crlmm (version 1.30.0)

predictionRegion: Prediction regions for integer copy number

Description

Bivariate normal prediction regions for integer copy number. Copy numbers 0-4 allowed.

Usage

predictionRegion(object, copyNumber)

Arguments

object

A CNSet object.

copyNumber

Integer vector. 0-4 allowed.

Value

A list named by the genotype. `NULL' refers to copy number zero, `A' is a hemizygous deletion, etc. Each element is a list of the means (mu) and covariance (cov) for each marker stored as an array. For `mu', the dimensions of the array are marker x allele (A or B) x batch. For `cov', the dimensions of the array are marker x 3 (varA, cor, and varB) x batch.

Details

We fit a linear regression for each allele to the diallic genotype cluster medians. Denoting the background and slope by nu and phi, respectively, the mean for the bivariate normal prediction region is given by

mu_A = nu_A + CA * phi_A

and

mu_B nu_B + CB * phi_B

The variance and correlation of the normalized intensities is estimated from the diallelic genotype clusters AA, AB, and BB on the log-scale. For copy number not equal to two, we assume that the variance is approximately the same for copy number not equal to 2.

References

Scharpf et al., 2011, Biostatistics.

See Also

posteriorProbability, genotypes

Examples

data(cnSetExample)
pr <- predictionRegion(cnSetExample, copyNumber=0:4)
names(pr)
## bivariate normal prediction region for NULL genotype (homozygous deletion)
str(pr[["NULL"]])