Companion GitHub to "Pitfalls in re-analysis of observational omics studies: a post-mortem of the human pathology atlas."

This GitHub page serves as a repository for reproducing all analyses and figures from our paper, Pitfalls in re-analysis of observational omics studies: a post-mortem of the human pathology atlas, that is available on bioRxiv: https://www.biorxiv.org/content/10.1101/2020.03.16.994038v2.

This GitHub page contains two main folders, BRCA and LIHC, to reproduce the analyses for the breast cancer and liver cancer case studies, respectively.

Content of the repository

Both folders contain;

1. A data folder, containing .xlsx files that were produced in the analysis, for instance, to generate Tables S1-4
2. A figures folder, containing all figures that were displayed in the main paper and in the supplementary materials.

In addition, the folders contain the scripts required to reproduce the analysis (in R markdown format) More specifically;

3. Both the BRCA and LIHC folders contain a script to get the required data from the TCGA database ("getData") and to run our revisited workflow ("revisited").
4. In addition, the BRCA folder contains a script to regenerated some of the results communicated by Uhlen et al. ("original") and a script to showcase the issues regarding this analysis ("original_problems"). To run these scripts, getData.Rmd should be run first.

Workflow reproducibility

Note that, in order to guarantee the reproducibility of our workflow, which requires the permanently changing TCGA database and R packages, we have provided a docker instance, which we used to generate all results for our paper. As such, the entire workflow remains reproducible at the snapshot date 2019-10-06.

To run docker locally, you should;

1. Install repo2docker (https://repo2docker.readthedocs.io)
2. Launch repo2docker through the command line:

jupyter-repo2docker https://github.com/statOmics/pitfallsOfHumanPathologyAtlas

Note that this step may take a while (±30 minutes) as an entirely new R instance is build for the snapshot date 2019-10-06.

3. Click the link that appears in the command line
4. In the top right corner of the jupyter localhost, click new < Rstudio
5. The scripts can be run in the new Rstudio session

Issues and comments

Please use https://github.com/statOmics/pitfallsOfHumanPathologyAtlas/issues to submit issues regarding reproducing our workflow and the comments you might have.