Biology Reference
In-Depth Information
Mean
Normalized
Expression
Mean
Normalized
Expression
0.2
0.0
0.0
5
10
15
0
4
8
-0.2
-0.2
-0.4
12
-0.4
-0.6
-0.6
-0.8
-0.8
1.0
1.5 2.0 2.5
Microarray standard deviation
3.0
3.5
4.0
0.25
0.5
1
2
Dispersion
4
8
16
FIGURE 5.2 Association between cross-platform correlation and the standard deviation / dispersion. These plots
show that expression level variation across all groups is an important factor for predicting cross-platform correlation.
For both of the high content platforms (microarray, left; miRNA-Seq, right), a better correlation with PCR is accom-
panied by a larger total variability (within- plus between-group) in miRNA expression. This applies across the entire
set and is not a within group measurement of variance. Circle size represents relative overall expression of miRNAs.
discovery or predictive model performance. In general, we expect that overall variance will
drive correlation of expression signal between platforms ( Fig. 5.2 ).
5.2.3 Validation of a Classifier
Once a model is identified and a diagnostic test is developed, the model typically undergoes
a rigorous validation process to determine its analytical and clinical performance. Although
validation of a model can vary depending on the clinical context, an important reference
includes the guidelines published by CLIA and College of American Pathologists (CAP) for
the design and development of a LDT. These guidelines prescribe the assay of specific perfor-
mance parameters for the test, including sensitivity, specificity, precision, and linearity. The test
validation process for the development of the miR Inform ™ Pancreas test, which is applicable to
the development of diagnostic tests in general, has been reviewed elsewhere [14] , and will be
briefly discussed.
Analytical sensitivity - the ability of an assay to produce a positive test result when the
target is present - is evaluated by measuring the lower limit of detection and lower limit of
quantification of analytes, typically titrated into a background of yeast tRNA. The analytical
sensitivity is also commonly tested with RNA extracted from clinical specimens. Analytical
specificity - the ability of an assay to produce a negative test result when the molecular tar-
get is not present - is evaluated by testing for non-specific amplification. For validation of
the miR Inform Pancreas assay, analytical specificity was evaluated by comparing the ampli-
fication of synthetic miR-196a to that of closely related miR-196b, a no-template control and
Search WWH ::




Custom Search