Prediction of a patient´s response to a specific drug treatment using artificial neural networks

We demonstrate the ability of artificial neural networks (ANNs) in predicting the response of patients suffering from a specific disease or disorder to a specific drug. As a case study, we show that ANNs can be used to predict sickle cell anemia patients´ response to hydroxyurea treatment. Hydroxyurea is an orally delivered medication that partially alleviates the symptoms of sickle cell anemia. The studies described were undertaken to develop the ability to identify those patients who will not benefit sufficiently from hydroxyurea to warrant the risk of its deleterious side effects by the use of neural networks. The trained artificial neural networks were capable of predicting the potential response of a patient to hydroxyurea with 86% and as high as 100% accuracy, depending on the definition of a positive response. This prediction was achieved by training the network with the 23 collected parameters. Furthermore, if was possible to reduce the 23 input dimensional space into only 8 by performing variable selection