Classification of cancers based on copy number variation landscapes

• Published in: Biochimica et biophysica acta Vol. 1860; no. 11; pp. 2750 - 2755
• Main Authors: Zhang, Ning, Wang, Meng, Zhang, Peiwei, Huang, Tao
• Format: Journal Article
• Language: English
• Published: Netherlands Elsevier B.V 01.11.2016
• Subjects: Cancer type; Chromosomes - genetics; DNA; DNA copy number variants; DNA Copy Number Variations - genetics; Feature selection; genes; Genomics - methods; Humans; IFS; landscapes; mutation; Mutation - genetics; neoplasms; Neoplasms - genetics; Sequential minimal optimization; Cancer type; DNA copy number variants; Feature selection; Sequential minimal optimization; IFS
• ISSN: 0304-4165; 0006-3002; 1872-8006
• DOI: 10.1016/j.bbagen.2016.06.003

Genomic alterations in DNA can cause human cancer. DNA copy number variants (CNV), as one of the types of DNA mutations, have been considered to be associated with various human cancers. CNVs vary in size from 1bp up to one complete chromosome arm. In order to understand the difference between different human cancers on CNVs, in this study, we developed a method to computationally classify six human cancer types by using only CNV level values. The CNVs of 23,082 genes were used as features to construct the classifier. Then the features are carefully selected by mRMR (minimum Redundancy Maximum Relevance Feature Selection) and IFS (Incremental Feature Selection) methods. An accuracy of over 0.75 was reached by using only the CNVs of 19 genes based on Dagging method in 10-fold cross validation. It was indicated that these 19 genes may play important roles in differentiating cancer types. We also analyzed the biological functions of several top genes within the 19 gene list. The statistical results and biological analysis of these genes from this work might further help understand different human cancer types and provide guidance for related validation experiments. This article is part of a Special Issue entitled "System Genetics" Guest Editor: Dr. Yudong Cai and Dr. Tao Huang. •The authors did a comprehensive genome wide analysis of 23,082 CNVs in 3480 cancer patients from six cancer types.•19 discriminative genes among cancer types were identified and their prediction accuracy was over 0.75.