An empirical Bayes change point model for gene expression timecourse data. Time-course experiments are commonly conducted to capture temporal changes. It is generally of interest to detect if any changes happen over time, which we define as a detection problem (Q1). If there is a change, it is informative to know when the change is, which we define as an identification problem (Q2). It is often desired to control Type I error rate at a nominal level while applying a testing procedure to detect or identify these changes. The EBtimecourse model provides an unified multiple-testing framework built upon an empirical Bayes change-point model to solve these two problems. The detail is described in our paper "An Empirical Bayes Change Point Model for Transcriptome Time Course Data" published in Annals of Applied Statistics. https://doi.org/10.1214/20-AOAS1403
In the figure, Xij means the expression level of gene i at time point j.
Required R packages: tensorflow, foreach.
Use soure("EBtimecourse.R") to load the function in the R enviroment.
The function only accepts one conditional timecourse data. The input data should be log scaled microarray data or normalized and log scaled RNA-seq data. If you have two conditional data, say cases and controls, you can transform it to one conditional data: case - control for paired data, averaged cases - averaged controls for unpaired data.
Parameters for EBtimecourse function:
exp.dat - matrix of data, rows stand for genes and columns stand for time points.
timepoint - number of time points
replicate - a vector of number of replicates for each time point.
FDR - expected false discovery rate (default 0.1).
learning_rate - learning rate for the Adam optimizer (default 0.001).
max_iter - max iterations (defualt 1e5).
rel_tol - relative tolerance threshold for optimizing termination (defualt 1e-10).
threads - number of threads.
If you have a data of 8 time points, and each time point has 3 replicates, then timepoint=8 and replicate=c(3,3,3,3,3,3,3,3).
The output of EBtimecourse function is a list:
cp.index - index (row index of the data matrix) of genes detected to have change points at the FDR level (result of Q1).
cp.position - matrix stores index of changed genes and the position of change points at the FDR level (result of Q2).
mle_parameter - hyper-parameters estimated by MLE. There are five hyper-parameters:
"P": 1-P is the prior probability to have change points.
"mu0", "kappa0", "alpha0", "beta0" are the hyper-parameters of the NNG model.
ll - likelihood.
cp.position shows the length of the homogeneous sequences separated by the two change points. If there are 8 time points, one genes have a change point between 2nd and 3rd point, and another change point is between 5th and 6th point, then cp.position for this gene will be c(2,3,3). Thus, it is intuitive to get the position of change points by cp.position.
A detailed tutorial of using EBtimecourse to a simulated data is provided in Tutorial.ipynb.
Simulate_sample stores code for simulation experiments.
data stores the real data we used in the paper.
Here is version information of my R enviroment:
R version 3.6.0 (2019-04-26)
Platform: x86_64-apple-darwin18.6.0 (64-bit)
Running under: macOS 10.15.4
Matrix products: default
BLAS: /System/Library/Frameworks/Accelerate.framework/Versions/A/Frameworks/vecLib.framework/Versions/A/libBLAS.dylib
LAPACK: /usr/local/Cellar/openblas/0.3.6_1/lib/libopenblasp-r0.3.6.dylib
locale:
[1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8
attached base packages:
[1] stats graphics grDevices utils datasets methods base
other attached packages:
[1] foreach_1.4.7 tensorflow_2.0.0
loaded via a namespace (and not attached):
[1] compiler_3.6.0 magrittr_1.5 Matrix_1.2-17 tools_3.6.0 whisker_0.4 base64enc_0.1-3 Rcpp_1.0.3
[8] reticulate_1.13 codetools_0.2-16 grid_3.6.0 iterators_1.0.12 jsonlite_1.6 tfruns_1.4 lattice_0.20-38
Tian Tian [email protected]