zengmudong/High-dimensional-mediation-analysis

High-dimensional mediation analysis for selecting DNA methylation Loci mediating childhood trauma and cortisol stress reactivity

High-dimensional-mediation-analysis

High-dimensional mediation analysis for selecting DNA methylation Loci mediating childhood trauma and cortisol stress reactivity

Supplementary material to the manuscript High-dimensional mediation analysis forselecting DNA methylation Loci mediating childhood trauma and cortisol stress reactivity.

1. Download data and set up correct paths

• Step 1. Download data E-GEOD-77445.processed.1.zip
  (https://www.ebi.ac.uk/arrayexpress/files/E-GEOD-77445/E-GEOD-77445.processed.1.zip) as well as E-GEOD-77445.processed.2.zip (https://www.ebi.ac.uk/arrayexpress/files/E-GEOD-77445/E-GEOD-77445.processed.2.zip). Unzip these two zip files in the data folder.

    Note that there is no need to download `E-GEOD-77445.sdrf.txt`, which has 
    different row length and cannot be read into R directly. We have downloaded 
    and manually cleaned it. We saved it as `X.tsv` in the `data` folder 
    because this file contains information about exposure variable.
  

• Step 2. Open High-dimensional-mediation-analysis-JASA.Rproj. This step will help you to correctly setup working directory.

Now you are ready to run the data analysis related to case study in Section 3

2. Real Data Analysis

• Step 1. Open case_study_main.Rmd in the Rstudio.

      Before running this R markdown, you have to set the `data` folder's absolute path correctly. 
  
      To this end, we may have to change the path 
  
          PATH = "E:/JASAACS20210115Code/data/"
  
      in the R markdown file to the corresponding one in which you store your files. 
      
      Caution: the PATH should end with `/` or `\\`.
  

• Step 2. After setting the path correctly, you can click Knit (in the upper left corner) to run the entire markdown.

      Alternatively, you may run the R code in individual cells one by one.
  

You may need to wait for a few minutes to knit this R markdown. For your convenience, we have saved its output as case_study_main.html and case_study_main.pdf. Thus, you may have a quick check.

3. Run simulation for this paper

• Step 1. Open High-dimensional-mediation-analysis-JASA.Rproj. This step will help you to correctly setup working directory.

• Step 2. Open and run simulationPreprocessing.R in R studio

• Step 3. Open and run Simulation_no_confounders.R for simulation studies without confounding variables (section 4.1) This may take two weeks to finish the simulation in this part.

• Step 4. Open and run Simulation_confounders.R for simulation studies with confounding variables (section 4.2) This may take two weeks to finish the simulation in this part.

• Step 5. Open and run Simulation_compare_sigma1.R for comparing $\hat{\sigma}_1$ by three methods: our proposed new method, oracle and Zhou et al (2020) (section 4.2 Figure 3) This may take 10 hours to finish the simulation in this part.

• Step 6. Open and run Simulation_global_test_compare.R for comparing our method with oracle model, Zhou et al. (2020) and global test for both non-confounding and confounding cases (section S.3)
  This may take two months to finish the simulation in this part.

R code in this folder are:

1. utils_mediation.R stores all the functions that will be used in case_study_main.Rmd.
2. simulationPreprocessing.R loads the processed data and prepare for the simulations (section 4). Please run through this file before the following files, because all the simulations depend on this file's result.
3. Simulation_no_confounders.R performs simulation studies without confounding variables (section 4.1).
4. Simulation_confounders.R performs simulation studies with confounding variables (section 4.2).
5. Simulation_compare_sigma1.R compares estimated $\hat{\sigma}_1$ among three methods: our proposed new method, oracle and Zhou et al (2020) (section 4.2 Figure 3).
6. Simulation_global_test_compare.R compares our method with oracle model, Zhou et al. (2020) and global test for both non-confounding and confounding cases (section S.3).

#Other files:

7. `case_study_main.Rmd' is a R markdown file for real data analysis

8. case_study_main.html is a markdown file generated by case_study_main.Rmd which is easy-to-read in any browser (such as Chrome).

9. case_study_main.pdf is a markdown file generated by case_study_main.Rmd.

10. README.md is a markdown file providing the instruction of using R markdown and R code in this folder.

11. README.html is a file generated by README.md

12. README.pdf is a file generated by README.md

Note that simulationPreprocessing.R file requires R package freebird, which should be installed from the GitHub. If you fail to install freebird package, this likely is because your computer has an old version of Rtools. To update Rtools, you may go to the following link:

https://cran.r-project.org/bin/windows/Rtools/rtools40.html

and follow its instruction to install Rtools and setup it correclty.

• After installing and setting path of Rtools successfully, rerun codes in line 6 -13 of simulationPreprocessing.R.