The ever increasing number of experimentally resolved crystal structures su
pports the possibility of fully empirical crystal structure prediction for
small organic molecules. Empirical methods promise to be significantly more
efficient than methods that attempt to solve the same problem from first p
rinciples. However, the transformation from data to empirical knowledge and
further to functional algorithms is not trivial and the usefulness of the
result depends strongly on the quantity and the quality of the data. In thi
s work, a simple scoring function is parameterized to discriminate between
the correct structure and a set of decoys for a large number of different m
olecular systems. The method is fully automatic and has the advantage that
the complete scoring function is parametrized at once, leading to a self-co
nsistent set of parameters. The obtained scoring function is tested on an i
ndependent set of crystal structures taken from the P1 and P <(1)overbar>1
space groups. With the trained scoring function and FlexCryst, a program fo
r small-molecule crystal structure prediction, it is shown that approximate
ly 73% of the 239 tested molecules in space group P1 are predicted correctl
y. For the more complex space group P <(1)overbar> 1, the success rate is 2
6%. Comparison with force-field potentials indicates the physical content o
f the obtained scoring function, a result of direct importance for protein
threading where such database-based potentials are being applied.