chipPCR (version 1.0-2)

CPP: Curve Pre-processor

Description

CPP encompasses a set of functions to pre-process an amplification curve. The pre-processing includes options to normalize curve data, to remove background, to remove outliers in the background range and to test if an amplification is significant.

Usage

# S4 method for numeric,numeric
CPP(x, y, smoother = TRUE, method = "savgol", 
				trans = FALSE, method.reg = "lmrob", 
				bg.outliers = FALSE, median = FALSE, 
				method.norm = "none", qnL = 0.03, amptest = FALSE, 
				manual = FALSE, nl = 0.08, bg.range = NULL, ...)
# S4 method for matrix,missing
CPP(x, y, smoother = TRUE, method = "savgol", 
			       trans = FALSE, method.reg = "lmrob", 
			       bg.outliers = FALSE, median = FALSE, 
			       method.norm = "none", qnL = 0.03, amptest = FALSE, 
			       manual = FALSE, nl = 0.08, bg.range = NULL, ...)
# S4 method for data.frame,missing
CPP(x, y, smoother = TRUE, 
				   method = "savgol", trans = FALSE, 
				   method.reg = "lmrob", bg.outliers = FALSE, 
				   median = FALSE, method.norm = "none", 
				   qnL = 0.03, amptest = FALSE, 
				   manual = FALSE, nl = 0.08, bg.range = NULL, ...)

Arguments

x

is a vector containing the time or cycle values or a matrix or data frame containing both time or cycle values and fluorescence.

y

is a vector containing the fluorescence values. Omitted if x is a data frame or matrix.

smoother

logical parameter which indicates if smoother should be used.

method

a vector of names defining which smoothing method should be used. The Savitzky-Golay smoothing filter is the default smoother. Use "lowess" for LOWESS smoother (locally-weighted polynomial regression), "mova" for moving average, "savgol" for Savitzky-Golay smoothing filter, "smooth" for cubic spline smooth, "spline" for standard cubic spline smooth, "supsmu" for Friedman's SuperSmoother, "whit1" for weighted Whittaker smoothing with a first order finite difference penalty, "whit2" for weighted Whittaker smoothing with a second order finite difference penalty or "all" for all implemented smoothing algorithms.

trans

defines if the slope of the background range in a curve should be corrected by a linear regression.

method.reg

defines the method ("rfit", "lmrob", "rq") for the robust linear regression. If equal to "least", CPP uses linear regression.

bg.outliers

is a logical argument which to remove outliers in the background range.

median

If set to TRUE, median is used instead of mean in outlier replacement. The mean is used by default.

method.norm

is a argument to use a "none", "minm", "max", "lugn", or "zscore" normalization.

qnL

is the quantile to be used for the quantile normalization.

amptest

is a logical operator which is used to set a test for a positive amplification.

manual

is used to test for a fixed threshold value of the background.

nl

is a value used as fixed threshold value for the background.

bg.range

is a numeric vector of length 2 pointing the background range. If NULL, the background range is calculated by bg.max function.

...

dot operator for diverse arguments of smoother for details).

Value

A list where each element represents the pre-process data and parameters for the background region (BG) of a qPCR experiment.

Details

CPP uses the bg.max function to estimate automatically the start of the amplification process. In the background range there is often noise which makes it harder to determine a meaningful background value. Therefore CPP can optionally remove outliers by finding the value with largest difference from the mean as provided by the rm.outlier function. This function also tries to prevent calculations of non amplified signals.

The slope of the background range is often unequal to zero. By setting the parameter trans it is possible to apply a simple correction of the slope. Thereby either a robust linear regression by computing MM-type regression estimators, a nonparametric rank-based estimator or a standard linear regression model. Care is needed when using trans with time series (see lm for details).

See Also

Normalization: normalizer

Smoothing: smoother

Robust linear regression: lm.coefs

Examples

# NOT RUN {
# Function to pre-process an amplification curve.
# Take a subset of the C17 data frame.
data(C17)
default.par <- par(no.readonly = TRUE)
par(mfrow = c(2,1))
plot(NA, NA, xlab = "Time [sec]", ylab = "refMFI", 
     main = "HDA Raw Data", 
     xlim = c(0, 2500), ylim = c(0,1.1), pch = 20)
for (i in 3:5) {
  lines(C17[1:50, 1], C17[1:50, i], col = i - 2, 
	type = "b", pch = 20)
}

legend(50, 0.5, c("55 degrees Celsius", "60 degrees Celsius", "65 degrees Celsius"), 
	col = c(1,2,3), pch = rep(20,3))

# Use CPP to pre-process the data by removing the missing value and 
# normalization of the data
plot(NA, NA, xlab = "Time [sec]", ylab = "refMFI", 
     main = "Curve Pre-processor Applied to HDA Data", 
     xlim = c(0, 2500), ylim = c(0,1.1), pch = 20)
for (i in 3:5) {
  y.cpp <- CPP(C17[2:50, 1], C17[2:50, i], method.norm = "minm", 
	      bg.outliers = TRUE)$y.norm
  lines(C17[2:50, 1], y.cpp, col = i - 2, 
	type = "b", pch = 20)
}
legend(50, 1, c("55 degrees Celsius", "60 degrees Celsius", "65 degrees Celsius"), 
	col = c(1,2,3), pch = rep(20,3))

par(default.par)
# }