Special Cases - Newport iServer MicroServer iTHX-M Operator's Manual

Special Cases

Appendix D | Bayesian Robust Linear Model with Mahalanobis (BRLMM) Distance Classifier Al-
Again, this has the intuitively sensible property that when there is
little or no labeled data available, the estimate of cluster centers will
be driven mainly by the prior estimate m. When there is alot of data
available for a given genotype, the estimate will be driven by v. Loosely
speaking, this update rule has the form of a weighted average of the
prior and observed data, with the following two specifications:
• The prior has weight inversely proportional to M, which is the
variance-covariance matrix of cluster centers in the SNPs used to
build the prior.
• The observed data for the particular SNP has weight inversely
proportional to NS, which is a product of the number of
observations and the variance in the observed data.
With these posterior estimates of center and spread for each cluster,
genotypes and confidences are then determined as outlined in the
previous section.
The preceding algorithm assumes that the observations for each SNP
are well described by prototypes for each genotype. However, for SNPs
on the X chromosome, there are distinct clusters for each gender due
to males having one fewer copy of the X chromosome. This not only
changes the location of the cluster centers for XY individuals, but the
SNPs located on chrX may end up being called heterozygote.
Therefore the chrX SNPs are treated differently for XX individuals
than for XY individuals. Note that the special treatment of chrX SNPs
described here is only applied to SNPs on chrX in the non-
pseudoautosomal region. For the rest of this section, chrX should be
interpreted as chrX excluding the pseudoautosomal region.
The difference between XY and XX individuals is detected by the seed
calls from DM. XY individuals are estimated as those having
heterozygosity less than 7.5% on chrX. The remaining individuals are
classified as XX. For each chrX SNP, XX individuals and XY
individuals are treated as separate data sets.