Using rule induction for prediction of self-injuring behavior in animal models of development disabilities

The data mining system LERS (Learning from Examples using Rough Sets) was used to assess whether animal models of varying basal ganglia dopamine concentrations could be distinguished based on their behavioral responsiveness to a dopamine agonist, GBR12909. GBR12909 causes its agonist effects by increasing synaptic concentrations of dopamine. The three animal models included rats depleted as neonates of striatal dopamine, rats with hyper-innervation of striatal dopamine and control rats with normal striatal dopamine concentrations. The groups received five injections of GBR12909 and were observed for stereotyped and self-injurious behaviors immediately following the injections and six hours after injections. The data mining system LERS induced rules that indicated which of the injections caused several behaviors to be exhibited and which injections caused more focused behaviors. Prediction error rate analysis enable us to determine whether the pattern of behaviors displayed following GBR12909 administration could be distinguished among animal models. Differences in the rule sets formed for each group for each injection enables the prediction of the stereotyped behaviors that may occur prior to occurrence of self-injurious behavior. The ability to predict the occurrence of self-injurious behaviors in the animal models greatly increases our change of suppressing these behaviors through behavioral or pharmacological intervention