Newport iServer MicroServer iTHX-M Operator's Manual page 366

358
Affymetrix GeneChip ®
The first step in the workflow is to normalize the probe intensities and
calculate allele signal estimates for each SNP in each experiment. The
allele signal estimates are then transformed to a two-dimensional space
in which the underlying genotype clusters are well behaved in terms
of having similar variances for each of the clusters.
In parallel, an initial guess for each SNP's genotype using the DM
approach (with a confidence threshold set to 0.17 for high stringency)
is derived.
The threshold of 0.17 is user-definable. See
Settings (page 371).
Next a subset of SNPs is examined to identify cases where there are at
least a certain minimum number of examples of each of the three
genotypes according to the initial guesses. This subset of SNPs is used
to estimate a prior distribution on the typical cluster centers and
variance-covariance matrices. A default of 10,000 SNPs is used. To
adjust the number of SNPs needed to generate the prior, see BRLMM
Mapping Algorithm Settings
Each SNP is examined, and the cluster centers and variances implied
by the initial genotype guesses are combined with the prior
information in an ad hoc Bayesian procedure to derive a posterior
estimate of cluster centers and variances; it is principally this step that
distinguishes BRLMM from RLMM.
Finally, a genotype and a confidence score are assigned to each
observation according to its Mahalanobis distance from the three
cluster centers. The default confidence threshold of 0.5 is used to set
the confidence score threshold.
Genotyping Analysis Software User's Guide
page 371.
BRLMM Mapping Algorithm