[Bioperl-l] error during installation

[Peng Du]

Thanks for all your kind replies, based on my current knowledge, it might be worth making dpAlign at work even in the absence of BioPerl-ext package.
Thank you.

Peng

On 2012/06/20, at 2:38, "Fields, Christopher J" <cjfields at illinois.edu> wrote:

> On Jun 19, 2012, at 12:21 PM, Leon Timmermans wrote:
> 
>> On Tue, Jun 19, 2012 at 5:25 PM, Du, Peng <du at ibio.jp> wrote:
>>> As to the Clustalw problem, it turned out that, I didn't install it in
>>> the first place, I am very sorry for the stupid mistake. It is alright
>>> now. But, as to dpAlign, I compared the alignment of non coding
>>> regions of two species with dpAlign (in an Intel x86 but much slower
>>> machine where dpAlign is good) and Clustalw respectively and checked
>>> the output, I found that, each with default parameters, the alignments
>>> by dpAlign were better than that by Clustalw, and there was a
>>> tremendous difference in the outputs in terms of A<=>G, C<=>T, AG<=>CT
>>> substitutional rate and divergence by Tamura-Nei Model between the two
>>> alignments. The values are much smaller with dpAlign.  Since I am a
>>> newbie, I don't quite understand why there was such a difference. Is
>>> it might because of the Clustalw is designed for multiple alignment or
>>> the default parameters are not appropriate?
>>> 
>>> I am now investigating other tools available in the main distribution
>>> of  BioPerl, but I have no idea whether there is a good substitution
>>> of dpAlign in context of non coding region alignment.
>> 
>> ClustalW uses a heuristic algorithm, it cannot produce a solution that
>> is guaranteed to be optimal. dpAlign uses the Miller-Myers algorithm
>> (a variant on Needleman-Wunch), that does make such guarantees. It
>> should be slower but may produce a better alignment.
>> 
>> Leon
> 
> Agreed.  I think it would be worth pulling this out and releasing it on it's own, maybe including Bio::Tools::dpAlign with it as well.  The key problem is one of maintenance; this code was pretty much written by one developer.  I'll see if I can track Yee Man (the developer) down.
> 
> chris