Note: some of the R-packages that DEtime relies on are no longer supported. The model has been re-implemented in the GPcounts python package and can be used with a Gaussian or counts (negative binomial) likelihood. See https://github.com/ManchesterBioinference/GPcounts/blob/master/demo_notebooks/Branching_GPcounts.ipynb for an example application there to single-cell pseudotime branching. If using the DEtime kernel in GPcounts then please cite the GPcounts paper https://doi.org/10.1093/bioinformatics/btab486 as well as the original DEtime paper

DEtime is an R package for two-sample time series analysis using Gaussian process methods. This package implements the Gaussian regression framework for perturbation time point inferrence in a two sample case.

The paper describing this package is available at DOI: https://doi.org/10.1093/bioinformatics/btw329 and arXiv: http://arxiv.org/abs/1602.01743. Please refer to the Jupyter notebook DEtime_illustration.ipynb for R codes about how to run the package.

There are two ways to install DEtime:

• use devtools package:

  • Install and load the devtools package to be able to directly install R packages hosted on github :

  install.packages("devtools")
  library(devtools)

  • To install DEtime type:

  install_github("ManchesterBioinference/DEtime")

• download the tarball file DEtime_1.0.tar.gz.

  • Install the dependent packages, spam and gptk, first:

  install.packages("spam")
  install.packages("gptk")

  • Install the tarball

  install.packages("DEtime_1.0.tar.gz", repos=NULL, type="source")

The package contains two main functions: DEtime_infer and DEtime_rank.

• DEtime_infer is the main function for perturbation point inference.

• DEtime_rank is used to filter out these silent genes before any focused perturbation point inference work.

The user is required to provide times, ControlData, PerturbedData etc to use these two functions. For explanation of these arguments, please refer to the vignettes asscoiated with this package.

• For straightforward perturbation time point inference without ranking,

library(DEtime)

### inport simulated dataset
data(SimulatedData)

### go on with the perturbation time point inference
res <- DEtime_infer(ControlTimes = ControlTimes, ControlData = ControlData, PerturbedTimes = PerturbedTimes, PerturbedData = PerturbedData)

### Print a summary of the results
print_DEtime(res)
### plot results for all the genes
plot_DEtime(res)
}

• If ranking is needed,

library(DEtime)

### inport simulated dataset
data(SimulatedData)

### calculating the loglikelihood ratio for these tested genes. the result is saved into DEtime_rank.txt

res_rank <- DEtime_rank(ControlTimes = ControlTimes, ControlData = ControlData, PerturbedTimes, PerturbedData=PerturbedData, savefile=TRUE)
 
### get the index of these data with loglikelihood ratio larger than 1
idx <- which(res_rank[,2]>1)

### go on with the perturbation time inference if some of the data has passed the threshould test 
if (length(idx)>0){
     res <- DEtime_infer(ControlTimes = ControlTimes, ControlData = ControlData[idx,], PerturbedTimes = PerturbedTimes, PerturbedData=PerturbedData[idx,])
     ### Print a summary of the results
     print_DEtime(res)
     ### plot results for all the genes
     plot_DEtime(res)
  }