SAR model development should be a continuous process involving formulation
then experimental testing of the model, incorporation of test results into
the database, then refinement of the model using the expanded database. The
larger database affords greater confidence in its ability to predict the b
iological response. This iterative procedure was employed with a recently d
eveloped structure-activity relationship (SAR) model of human skin irritati
on. Based on a ''leave-one-out'' cross validation, the mean sensitivity of
the initial model was 0.89, the mean specificity was 0.74. A clinical valid
ation study was conducted to assess the ability of the model to predict hum
an skin irritation by esters commonly used as fragrance ingredients. Esters
that were found to cause irritation in rabbits, and that were within the p
redictive space of the SAR model, were selected for human testing using the
patch test procedure. Of the 34 rabbit irritants selected, 16 were predict
ed by the model to be positive and is were predicted to be negative. Patch
testing yielded two positive esters, allyl heptanoate and allyl cyclohexane
propionate. These test results were incorporated into the database to refin
e the SAR model. Best subsets regression and linear discriminant analysis w
ere used to generate 10 submodels consisting of 10 irritants and 50 non-irr
itants randomly selected from the new database. Physicochemical parameters
associated with irritant esters, when compared with non-irritant esters, di
ffered somewhat from those identified in the original model. Irritant ester
s had lower solubility parameter and water solubility, higher Hansen disper
sion and Hansen hydrogen bonding, and lower sum of partial positive charges
, when compared with non-irritant esters. The sensitivity of the new model
is 0.69 and specificity is 0.67. The results of this study indicate that SA
R models based on limited data may not accurately predict the activity of u
nknown chemicals even though the computationally-derived sensitivity and sp
ecificity of the models are high. This finding emphasizes the need for expe
rimental validation of models and their refinement as new data become avail
able.