tc: Time Course Analysis for Detecting Differentially Expressed...
In acde: Artificial Components Detection of Differentially Expressed Genes

Description Usage Arguments Details Value Warning Author(s) References See Also Examples

Performs the Time Course Analysis from Acosta (2015) for detecting differentially expressed genes in time course experiments for gene expression data.

tc(data, designs, tPoints = NULL, 
    method = c("active vs complementary", "groups conformation"), 
    activeTP = NULL, alpha = 0.05, B = 100, lambda = 0.5, 
    PER = FALSE, BCa = FALSE, gamma = 0.95, R = 1000, ...)

`data`	a list with matrices or data.frames representing genes' expression levels at each time point. The rows of each matrix correspond to the genes in the experiment, and the columns correspond to the replicates. Treatment replicates are to the left, control replicates to the right.
`designs`	a list with a vector for each time point, of length equal to the number of columns in the respective matrix or data.frame in `data`, with 1's for the treatment replicates and 2's for the control replicates.
`tPoints`	a character vector with the names of the timepoints.
`method`	if "active vs complementary", an analysis following the active vs complementary time points approach (Acosta, 2015) is performed. If "groups conformation", an analysis following the groups conformation through time approach (Acosta, 2015) is performed. The default is both.
`activeTP`	numeric. The index of the active timepoint in `tPoints` for the active vs complementary time points approach. If `NULL` (default), the active time point is selected via inertia ratios.
`alpha`	between 0 and 1. Desired level for controlling the false discovery rate (FDR).
`B`	Number of bootstrap or permutation replications for estimating the FDR.
`lambda`	Parameter for the estimation of pi0 and, further, the estimation of the FDR (see Storey, 2002).
`PER`	If `FALSE` (default), bootstrap replications are used to estimate the FDR. If `TRUE`, permutation replications are used instead.
`BCa`	If TRUE, a BCa confidence upper bound for the FDR is computed (see Efron and Tibshirani, 1994).
`gamma`	Confidence level for the FDR's BCa upper confidence bound.
`R`	Number of bootstrap replications for the computation of the FDR's BCa upper confidence bound.
`...`	additional arguments for parallel computation in `boot` function (see Details).

In the active vs complementary time points approach, the time point that maximizes the inertia ratio is selected as the active time point. Then, a Single Time Point Analysis (stp) is performed on this time point and plots of the behavior throughout the time course of the differentially expressed genes identified in this time point are displayed.

In the groups conformation through time approach, a Single Time Point Analysis (stp) is performed at each time point and plots are displayed showing the behaviour of the differential expression process throughout the time course; that is, how many genes are differentially expressed and how strong is the differential expression at each time point.

For details on the computations performed in this function, see Acosta (2015).

Additional parameters in the '...' argument are used for parallel computation in bootstrap calculations. These are supplied to calls to the boot function in package boot. With this in mind, the use of additional arguments must be restricted to arguments parallel and ncpus from function boot.

tc returns an object of class 'TC', which is a list with components:

`iRatios`	inertia ratios for each time point.
`gct`	results for the groups conformation through time approach. A list with an object of class '`STP`' for each timepoint.
`act`	results for the active vs complementary time points approach. A list with an object of class '`STP`' for each timepoint.
`activeTP`	the index of the active timepoint in `tPoints` used in the active vs complementary time points approach.
`tPoints`	a character vector with the names of the timepoints.
`call`	The matched call.

If argument BCa=TRUE, computations may take a considerable amount of time.

Juan Pablo Acosta (jpacostar@unal.edu.co).

Acosta, J. P. (2015) Strategy for Multivariate Identification of Differentially Expressed Genes in Microarray Data. Unpublished MS thesis. Universidad Nacional de Colombia, Bogot\'a.

Storey, J. D. (2002) A direct approach to false discovery rates. Journal of the Royal Statistical Society: Series B (Statistical Methodology), 64(3): 479–498.

Efron B. and Tibshirani R. J. (1994) An Introduction to the Bootstrap. Chapman & Hall/CRC, 1993.

stp for Single Time Point Analysis; plot.TC, print.TC, summary.TC.

## Time course analysis for 500 genes with 10 treatment 
## replicates and 10 control replicates
tPts <- c("h0", "12h", "24h")
n <- 500; p <- 20; p1 <- 10
Z <- vector("list", 3)
des <- vector("list", 3)
for(tp in 1:3){ des[[tp]] <- c(rep(1, p1), rep(2, (p-p1))) }
mu <- as.matrix(rexp(n, rate=1))
### h0 time point (no diff. expr.)
Z[[1]] <- t(apply(mu, 1, function(mui) rnorm(p, mean=mui, sd=1)))
### h12 time point (diff. expr. begins)
Z[[2]] <- t(apply(mu, 1, function(mui) rnorm(p, mean=mui, sd=1)))
#### Up regulated genes
Z[[2]][1:5,1:p1] <- Z[[2]][1:5,1:p1] + 
    matrix(runif(5*p1, 1, 3), nrow=5)
#### Down regulated genes
Z[[2]][6:15,(p1+1):p] <- Z[[2]][6:15,(p1+1):p] + 
    matrix(runif(10*(p-p1), 1, 2), nrow=10)
### h24 time point (maximum differential expression)
Z[[3]] <- t(apply(mu, 1, function(mui) rnorm(p, mean=mui, sd=1)))
#### 5 up regulated genes
Z[[3]][1:5,1:p1] <- Z[[3]][1:5,1:p1] + 5
#### 10 down regulated genes
Z[[3]][6:15,(p1+1):p] <- Z[[3]][6:15,(p1+1):p] + 4

resTC <- tc(Z, des)
resTC
summary(resTC)
plot(resTC)

## Not run: 
## Phytophthora Infestans Time Course Analysis (takes time...)
dataPI <- phytophthora
desPI <- vector("list", 4)
for(tp in 1:4){ desPI[[tp]] <- c(rep(1, 8), rep(2, 8)) }
resPI <- tc(dataPI, desPI)
resPI
summary(resPI)
plot(resPI)

## End(Not run)