The biological activity measured experimentally for a series of molecules w
as used to create a QSAR model using the parameters calculated with the sem
i-empirical method AM1 and the molecular volumes. The statistical methods,
such as partial least squares, pattern recognition techniques and principal
component analysis, were used to work with a large amount of data and esta
blish QSAR. Cross-validation was used to test the predictive capability of
the model. The analysis of the data allows one to draw some conclusions abo
ut the importance of some molecular parameters. The results can provide inf
ormation about how to model better drugs. (C) 2001 Elsevier Science B.V. Al
l rights reserved.