TEGS-CN: A Statistical Method for Pathway Analysis of Genome-wide Copy Number Profile

The effects of copy number alterations make up a significant part of the tumor genome profile, but pathway analyses of these alterations are still not well established. We proposed a novel method to analyze multiple copy numbers of genes within a pathway, termed Test for the Effect of a Gene Set with Copy Number data (TEGS-CN). TEGS-CN was adapted from TEGS, a method that we previously developed for gene expression data using a variance component score test. With additional development, we extend the method to analyze DNA copy number data, accounting for different sizes and thus various numbers of copy number probes in genes. The test statistic follows a mixture of X(2) distributions that can be obtained using permutation with scaled X(2) approximation. We conducted simulation studies to evaluate the size and the power of TEGS-CN and to compare its performance with TEGS. We analyzed a genome-wide copy number data from 264 patients of non-small-cell lung cancer. With the Molecular Signatures Database (MSigDB) pathway database, the genome-wide copy number data can be classified into 1814 biological pathways or gene sets. We investigated associations of the copy number profile of the 1814 gene sets with pack-years of cigarette smoking. Our analysis revealed five pathways with significant P values after Bonferroni adjustment (<2.8 × 10(−5)), including the PTEN pathway (7.8 × 10(−7)), the gene set up-regulated under heat shock (3.6 × 10(−6)), the gene sets involved in the immune profile for rejection of kidney transplantation (9.2 × 10(−6)) and for transcriptional control of leukocytes (2.2 × 10(−5)), and the ganglioside biosynthesis pathway (2.7 × 10(−5)). In conclusion, we present a new method for pathway analyses of copy number data, and causal mechanisms of the five pathways require further study.