Format

Send to

Choose Destination
PLoS Comput Biol. 2014 Aug 28;10(8):e1003765. doi: 10.1371/journal.pcbi.1003765. eCollection 2014 Aug.

Identification of allelic imbalance with a statistical model for subtle genomic mosaicism.

Author information

1
Department of Epidemiology, The University of Texas MD Anderson Cancer Center, Houston, Texas, United States of America; Division of Biostatistics, The University of Texas School of Public Health, Houston, Texas, United States of America.
2
Department of Epidemiology, The University of Texas MD Anderson Cancer Center, Houston, Texas, United States of America; Human & Molecular Genetics Program, The University of Texas Graduate School of Biomedical Sciences, Houston, Texas, United States of America.
3
Department of Epidemiology, The University of Texas MD Anderson Cancer Center, Houston, Texas, United States of America; Division of Biostatistics, The University of Texas School of Public Health, Houston, Texas, United States of America; Human & Molecular Genetics Program, The University of Texas Graduate School of Biomedical Sciences, Houston, Texas, United States of America.

Abstract

Genetic heterogeneity in a mixed sample of tumor and normal DNA can confound characterization of the tumor genome. Numerous computational methods have been proposed to detect aberrations in DNA samples from tumor and normal tissue mixtures. Most of these require tumor purities to be at least 10-15%. Here, we present a statistical model to capture information, contained in the individual's germline haplotypes, about expected patterns in the B allele frequencies from SNP microarrays while fully modeling their magnitude, the first such model for SNP microarray data. Our model consists of a pair of hidden Markov models--one for the germline and one for the tumor genome--which, conditional on the observed array data and patterns of population haplotype variation, have a dependence structure induced by the relative imbalance of an individual's inherited haplotypes. Together, these hidden Markov models offer a powerful approach for dealing with mixtures of DNA where the main component represents the germline, thus suggesting natural applications for the characterization of primary clones when stromal contamination is extremely high, and for identifying lesions in rare subclones of a tumor when tumor purity is sufficient to characterize the primary lesions. Our joint model for germline haplotypes and acquired DNA aberration is flexible, allowing a large number of chromosomal alterations, including balanced and imbalanced losses and gains, copy-neutral loss-of-heterozygosity (LOH) and tetraploidy. We found our model (which we term J-LOH) to be superior for localizing rare aberrations in a simulated 3% mixture sample. More generally, our model provides a framework for full integration of the germline and tumor genomes to deal more effectively with missing or uncertain features, and thus extract maximal information from difficult scenarios where existing methods fail.

PMID:
25166618
PMCID:
PMC4148184
DOI:
10.1371/journal.pcbi.1003765
[Indexed for MEDLINE]
Free PMC Article

Supplemental Content

Full text links

Icon for Public Library of Science Icon for PubMed Central
Loading ...
Support Center