Hello everyone,
I'm currently re-working my dipole coupling cross-validation methods,
and was wondering if I can treat my "total" R-work and "total" R-free as
weighted summations of the individual R-work/free for ea. individual
coupling type?
For example, I take a 90% working set (consisting of 5 dipole types) and
a 10% test set, use normal force constants for my working dipole
classes, zero for the test classes (also not calling these until the
final analysis) and report the R-work and R-free for each coupling
type. To obtain the "total" R-work/free I do the following (after
evaluation of the rms and violations for each coupling as usual):
{ TOTAL WORKING R-FACTOR }
evaluate ($RJNHw=$rmsJNHw*100*50/13.008) {* make sure only the num
dipoles in the working set are entered *}
evaluate ($RJNCw=$rmsJNCw*100*44/13.008)
evaluate ($RJHCw=$rmsJHCw*100*44/13.008)
evaluate ($RJCCw=$rmsJCCw*100*44/13.008)
evaluate ($RJCHw=$rmsJCHw*100*51/13.008)
evaluate
($rms_sani_ALLw=($rmsJNHw*50+$rmsJNCw*44+$rmsJHCw*44+$rmsJCCw*44+$rmsJCHw*51)/233)
evaluate ($R_sani_ALLw=($RJNHw+$RJNCw+$RJHCw+$RJCCw+$RJCHw)/233)
{ TOTAL FREE R-FACTOR }
evaluate ($RJNHf=$rmsJNHf*100*6/13.008) {* make sure only the num
dipoles in the test set are entered *}
evaluate ($RJNCf=$rmsJNCf*100*5/13.008)
evaluate ($RJHCf=$rmsJHCf*100*5/13.008)
evaluate ($RJCCf=$rmsJCCf*100*5/13.008)
evaluate ($RJCHf=$rmsJCHf*100*6/13.008)
evaluate
($rms_sani_ALLf=($rmsJNHf*6+$rmsJNCf*5+$rmsJHCf*5+$rmsJCCf*5+$rmsJCHf*6)/27)
evaluate ($R_sani_ALLf=($RJNHf+$RJNCf+$RJHCf+$RJCCf+$RJCHf)/27)
--
where the subscripts w and f refer to working and free respectively.
The lines:
evaluate
($rms_sani_ALLw=($rmsJNHw*50+$rmsJNCw*44+$rmsJHCw*44+$rmsJCCw*44+$rmsJCHw*51)/233)
evaluate ($R_sani_ALLw=($RJNHw+$RJNCw+$RJHCw+$RJCCw+$RJCHw)/233)
--
are where I have weighted and summed the terms, and I just wanted to see
If you guys think this is ok? To report R-free and R-work in the same
script I found I can't use the 'ALL' class to generate an R-work, as
this then incorporates the classes in the test set too.
I have not combined my classes into a single working set and a single
test set as I need to scale the force constants for each dipole type
differently.
I think the weighted summation is ok, but if someone can validate this
for me I would be most grateful.
Best regards,
Haydyn
--
-----------------------
Haydyn Mertens
PhD student
Dept. Biochemistry
University of Melbourne
Australia
-----------------------
