Automatic Outlier Removal - Applied Biosystems 7900HT User Manual

Automatic Outlier Removal

About Outlying
Replicate Wells
About the
Automatic Outlier
Removal
Regarding
Amplification
Beyond
Reproducible
Limits
Applied Biosystems 7900HT Fast Real-Time PCR System and SDS Enterprise Database User Guide
For any PCR, experimental error may cause some wells to amplify insufficiently or
not at all. These wells typically produce C
average for the associated replicate wells. If included in the calculations, the C
values from these wells indicate an incorrect level of relative gene expression by
skewing the average for the replicate group.
The SDS software offers algorithmic identification and removal of outlying data for
studies consisting of replicate populations of three or more wells. The statistical
method used by the software is based on Grubbs outlier removal (also known as the
Maximum Normalized Residual Test), which permits the exclusion of a single outlier
in a population consisting of as few as three replicates.
The software applies the Grubbs tests at different stages in the ∆∆C
depending on the type of endogenous control used in the study. For non-multiplex
studies, the software removes outliers from replicate groups immediately after
calculating threshold cycles (C
algorithm following the ∆C
C s of the mean for the associated replicate group, the software does not remove it.
T
The outlier removal feature is optional and can be activated from the Analysis
Note:
Settings dialog box (see
manually as explained on
In addition to the Grubbs tests, the algorithm used by the SDS software features
additional rules for dealing with SDS data. Unless the majority of samples in a
replicate population are at maximum cycle, the algorithm ignores max cycle wells
(wells with C s equal to or greater than the Maximum cycle value).
T
The experimental reproducibility of C
decreases as the starting copy number approaches extremely low levels (less than
100 copies). Therefore, PCR targets with low starting copy number and subsequent
high C values may have significantly diminished statistical reproducibility. To avoid
T
problems with reproducibility for low-expression targets, the limits of reproducibility
must be determined experimentally for each relative quantification assay (all targets
and the endogenous control).
values that differ significantly from the
T
). For multiplex studies, the software applies the
T
calculation. Also, if an apparent outlier is within 0.25
T
page 6-29). Outliers can also be identified and eliminated
page 6-49.
values for relative quantification assays
T
September 1, 2004 11:39 am, CH_Real-Time.fm
calculation,
T
DRAFT
Overview
T
6-19