GPHMM

GPHMM models chromosomal aberrations and genotyping signal biases in whole genome SNP array data from tumor samples to detect copy number alterations and loss of heterozygosity (LOH).

Key Features:

• Global parameter integration: Incorporates global parameters into a Hidden Markov Model to account for sample-wide effects on SNP array signals.
• Hidden Markov Model framework: Uses an HMM to segment genomic states relevant to copy number and LOH.
• Baseline shift modeling: Quantitatively models signal baseline shifts arising from aneuploidy.
• Normal cell contamination modeling: Explicitly models contamination from normal cells (tumor purity effects) in genotyping signals.
• GC content bias correction: Accounts for GC content bias that distorts SNP array signal intensities.
• Expectation-Maximization estimation: Employs an Expectation-Maximization (EM) algorithm for parameter estimation.
• Low-purity sensitivity: Capable of identifying chromosomal rearrangements in samples with tumor cell content as low as 10%.
• SNP array compatibility: Operates on whole genome SNP array genotyping data for genome-wide analysis.
• Quality-control outputs: Produces global parameter estimates useful for data quality control and outlier detection in cohorts.

Scientific Applications:

• Chromosomal aberration detection: Detection and characterization of chromosomal rearrangements and copy number alterations from SNP arrays.
• LOH analysis: Identification and mapping of loss of heterozygosity regions in tumor genomes.
• Tumor purity and aneuploidy assessment: Estimation of tumor cell content and aneuploidy-related baseline shifts from genotyping signals.
• Cohort data quality control: Quality-control assessment and outlier detection in SNP array cohort studies using global parameter estimates.

Methodology:

Integration of global parameters into a Hidden Markov Model and parameter estimation via an Expectation-Maximization (EM) algorithm to quantitatively model baseline shifts (aneuploidy), normal-cell contamination, and GC content bias and to identify chromosomal rearrangements from whole genome SNP array genotyping signals.