rankReads {fcros}

This function computes a score to assess the significance of sequencing values

Description

Implementation of two methods based (1) on the coefficient of variation or (2) on the fold change rank ordering statistics for detecting genes with significant sequencing values (gwssv). A score is obtained for each gene and a threshold allows to select the number of gwssv.

Usage

rankReads(xdata, cont, test, meth=0, Ttimes=10, err=0.1, trim.opt=0,
                        rseed=60)

Arguments

xdata

A matrix or a table containing sequencing dataset. The rownames of xdata is used for the output idnames.

cont

A vector containing the label names of the control samples: cont = c("cont01", "cont02", ...).

test

A vector containing the label names of the test samples: test = c("test01", "test02", "test03", ...).

meth

This parameter allows to specify the approach to use. The value 0 (defaul) means the coefficient of variation is used. When non-zero value is given, the fcros method used: meth = 0

Ttimes

The number of perturbed data to use. The value 10 (default) means that the dataset is used 20 times and small uniform values are added at each time: Ttimes = 10

err

This is the amount of the uniform values to add to count values. The value 0.1 (default) is used: err = 0.1

trim.opt

A scalar between 0 and 0.5. The value 0.25 (default) means that 25% of the lower and the upper rank values of each gene are not used for computing its statistics "ri", i.e. the inter-quartile range rank values are averaged: trim.opt = 0.25

rseed

This value allow to set the computer random generation seed value in order to be able to have the same results for runs performed at different times: rseed = 58

Details

Label names appearing in the parameters "cont" and "test" should match with some label names in the columns of the data matrix "xdata". It is not necessary to use all label names appearing in the columns of the dataset matrix. For a general purpose dataset, one of these parametere can be empty.

Value

This function returns a data frame containing 10 components when meth=1 and 3 components when meth=0

idnames

A vector containing the list of IDs or symbols associated with genes

score

coefficient of variation (meth=0) or Fisher-Snedecor test p-value (meth=1). Smaller (higher) values are associated with genes with significant (non significant) sequencing values.

moy

trimmed means associated with genes (when meth=0).

ri

The average of rank values associated with genes when meth=1. These values are rank values statistics leading to f-values and p-values (when meth=1).

FC

The fold changes for genes in the dataset. These fold changes are calculated as a ratio of averages from the test and the control samples. Non log scale values are used in the calculation (when meth=1).

FC2

The robust fold changes for genes. These fold changes are calculated as a trimmed mean of the fold changes or ratios obtained from the dataset samples. Non log scale values are used in the calculation (when meth=1).

f.value

The f-values are probabilities associated with genes using the "mean" and the "standard deviation" ("sd") of the statistics "ri". The "mean" and "sd" are used as a normal distribution parameters (when meth=1).

p.value

The p-values associated with genes. These values are obtained from the fold change rank values and one sample t-test (when meth=1).

Author(s)

Doulaye Dembele [e-mail]

References

Dembele D, manuscript under preparation

Examples

   data(bott);
   cont <- c("SRX033480", "SRX033488", "SRX033481");
   test <- c("SRX033493", "SRX033486", "SRX033494");
   n <- nrow(bott);

   x2 <- tcnReads(bott[,c(cont,test)])
   idx.ok <- (apply(x2, 1, sum) != 0)
   xdata <- x2[,c(cont,test)]
   rownames(xdata) <- bott[,1]
   idx.ok <- (apply(x2, 1, sum) != 0)
   tt2 <- sum(idx.ok)

   raf10.cv <- rankReads(xdata, cont, test, meth=0)
   raf10.pv <- rankReads(xdata, cont, test, meth=1)
   score.cv <- -log10(sort(raf10.cv$score))
   score.pv <- -log10(sort(raf10.pv$score))
   tmp <- scoreThr(score.cv, 2500, 3500)
   tmp

   tmp <- scoreThr(score.pv, 2500, 3500)
   tmp

   op <- par(mfrow = c(1,2))
   plot(score.cv, xlab = "index of genes",
      ylab = "-log10(sorted(score)", main = "rs.cv", type = "l",
      col = "blue", panel.first = grid())
   plot(score.pv, xlab = "index of genes",
      ylab = "-log10(sorted(score)", main = "rs.pv", type = "l",
      col = "blue", panel.first = grid())
   par(op)