paperKB
coga / coga-kb
Processing
Help
Sign in

Chunk #11 — MATERIALS AND METHODS — Statistical model — QuantiSNP: an Objective Bayes Hidden-Markov Model — Emission probabilities

Source
QuantiSNP: an Objective Bayes Hidden-Markov Model to detect and accurately map copy number variation using SNP genotyping data.
Embedded
yes

Text

Let r denote the log R ratio and b the B allele frequencies. These values are assumed to be independent given the hidden state z and all other model parameters θ. The emission probabilities are defined as a mixture of Gaussian and uniform distributions, 3 4 where Rmax = 6 is defined by the lowest observed value for the intensity R. The uniform distribution in each case acts as a non-informative state for capturing outliers in the data. As the B allele frequencies are in the range 0 < b < 1, we use half-normal distributions (G+) for the homozygous genotypes with fixed location parameters (0 and 1, respectively). The EM updating of the variance parameters is then the same as with the full normal distribution.