Predictive value of recurrent DNA copy number variations

Recurrent copy number variations across multiple samples are increasingly used to identify the genes and the genomic locations that are statistically and biologically significant and correlated with certain diseases. In this paper, we evaluate the predictive power of copy number variations for detecting autism. We consider both recurrent copy number variations at one location and correlated recurrent copy number variations at multiple locations. In each case, we compare the ability of k-means and Fuzzy c-means algorithms to correctly classify autistic samples. Finally, we apply our proposed techniques on 51 samples of 25 apparently healthy and 26 autistic children.