|Grant Number:||5R03CA105956-02 Interpret this number|
|Primary Investigator:||Betensky, Rebecca|
|Project Title:||Statistical Methods for Analysis of Quantitative Loh|
DESCRIPTION (provided by applicant): Loss of heterozygosity (LOH) of chromosomal regions of tumors is of interest as it is suggestive of the presence of a tumor suppressor gene. Allelic losses on chromosomes 1p and 19q have been found frequently in oligodendrogliomas. Further, LOH on chromosomes 1p and 19q is of prognostic interest, as it has been shown to be highly associated with response to chemotherapy and long survival in patients with gliomas. Previous qualitative analyses of LOH in oligodendroglioma used three or four distally-located CA-repeat polymorphism markers to assess LOH. Loss at a marker was inferred if the ratio of the allele ratio in a normal sample to that in a tumor sample exceeded a fixed threshold. The tumor was then scored as LOH if LOH was observed at all informative markers. Recently, a "medium throughput" quantitative method for assessing LOH at several markers was applied to gliomas by Dr. Bogler. This methodology raises the possibility of a refined LOH-based classification scheme for brain tumors that may be even more predictive of clinical outcomes than the qualitative, lower throughput LOH methodology. The primary goal of this proposal is to investigate optimal ways of using the multiple quantitative allele ratios to predict clinical outcomes, and thereby develop an LOH-based classification scheme. This goal coincides with priorities for the detection and diagnosis of brain tumors identified by the NCI/NINDS sponsored Brain Tumor Progress Review Group: "to develop a molecular-based classification scheme for brain tumors that can be used to predict tumor behavior." The methodology developed here is not limited to LOH experiments, but could be applied also to array CGH experiments. A secondary goal of this proposal is to identify and apply unbiased tests for linkage between marker loci and a gene with functional significance.
Supervised Bayesian latent class models for high-dimensional data.
Authors: Desantis SM, Houseman EA, Coull BA, Nutt CL, Betensky RA
Source: Stat Med, 2012 Jun 15;31(13), p. 1342-60.
EPub date: 2012 Apr 11.
Testing goodness of fit of a uniform truncation model.
Authors: Mandel M, Betensky RA
Source: Biometrics, 2007 Jun;63(2), p. 405-12.
Feature-specific penalized latent class analysis for genomic data.
Authors: Houseman EA, Coull BA, Betensky RA
Source: Biometrics, 2006 Dec;62(4), p. 1062-70.