Abstract
QSAR models of log
BB for a diverse set of 113 drug molecules were developed employing structural descriptors using CODESSA-PRO and ISIDA approaches. The models were successfully validated using the central nervous system activity data of an external test set of 40 drug molecules.
Experimental blood–brain partition coefficients (log
BB) for a diverse set of 113 drug molecules are correlated with computed structural descriptors using CODESSA-PRO and ISIDA programs to give statistically significant QSAR models based respectively, on molecular and on fragment descriptors. The linear correlation CODESSA-PRO five-descriptor model has correlation coefficient
R
2
=
0.781 and standard deviation
s
2
=
0.123. The ‘consensus model’ of ISIDA gave
R
2
=
0.872 and
s
2
=
0.047. The developed models were successfully validated using the central nervous system activity data of an external test set of 40 drug molecules.