[EMBOSS] EMBOSS equivalents for GCG programs

[Peter Rice]

Derek Gatherer wrote:

> I have to confess, I'm a bit baffled by this.  Obviously, EMBOSS needle 
> would do a NW alignment, but the gap opening penalty default in needle 
> is 10 and the gap extension is 0.5.  I'm not sure what is implied in Gap 
> by "gap creation", or whether or not needle allows any end gap 
> penalties.  Does anybody have any particular recommendations about the 
> appropriate needle parameters to use to emulate Gap's behaviour?

In a Needleman-Wunsch global alignment, the entire length of each 
sequence is aligned. This can be thought of as an overlap between the 
two sequences (one can be completely within the otehr, or their ends can 
overlap).

This leaves no penalty for the hanging ends of the overlap. In 
bioinformatics, it is usually reasonable to assume that the sequences 
are incomplete and there should be no penalty for failing to align the 
missing bases.

In GCG's gap, "endweight" means that the hanging ends are treated as two 
more rather long gaps. This tends to force the ends of the sequences to 
be aligned (whether they should be or not) to reduce the number of gaps, 
and is usually not the option you want.

it would not be too difficult to support endweights in needle (the 
difference is in setting up the gap penalties in the matrix at the start 
and end, the rest is the same).

But ... is it useful?

> I'm also looking for an equivalent of the GCG program Diverge, that 
> calculates Ka/Ks values.  I think there is no equivalent to this in 
> EMBOSS, but I may have missed something?

diverge is a strange best that was not too happy with large input data 
sets. GCG replaced it by "newdiverge" some time back with a different 
algorithm.

EMBOSS supports phylip 3.5, which has distance methods, as an EMBASSY 
package. You may find what you need there, and phylip 3.6 will be stable 
soon and will appear as a new and more closely integrated EMBASSY 
package soon after.

Again ... would phylip 3.6 in EMBOSS be useful?

regards,

Peter Rice