Biology Reference
In-Depth Information
yeast tRNA. Targets were routinely not detected in the no-template control, while yeast
tRNA targets were detected at levels that lie outside the range of reliable detection. Precision
refers to the ability of an assay to produce the same result under repetition. The assessment
of precision should demonstrate minimal within-run, run-to-run, lot-to-lot, and operator-
to-operator variability. Finally, linearity is the ability of an assay to produce a linear rela-
tionship to known standards or dilutions. Linearity should cover the reportable range of
the final test score, and the expected concentrations of target values from clinical samples.
For clinical validation of a diagnostic test, the key performance parameters to be tested are
clinical sensitivity and clinical specificity. Clinical or diagnostic sensitivity is the ability of
an assay to produce a positive test result when the patient has the disease or clinical con-
dition. Conversely, clinical or diagnostic specificity is the ability of an assay to produce a
negative test result when the patient sample does not have the disease or clinical condition.
Reasonable estimates for test results can vary from test to test, but it is critical that whatever
criteria are put in place that they are well documented, reproducible and defensible.
5.2.4 miRNA Normalization
The accurate measurement of changes in miRNA expression levels is a critical objective
in biomarker studies. Even modest differences in miRNA quantification can be biologically
significant [48,49] , and careful attention must be paid to experimental design and data inter-
pretation to ensure reproducible measurements among samples. When comparing sample
groups, such as normal and diseased biopsies, the sources of variation are manifold, and may
include technical factors, such as sample procurement, handling, stabilization and storage,
RNA extraction, and the efficiency of target quantification, or biological differences, such as
heterogeneous and varied cellular composition. Moreover, the total miRNA population is
only a fleeting fraction of the bulk RNA in a sample, and miRNA yields can vary from one
specimen to the next. To address these challenges, many RNA-based assays invoke normali-
zation as a statistical strategy to correct as much of the sample-to-sample measurement vari-
ation as possible, leaving only the biological fingerprint of miRNA expression differences to
reflect phenotypic differences between experimental groups.
Because there is no such thing as a 'universal normalizer', these reference genes must be
identified through empirical studies that are faithful to the experimental workflow. For
RT-qPCR studies that aim to identify individual miRNA targets associated with a particular
disease, a number of different approaches have been described [48-52] . Each of these methods
recognizes the value of quantifying targets against endogenous RNA molecules that can cor-
rect for the noise inherent across the complex workflows that are required to reveal miRNA
levels. Two of these methods deserve further discussion; these address reduced or large-scale
experimental designs, respectively, and are effective for applications of miRNA biomarker dis-
covery in diverse sample types, including challenging specimens such as biofluids.
In cases where samples are limiting, or when the highest possible performing (typically
singleplex) RT-qPCR assays are desired, a reduced scale experimental plan is warranted [50] .
In this approach, a short list of reference gene candidates are identified from related stud-
ies [48] or current best evidence, and these are evaluated across the samples and workflows
of interest. Typically, these candidates are themselves miRNAs, since other species such as
ribosomal RNA have been shown to be statistically inferior [48] . Open source programs such
Search WWH ::




Custom Search