[Bioperl-l] Bio::Score of interest?

[Sendu Bala]

Chris Fields wrote:
>> Hilmar Lapp wrote:
>>> So you basically want to attach semantic information to a number, and
>>> type the number thereby?
>> Basically, I want to be able to stick a bunch of (different kinds of)
>> numbers into an object, and later get the 'best' one out (of a
>> particular kind), or sort multiple of those objects.
> 
> The 'best one' might be tricky when dealing with different kinds of scores,
> esp. scores calculated different ways.

I didn't make myself very clear, but you don't compare different kinds 
of scores. When you want to compare two different Score objects, each of 
which may contain multiple different kinds of scores, you pick the kind 
of score you're interested in, and for that kind of score ask which 
object has the 'best' score. I can't readily think of any exceptions to 
the rule that 'best' is either the higher score or the lower score, 
depending on what kind of score you've chosen.

I may not have made myself clear in another way. One of the ideas behind 
a Bio::Score is to have a container object for multiple different kinds 
of scores (and even multiple values per kind) all generated by one 
program in one analysis on one data set.
The container then lets you pick the kind of score you want to work with 
and compare its scores with those in other Bio::Score objects that 
contain the same kind of score (most probably, ones made by the same 
analysis program but on different data sets).

Furthermore, the kind of score you want to work with could have multiple 
values from that single analysis. So the container also lets you 
summarise these values (eg. average them) before trying to compare with 
another Score object. Often, it may be that for a certain kind of score 
it makes sense (it is intended by the score-generating program) to 
always summarise the values in a certain way. So the container needs to 
know about that and 'do the right thing' so the user can just compare 
things without having to trouble himself.

So this is why I feel that to just 'use an ontology' isn't enough. 
Certainly one ought to be used when defining the kinds, but you need 
some single interface with useful methods that lets you deal with the 
actual score values easily.