The stu
dy an
d pre
diction of kinase function (kinomics) is of major importance for proteome research
dueto the wi
desprea
d distribution of kinases. However, thepre
diction of protein function base
d on the similaritybetween a functionally annotate
d 3D template an
d a querystructure may fail, for instance, if a similar proteinstructure cannot be i
dentifie
d. Alternatively, function canbe assigne
d using 3D-structural empirical parameters. Inprevious stu
dies, we intro
duce
d parameters base
d onelectrostatic entropy (
Proteins 2004,
56, 715) an
d molecular vibration entropy (
Bioinformatics 2003,
19, 2079) butignore
d other important factors such as van
der Waals(v
dw) interactions. In the work
describe
d here, we
define3D-v
dw entropies (
deg.gif">
k) an
d use them for the first time to
derive a classifier for protein kinases. The mo
del classifiescorrectly 88.0% of proteins in training an
d more than85.0% of proteins in vali
dation stu
dies. Principal components analysis of heterogeneous proteins
demonstrate
dthat
deg.gif">
k co
dify information that is
different to that
describe
d by other bulk or fol
ding parameters. In a
dditionalvali
dation experiments, the mo
del recognize
d 129 out of142 kinases (90.8%) an
d 592 out of 677 non-kinases(87.4%) not use
d above. This stu
dy provi
des a basis forfurther consi
deration of
deg.gif">
k as parameters for the empirical search for structure-function relationships.