REST {qpcR} | R Documentation |
This function is in principle a port of the REST software to R (with minor modifications). The paradigm behind the REST approach is to permutate threshold cycle values between controls and samples, calculate ratios for each permutation and finally construct a confidence interval on the obtained cumulative distribution. This approach has been replicated here with modifications as described under 'Details'.
REST(eff.goi, eff.ref, data, B = 10000, alpha = 0.05, replace = FALSE, bootstrap = NULL, perm = c("pairwise", "single"))
eff.goi |
the PCR efficiency of the gene-of-interest. |
eff.ref |
the PCR efficiency of the reference gene. |
data |
a dataframe with the threshold cycles for control ref, control goi, sample ref and sample goi. STRICTLY in that order. |
B |
The number of permutations. |
alpha |
the confidence level. |
replace |
logical. Should permutations be conducted with replacement? |
bootstrap |
The proportion of data to be used for the permutation (bootstrapping). Must be between 0 and 1, default is 1. |
perm |
type of permutation. 'pairwise' (default) is the usual REST procedure ('pairwise fixed reallocation'). See 'Details'. |
The following modifications were made to (enhance) the original REST version.
a) Permutation sampling can be controlled with/without replacement and data subsetting (bootstrap).
b) Additionally to the pairwise fixed reallocation, a permutation regime without fixing control and GOI of reference/sample
can be conducted. Thus, all four input columns of data
are permutated. This increases the number of non-redundant
permutations, but contradicts the pradigm of a pairwise observation (two PCRs on the same sample). The user may judge, which
approach is more sensible.
c) The list item uniques
reveals the redundancy of permutations (=> no information gain for confidence interval).
A list with the following components:
ratios |
the sorted ratios obtained from the permutations. |
conf |
the confidence interval for alpha . |
uniques |
a dataframe with the UNIQUE permutations. Reveals redundancy in dependence of B . |
Andrej-Nikolai Spiess
Relative Expression Software Tool (REST) for group wise comparison and
statistical analysis of relative expression results in real-time PCR.
Pfaffl MW, Horgan GW & Dempfle L.
Nucl Acids Res 2002, 30: E36.
### artificial data with very low variance ### in threshold cycles (~0.5 ### See how this still gives a ### relatively large confidence interval! DATA <- data.frame(ref.C = rnorm(10, 26 , 0.1), goi.C = rnorm(10, 27, 0.1), ref.S = rnorm(10, 26, 0.1), goi.S = rnorm(10, 24, 0.1)) res <- REST(eff.goi = 2.00, eff.ref = 1.97, data = DATA, B = 10000, perm = "pairwise") boxplot(res$ratios, col = 2, outline = FALSE)