Brlmm Mapping Algorithm Settings - Newport iServer MicroServer iTHX-M Operator's Manual

BRLMM Mapping Algorithm Settings

Appendix D | Bayesian Robust Linear Model with Mahalanobis (BRLMM) Distance Classifier Al-
minimum number of samples required is six. However, performing a
run with only six samples is not advised.
Another consideration is the extent to which data sets can be
combined. On the one hand, this should help in terms of increasing
the number of observations, particularly for rare genotypes, thereby
improving the performance on rarer genotypes. On the other hand, the
validity of combining data sets depends on the degree to which the
combined data sets have the same underlying probe intensity
distribution, probe effects, cluster centers and cluster variances. A
combination of data sets from different labs can change performance
slightly in either direction, and understanding the criteria under
which it will and won't succeed remains an area of future work.
The BRLMM mapping algorithm incorporates user modifiable
settings which influence call rate and call accuracy. The default
settings are set to allow high call rates with better than 99% accuracy
when analyzing samples which meet the 93% call rate cutoff value
using DM at 0.33. Care should be taken when changing these settings
as adjusting the parameters to increase call rates may decrease the
accuracy. Table D.2 on page 373
changing a value affects the analysis.
To view the user-modifiable algorithm settings, click the
1.
Mapping Algorithm Settings button
bar; or
Select Tools
bar.
The Mapping Algorithm Settings dialog box opens
explains the settings and how
Mapping Algorithm Settings from the menu
in the Settings shortcut
(Figure D.5).