Classification techniques for recurrent DNA copy number data

Genetic instabilities in the human genome are frequently exhibited in the form of DNA copy number variations. In this paper, we propose a framework to evaluate the predictive power of recurrent copy number variations at multiple genomic sites for detecting genetic diseases. Moreover, we compare the ability of two well known clustering algorithms, k-means and Fuzzy c-means, to correctly classify diseased samples with respect to the control samples. Finally, we apply our proposed techniques on 51 samples of 25 apparently healthy and 26 autistic children. Our results show that using CNVs at multiple sites will considerably increase classification performance when compared with the traditional classifiers that focus on a single CNV.