Both pipelines from the manuscript are available here as lasso_pipeline.R and rf_pipeline.R. They both use the same input files.

The LASSO pipeline contains more post-processing and visualizations than the RF pipeline. If you want to create identical PCA plots for the RF pipeline simply run a PCA on all your markers beforehand (see code in lasso_pipeline.R if unfamiliar with how to do this). Then take the set of selected features output from the RF pipeline and run a PCA using only the selected features. Note that the {}_transcriptImportance.txt file output from the RF pipeline contains variable importance values for all features, you need to filter this based on your desired significance cut-off for your data.

The example directory contains data to run a small example analysis. This will demonstrate all output statistics and visualizations. This is a random subset of the Poplar data presented in the paper so do not expect to obtain the same results.

The meta-data used for both the Zhu et al. and Mok et al. datasets are included in the malaria data directory. We have included this to make it easier for future investigators as some data was generated via text-scraping.

The paper describing the Mok et al. data can be found here: https://pubmed.ncbi.nlm.nih.gov/25502316/. Their transcriptome data can be downloaded from here: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE59097.

The paper describing the Zhu et al. data can be found here: https://www.nature.com/articles/s41467-018-07588-x. Their transcriptome data can be downloaded from here: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE121505.

To use the code with your own data, open the pipeline file in your favorite editor and simply change the strings in the first lines:

samples - The path to the samples file

expr - The path to the expression file

outname - Desired name of output directory to save stats and visualizations

subtit - The string to use as a subtitle on each plot. Not used in the RF pipeline as no plotting is done.

Once changed, the pipeline can easily be run using the Rscript command.

The pipeline requires two main files:

1. The samples file: this is a tab-separated text file with two columns. Headers must be "Genotype_ID" and "Pheno". The Genotype_ID column should contain genotype codes to match samples to the expression data. The Pheno column should contain phenotypic class values for each sample.

2. The expression data file: this is a tab-separated text file. The first column must be named "Genotype_ID" and these values should match those in the samples file. All subsequent columns should contain the markers to look for associations with.