jakesylvestre / cing

Automatically exported from code.google.com/p/cing
0 stars 0 forks source link

Residue ROG-score calculation: valSets.cfg parameters #299

Closed GoogleCodeExporter closed 9 years ago

GoogleCodeExporter commented 9 years ago
The code for '_criticizeResidue' in 'validate.py'
(http://www.google.com/codesearch#uJdEl-4UXoc/trunk/cing/python/cing/core/valida
te.py&q=validate.py%20package:http://cing\.googlecode\.com&l=446) suggests that 
the what_check scores for 'BBCCHK', 'C12CHK', 'RAMCHK' are only compared to a 
negative threshold (please note the 'smaller than' characters '<':

            if actualValue < thresholdValueBad: # assuming Z score
                comment = 'whatif %s value %s <%8.3f' % (key, actualValueStr, thresholdValueBad)
#                NTdebug(comment)
                residue.rogScore.setMaxColor( COLOR_RED, comment)
            elif actualValue < thresholdValuePoor:
                comment = 'whatif %s value %s <%8.3f' % (key, actualValueStr, thresholdValuePoor)
#                NTdebug(comment)
                residue.rogScore.setMaxColor( COLOR_ORANGE, comment)
            #endif

thresholdValueBad and thresholdValuePoor are defined by:

            thresholdValuePoor = valSets[ 'WI_' + key + '_POOR'  ]
            thresholdValueBad = valSets[ 'WI_' + key + '_BAD' ]

valSets.cfg seems to contain the actual values
(http://www.google.com/codesearch#uJdEl-4UXoc/trunk/cing/python/cing/valSets.cfg
&q=valsets%20package:http://cing\.googlecode\.com):

# The naming of the What If settings encodes their meaning.
WI_RAMCHK_POOR =  -1.0
WI_RAMCHK_BAD  =  -1.3
# This has become too restrictive in 2010 Whatif version
#WI_BBCCHK_POOR =  10.0
#WI_BBCCHK_BAD  =  3.0
WI_BBCCHK_POOR =  -999.9
WI_BBCCHK_BAD  =  -999.9
WI_C12CHK_POOR =  -0.9
WI_C12CHK_BAD  =  -1.2

There seem to be only negative values. Worse, for BBCCHK, the threshold seems 
to be -999.9 which I suppose is some kind of weird what_if internal standard.

Original issue reported on code.google.com by robinvan...@gmail.com on 11 Jul 2011 at 10:13

GoogleCodeExporter commented 9 years ago
Sorry Jurgen, but for some reason I cannot assign tags and other things to the 
issue report.

Original comment by robinvan...@gmail.com on 11 Jul 2011 at 10:14

GoogleCodeExporter commented 9 years ago
The BBCCHK is no longer used for coloring residues by CING since the WI 
database become of too high quality. The disablement is done by setting it to a 
real low value such as -999.9 although -1 would have been enough since the 
range of expected values is [0,80].

I don't understand your other remarks.

Original comment by jurge...@gmail.com on 11 Jul 2011 at 12:56

GoogleCodeExporter commented 9 years ago
The what-if checks return z-scores. In order to select residues that are within 
a couple of sigma of the mean, one would need to discard residues with z-scores 
lower than a negative threshold, and residues with z-scores higher than a 
positive threshold.

In other words: valSets.cfg only contains NEGATIVE values for 
WI_RAMCHK_POOR/BAD and WI_C12CHK_POOR/BAD. I am wondering whether the ROG-score 
calculation in validate.py only takes into account values that are LOWER than 
the NEGATIVE threshold
(actualValue < thresholdValuePoor/Bad), and not values that are HIGHER than the 
POSITIVE threshold.

I hope that will make it clear to you.

Original comment by robinvan...@gmail.com on 12 Jul 2011 at 1:11

GoogleCodeExporter commented 9 years ago
Ah, so you propose to include a limit on the upper values?
Now I understand. We've never had that request.

Original comment by jurge...@gmail.com on 12 Jul 2011 at 2:54

GoogleCodeExporter commented 9 years ago
Yes, in our opinion +4 sigma is as bad as -4 sigma.

Original comment by robinvan...@gmail.com on 12 Jul 2011 at 3:21