Newport iServer MicroServer iTHX-M, Operator'S Manual

Page: 369 / 414

Appendix D | Bayesian Robust Linear Model with Mahalanobis (BRLMM) Distance Classifier Al-
The cluster centers, estimated using DM, are determined. Each
plotted point corresponds to an estimate of the cluster center for one
genotype in one SNP, with color indicating genotype (AA = red,
AB = blue, BB = green). The black x’s denote the grand mean of
cluster centers for each genotype, and the black ellipses are derived by
taking the average within-cluster variance and covariance – thus the
ellipses are to be interpreted as representative of the typical variance of
a cluster. The figure above
Figure D.2 plots the signal strength
log2(S
A
+
S
B
) against the allele contrast (S
A
−
S
B
)/(S
A
+
S
B
), and the
contrast is transformed by a transformation we call the Cluster–
Center–Stretch (CCS) transformation.
Figure D.3 above shows examples of the Cluster Center Stretch (CCS)
transformation for different settings of K. The CCS transformation is
defined as asinh(K*Contrast)/asinh(K) where Contrast is defined as
(S
A
−
S
B
)/(S
A
+
S
B
). The effect of the transformation is to stretch
contrast values near zero (corresponding to heterozygous genotypes)
and to compress contrast values near
−
1 and
+
1 (corresponding to
homozygous genotypes). The purpose of the CCS transform is to
Figure D.3
Cluster Center Stretch (CCS) Transformation
GTYPE_manual.book Page 361 Wednesday, August 2, 2006 11:31 AM