Rearranging Computational Operations Greatly Improves the Accuracy of Genomic Copy Number Analysis of Tumors

• Activity Number: 301
• Type: Contributed
• Date/Time: Tuesday, August 2, 2011 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Stan Pounds*+ and Arzu Onar-Thomas and Cuilan Lani Gao
• Companies: St. Jude Children's Research Hospital and St. Jude Children's Research Hospital and St. Jude Children's Research Hospital
• Keywords: genomics ; cancer ; change-point ; microarray ; sequencing ; normalization

Abstract

In genomic copy number analysis of tumors, a series of computational data analysis operations are performed to address a set of statistical problems. A "typical" method (1) normalizes signals of tumors and controls, (2) compares normalized signals of tumors and controls, (3) identifies copy number change-point loci, and (4) infers the copy number status of genomic segments. By systematically describing the informative biological and statistical relationships between operations, we developed a new method that (1) compares unnormalized signals of tumors and controls, (2) iteratively normalizes signal differences and infers change-point loci, (3) identifies two-copy genomic segments, and (4) infers the copy number status of genomic segments. For a leukemia data set, the new method's results show 95-99% agreement with cytogenetic validation data for tumors with simple or complex genomes. In contrast, the "typical" approach shows 85-95% and 10% agreement for tumors with simple and complex genomes, respectively. Our results show that major gains in statistical accuracy may be achieved by rearranging the "typical" order of computational operations in the analysis of genomic data.