A QSAR - Bayesian neural network model to identify molecular properties causing eye irritation in cationic surfactants

QSAR models are frequently used to investigate and predict the toxicological effects of chemicals. Building QSAR models of the eye irritation potential of cationic surfactants is difficult, as the mechanism of action of these surfactants is still not fully understood. This report describes a data driven QSAR model to predict Maximum Average Scores (MAS in accordance to Draize) for cationic surfactants from the calculated molecular properties Log P, Log CMC and molecular volume, and the surfactant concentration. We demonstrate that a Bayesian Neural Network, a statistical non-linear regression approach that estimates the noise in the modelling data and error bars on the predictions, provided the most robust and accurate representation of the relationship between the MAS score and the molecular properties. A dataset of 20 in vivo rabbit eye irritation tests on 19 different cationic surfactants, obtained from historic in-house data and the scientific literature, was used to train the Bayesian neural network. The model was then used to simulate a large number of molecules to explore the relationship between MAS score and molecular properties. MAS vs. Log P showed bell shaped curve as expected. A higher concentration (> 20%) was required in order to elicit the eye irritancy response of molecules with a wide range of Log P. The simulated results were used to identify the range of molecular properties of cationic surfactants most likely to cause more than mild irritancy. Within the parameter space of the model, defined by the training data, the probability of causing severe irritation is highest for molecules with molecular volume < 320 Ao , while –2 < Log P <13 and –6 < Log CMC < 3. The simulated results were carried out at a concentration of 40%. For molecules with larger molecular volumes, the range of Log P and Log CMC for which these molecules would cause severe irritation i s narrowed. The model provides useful probabilistic predictions for the eye irritancy potential of new or untested cationic surfactants with physicochemical properties lying within the parameter space of the model. Introduction Eye irritancy potential in vivo is still based on the method described by Draize et al (1944). Chemical use solely on the basis of in vitro tests remains generally unacceptable (Speilmann et al, 1996) and there are neither complete replacements for the Draize test nor strategies which completely avoid the use of animals. Computer modelling based on a small number of carefully selected experiments can prove helpful in extending the knowledge domain whilst limiting the number of animals required for experimentation. QSARs are often hindered by a lack of quality in vivo data and sufficient understanding of the mechanisms of action (Cronin et al, 1995). Understanding of the eye irritation potential of chemicals is complex. Here we demonstrate that the Bayesian neural network (BNN) produces a robust model due to its capability of predicting noise in the experimental data and providing error bars on the predictions. We use this model to simulate a large number of molecular properties for a range of cationic surfactants and predict their corresponding MAS scores. We then identify and isolate the region of the molecular space that is most likely to result in severe eye irritation.