A perspective on the Statistical Methods to Identify Loss
of Heterozygosity in Matched Samples of Normal and Lung Tumor Tissues
and LOH Gene Networks
Aldi Kraja
Division of Statistical Genomics
Washington University in St Louis
Friday, April 11, 2008, 12:30–1:30 pm
GEMS classroom, 3rd Floor in
Shriner's Building
Coffee, tea, and cookies will be provided
Abstract
The recurring of the loss of heterozygosity (LOH), which represents the genotype change from a heterozygote genotype in the normal tissues to a homozygote genotype in the tumor tissues of the same subject, has been used for a long time to identify tumor suppressor genes and oncogenes. High-density genotyping platforms expand our ability to identify LOH regions, but defining the LOH borders remains a challenge. We developed and assessed statistics to identify LOH regions in 357 matched normal and lung adenocarcinoma tissues from the same patients (Tumor Sequencing Project), with the Affymetrix 250K SNP chip. The informative LOH statistic (the ratio of LOH SNPs to informative ones); the LOH corrected for genotyping error (estimated from apparent "gain" of heterozygosity); and sliding windows of informative LOH averaged over a region (R(x)) were performed. P-values were calculated from bootstrap resampling and LOH block boundaries were estimated using a Hidden Markov Model. LOH regions were found on chromosomes 5q, 6q, 9p, 17p, 18q, and 19p. To compare methods, we simulated SNPs for 400 matched pairs, comparable to the TSP data, which showed that using large R(x) sliding windows lowers its sensitivity, especially for short significant LOH regions. In combination, these methods appear to improve LOH boundary estimation, which can help identify carcinogenesis genes. A new perspective is added to these analyses by reporting a gene network analyzes of the LOH genes.