Clustering autism phenotypes
Autism is a neurodevelopmental disorder that includes a range of linked conditions all of which provide for similar manifestations. This plurality of causes is quite evident in secondary autism where incriminated etiologies include genetic, infectious, metabolic, and/or environmental causes. Recognizing that autism is not a unitary disorder but a syndrome made of subgroups is of significance when establishing therapeutic and outcome goals for patients. In this study we analyzed the grouping of patients within a dataset of autism spectrum disorder (ASD) patients according to clinical manifestations in their Autism Diagnostic Interview- Revised (ADI-R) screening. In all, 113 patients, all males, 5-16 years of age, complied with inclusionary criteria. A Random Decision Forest was used on the training set in order to select variables of importance within our ADI-R dataset. The selected variables were studied for their similarities by 2 clustering methods: hierarchical clustering and DIANA. Three clusters were identified in this way suggesting the existence of subgroups within our ADS patient population. Using clustering analysis may provide for more homogenous grouping of patients when testing the benefits of different therapeutic interventions.