Ligand bia
s can contribute
significantly to the number of fal
se po
sitive
s ob
served in a virtual
screeningcampaign. U
sing a receptor-ba
sed docking approach again
st a well-e
stabli
shed target of therapeutic importance,e
strogen receptor
![](/image<font color=)
s/gifchar
s/alpha.gif" BORDER=0> (ER
![](/image<font color=)
s/gifchar
s/alpha.gif" BORDER=0>), coupled with
several common
scoring function
s (ChemGua
ss, ChemGau
ss2,ChemScore, ScreenScore, ShapeGau
ss, and PLP), taken both individually and a
s a con
sen
su
s, we
sought toexamine the characteri
stic
s of molecule
s retrieved by each. It ha
s been previou
sly
shown that
scoring function
s(mainly empirical) exhibit bia
s in prioritizing more complicated molecule
s ari
sing from additive component
swithin the function. U
sing Spearmen'
s correlation coefficient, we
show that a large
set of de
scriptor
s calculatedfor a docked
set of molecule
s exhibit po
sitive correlation with the ranked po
sition in a hitli
st. Moreover,mo
st of the
se de
scriptor
s correlate well with MW. To thi
s end, rather than penalizing the docked
score ofall heavy molecular weight (MW) molecule
s and rewarding tho
se of lower MW, a
s i
s common practice, weexamine the impact of penalizing the
score only of tho
se molecule
s which were of higher MW, leavinglower MW molecule
s unaffected. Here, we introduce a new power function to aid the proce
ss. U
sing
scoringfrequency analy
si
s and SIFt fingerprint
s, we acheived a more meaningful analy
si
s of virtual
screening (VS)performance than with enrichment calculation
s, facilitating target-
specific VS method development.