[Bioperl-l] SiteMatrix changes
bix at sendu.me.uk
Fri Sep 1 12:25:57 UTC 2006
Sendu Bala wrote:
> skirov wrote:
>> Sounds OK with me. To summarize:
>> 1. Correction is disabled by default.
>> 2. Correction should be applied to all positions.
>> 3. Thresholds for IUPAC consensus can be user defined.
>> 4. A fix for IUPAC consensus calculation: change the defaukt behavior.
>> 5. Document the options
>> Does this sounds right?
> Yes, sounds good to me. I'll code those up shortly.
This is now done. I didn't quite do it the way you suggested: the
'threshold' for IUPAC consensus is implemented as significance level for
rounding the frequencies. This way we don't have to suffer some
arbitrary cutoff that does unexpected things. You may also want to check
the changes to the documentation (mostly in new() and the description)
to make sure I understood and explained everything well enough.
The consensus string now also enforces the supplied or default
threshold, and treats the threshold the way most people might think of
such a thing - the minimum acceptable value (inclusive). This doesn't
seem to actually change the answer for the few test matrices used by the
test scripts (though the test script answers have changed since we're
not doing pseudo-count correction anymore).
One issue is that there's no way for the user to decide to do
pseudo-count correction or not when using the PSM::IO modules. The
correction should probably be farmed out to a separate method. I don't
plan to do this myself.
More information about the Bioperl-l