https://github.com/wlandau/rnaseq_dispersion_2013

My Masters project
https://github.com/wlandau/rnaseq_dispersion_2013

Last synced: 8 days ago
JSON representation

My Masters project

• Host: GitHub
• URL: https://github.com/wlandau/rnaseq_dispersion_2013
• Owner: wlandau
• Created: 2012-08-23T08:02:43.000Z (about 12 years ago)
• Default Branch: main
• Last Pushed: 2020-10-27T16:21:05.000Z (about 4 years ago)
• Last Synced: 2024-10-11T18:25:30.901Z (about 1 month ago)
• Language: R
• Size: 13 MB
• Stars: 0
• Watchers: 6
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
These contain the code for "Dispersion Estimation and Its Effect on Test 

Performance in RNA-seq Data Analysis" by Will Landau and Peng Liu. 

DEPENDENCIES:

1) R version 2.15.3 (2013-03-01). See http://cran.r-project.org/ for details and usage.

2) The following R packages:

  abind version 1.4-0 (hosted on CRAN)

  AMAP.Seq version 1.0 (hosted on CRAN)

  Biobase version 2.18.0 (hosted on Bioconductor)

  clusterGeneration version 1.3.1 (hosted on CRAN)

  DESeq version 1.10.1 (hosted on Bioconductor)

  DSS version 1.0.0 (hosted on Bioconductor)

  edgeR version 3.0.8 (hosted on Bioconductor)

  ggplot2 version 0.9.3.1 (hosted on CRAN)

  hexbin version 1.26.1 (hosted on CRAN)

  iterators version 1.0.6 (hosted on CRAN)

  magic version 1.5-4 (hosted on CRAN)

  MASS version 7.3-23 (hosted on CRAN)

  multicore version 0.1-7 (hosted on CRAN)

  plyr version 1.8 (hosted on CRAN)

  QuasiSeq version 1.0-2 (hosted on CRAN)

  pracma version 1.4.5 (hosted on CRAN)

  reshape2 version 1.2.2 (hosted on CRAN)

The packages hosted on CRAN can be installed by typing the command in R:

  install.packages("packagename")

For help (for example, to find out how to locally install a package), type into R:

  ?install.packages

The packages hosted on Bioconductor can be installed by typing the following commands into R:

  source("http://bioconductor.org/biocLite.R")

  biocLite()

  biocLite("packagename")

For help on installing packages from Bioconductor, type:

  

  source("http://bioconductor.org/biocLite.R")

  ?biocLite

USAGE:

  The commands in example.r demonstrate a basic, step-by-step usage of the simulation code. 

  However, one can run the full code in batch mode in two steps:

1) Configure the makeArgs function in functions.r to set the simulation parameters:

  makeArgs = function(){

    print("Setting simulation parameters.")

    groups = rep(list(c(rep(1,3), rep(2,3)),

                      c(rep(1,3), rep(2,15)),

                      c(rep(1,9), rep(2,9))), times = 2)

    sim.settings = 1:6

  

    lapply(1:length(sim.settings),

           function(cond){list(

             nG = 10000, # number of genes (10000)

             blocksize = 50, # size of each group of correlated pseudo-genes (50)

             nreps = 30, # num. pseudo-datasets per simulation setting (30)

             ncores = 10, # 10

             p0.true = .8,

             sim.setting = sim.settings[cond],

             group = groups[cond][[1]]

           )})

  }

The user has the option to set:

  nG, the number of genes used in each pseudo-dataset.

  blocksize, the size of each group of pseudo-genes with correlated log fold changes

  nreps, the number of pseudo-datasets to generate per simulation setting

  ncores, the number of cores to use. With 10 cores, the workflow takes about a day.

  p0.true, the proportion of equivalently expressed pseudo-genes

2) Modify the path to Rscirpt in run.sh. For example, if Rscirpt is stored in the /usr/local/bin 

directory and run.sh has the line,

  rscript=/apps/R-2.15.3/bin/Rscript 

   

change it to:

  rscript=/usr/local/bin/Rscript

3) In a command line tool like Terminal in Mac OS X, change directories into the directory

containing functions.r. For example:

  cd /myHome/RNASeq_Dispersion_2013/code

4) In the same command line tool, type:

  ./run.sh 

The script takes about a day to run with 10 parallel CPU cores. 

It will create directories "rds" and "fig" to store the following results:

  rds

    args.rds: simulation parameters

    auc.rds: data structure storing the AUC (area under ROC curve) values

    countsI.rds: pseudo-counts for simulation setting I

    countsII.rds: pseudo-counts for simulation setting I

    countsIII.rds: pseudo-counts for simulation setting I

    countsIV.rds: pseudo-counts for simulation setting I

    countsV.rds: pseudo-counts for simulation setting I

    countsVI.rds: pseudo-counts for simulation setting I

    de.rds: array of indicators: 0 means that a pseudo-gene was truly equivalently expressed,

            1 means that a pseudo-gene was truly differentially expressed.

    delta.rds: array of log fold changes, delta. See the simulation section of the main paper.

    phi.rds: array of dispersions

    pvalsI.rds: p-values from the tests for differential expression, simulation setting I

    pvalsII.rds: p-values from the tests for differential expression, simulation setting II

    pvalsIII.rds: p-values from the tests for differential expression, simulation setting II

    pvalsIV.rds: p-values from the tests for differential expression, simulation setting IV

    pvalsV.rds: p-values from the tests for differential expression, simulation setting V

    pvalsVI.rds: p-values from the tests for differential expression, simulation setting VI

    rocI.rds: ROC (receiver operating characteristic) curves for simulation setting I

    rocII.rds: ROC (receiver operating characteristic) curves for simulation setting II

    rocIII.rds: ROC (receiver operating characteristic) curves for simulation setting III

    rocIV.rds: ROC (receiver operating characteristic) curves for simulation setting IV

    rocV.rds: ROC (receiver operating characteristic) curves for simulation setting V

    rocVI.rds: ROC (receiver operating characteristic) curves for simulation setting VI

    sessionInfo.rds: technical info of the current R session: computer architecture, packages, etc.

    true_dispsI.rds: dispersions phihat_g used to simulate pseudo-data for simulation setting I

    true_dispsII.rds: dispersions phihat_g used to simulate pseudo-data for simulation setting II

    true_dispsIII.rds: dispersions phihat_g used to simulate pseudo-data for simulation setting III

    true_dispsIV.rds: dispersions phihat_g used to simulate pseudo-data for simulation setting IV

    true_dispsV.rds: dispersions phihat_g used to simulate pseudo-data for simulation setting V

    true_dispsVI.rds: dispersions phihat_g used to simulate pseudo-data for simulation setting VI

    true_meansI.rds: geometric means ybar_g. used to simulate pseudo-data for simulation setting I

    true_meansII.rds: geometric means ybar_g. used to simulate pseudo-data for simulation setting II

    true_meansIII.rds: geometric means ybar_g. used to simulate pseudo-data for simulation setting III

    true_meansIV.rds: geometric means ybar_g. used to simulate pseudo-data for simulation setting IV

    true_meansV.rds: geometric means ybar_g. used to simulate pseudo-data for simulation setting V

    true_meansVI.rds: geometric means ybar_g. used to simulate pseudo-data for simulation setting VI

  fig

    auc1.pdf: AUC values for simulation setting I

    auc2.pdf: AUC values for simulation setting II

    auc3.pdf: AUC values for simulation setting III

    auc4.pdf: AUC values for simulation setting IV

    auc5.pdf: AUC values for simulation setting V

    auc6.pdf: AUC values for simulation setting VI

    hists.pdf: histograms of means and dispersions of Hammer and Pickrell datasets

               along with similar histograms of example Hammer-generated and Pickrell-generated

               pseudo-datasets

    meandisp_scatter.pdf: mean-dispersion relationships for Hammer, Pickrell, Hammer-generated,

                          and Pickrell-generated data.

    mse.pdf: mean squared error of transformed dispersions

    phi_vs_phiI.pdf: scatterplots of estimated dispersions vs. true dispersions for an example

                     pseudo-dataset from simulation setting I.

    phi_vs_phiII.pdf: scatterplots of estimated dispersions vs. true dispersions for an example

                     pseudo-dataset from simulation setting II.

    phi_vs_phiIII.pdf: scatterplots of estimated dispersions vs. true dispersions for an example

                     pseudo-dataset from simulation setting III.

    phi_vs_phiIV.pdf: scatterplots of estimated dispersions vs. true dispersions for an example

                     pseudo-dataset from simulation setting IV.

    phi_vs_phiV.pdf: scatterplots of estimated dispersions vs. true dispersions for an example

                     pseudo-dataset from simulation setting V.

    phi_vs_phiVI.pdf: scatterplots of estimated dispersions vs. true dispersions for an example

                     pseudo-dataset from simulation setting VI.

    roc.pdf: 2 example ROC curves using tests for differential expression on pseudo-data