether the wrapper would be better if it allowed each scoring option  
to be set as a separate field, or if they were incorporated in a file,  
or if (somehow) they were one of your galaxy history items.  Do you  
have any thoughts on that?

> I'm dealing with 454 reads and it's crucial to the scenario to have  
> the 5' ends aligned properly (in terms of sensitivity), so all the  
> YASRA-templates comprising heavy gap penalties perform fairly poor  
> as soon as there's a gap nearby the 5' end.

The idea behind having such severe gap penalties is that 454 often  
incorrectly calls the length of homopolymer runs, introducing what  
will look like short gaps.  As used within yasra (an assembler from  
the Miller Lab, not in galaxy) these settings are probably  
appropriate.  But this is not an ideal general solution because it can  
keep us from discovering true gaps.  A better solution would probably  
be to have less-severe gap penalties, with an additional context- 
related gap penalty (or reward) for gaps at homopolymer runs.   
However, it would be costly to add this to the alignment core inside  
lastz.

The problem with gaps or mismatches close to the end of a read is  
discussed in the lastz readme file, in the section "Y-drop Mismatch  
Shadow".  The situation can can be improved by using the --noytrim  
option.  This option was added to lastz after the wrapper was written,  
and so is not currently available from galaxy.  --noytrim tells lastz  
to accept a lower-than-maximal-scoring alignment if it can reach the  
end of the read.  I intend to add that into the yasra settings, but  
changing those raises an issue of backward compatibility that I need  
to resolve.

I hope that helps.  Please post a reply if I've left something  
unanswered or if you have other thoughts on this.

Bob H


On Jun 6, 2011, at 6:47 AM, Appelt, Uwe wrote:

> Hi @All,
>
> short version:
> Is it possible to specify more arguments in "Full parameter list"- 
> mode, particularly those that're implicitely described in YASRA-mode  
> (match/mismatch rewards and step-size (Z/step))?
>
> long version:
> I'm dealing with 454 reads and it's crucial to the scenario to have  
> the 5' ends aligned properly (in terms of sensitivity), so all the  
> YASRA-templates comprising heavy gap penalties perform fairly poor  
> as soon as there's a gap nearby the 5' end. However, heading to  
> "Full parameter list"-mode leads to excessive CPU-use, because  
> (presumably) the step-size (default=1?) is left unset. Also, it'd be  
> great to be able to alter not only gap penalties, but match/mismatch  
> rewards as well. (Obviously a mismatch would cause the same problem).
>
> Examples:
> AAAAAAAAAA (target)
> AAAAATAAAA (query)
>
> => I need the alignment boundary to be found at position 1, rather  
> than behind the mismatched T and this cleary doesn't work as long as  
> the mismatch penalty is too large (same applies for gap-open/extend  
> penalties).
>
> Thanks in advance and Cheers,
> Uwe
> ___________________________________________________________
> The Galaxy User list should be used for the discussion of
> Galaxy analysis and other features on the public server
> at usegalaxy.org.  Please keep all replies on the list by
> using "reply all" in your mail client.  For discussion of
> local Galaxy instances and the Galaxy source code, please
> use the Galaxy Development list:
>
>  http://lists.bx.psu.edu/listinfo/galaxy-dev
>
> To manage your subscriptions to this and other Galaxy lists,
> please use the interface at:
>
>  http://lists.bx.psu.edu/


___________________________________________________________
The Galaxy User list should be used for the discussion of
Galaxy analysis and other features on the public server
at usegalaxy.org.  Please keep all replies on the list by
using "reply all" in your mail client.  For discussion of
local Galaxy instances and the Galaxy source code, please
use the Galaxy Development list:

  http://lists.bx.psu.edu/listinfo/galaxy-dev

To manage your subscriptions to this and other Galaxy lists,
please use the interface at:

  http://lists.bx.psu.edu/

Reply via email to