Package installation and systems specification

The packages used to run the code is stored at packages.yml. One way to install these packages is conda env create -f packages.yml.

This codebase is designed specifically for HPCs using LSF. This codebase will not work on local machines or Slurm.

Data preparation

Atlas data

The atlas data should have the following format:

• The file should be a tsv file, with columns being the cell types or tissues, and the columns being the genes
• If the file has N columns, then the N-1 columns are the names of the cell types.
• The last column should have the name gene and contains either gene Ensembl IDs or gene symbols for each gene in the atlas data. A sample atlas data can be found at sample_data/sample_atlas_data/human_protein_atlas_all_tissues.tsv.

Proteomic summary statistics data

All the proteomic summary statistics data should be stored in a same folder. The naming of each summary statistic file should follow the format <disease_name>.csv. Each csv file should only contain the summary statistics of all proteins in one particular disease or condition, and different diseases are separated in different files. This codebase is designed to primarily deal with sumstats data in the format introduced from this paper. Examples of such sumstats data can be found at sample_data/sample_sumstats/Alcoholic_liver_disease.csv and sample_data/sample_sumstats/Nonalcoholic_fatty_liver_disease.csv.

If you would like to provide your own summary statistics, then it should follow the following format:

• It should contain at least these 4 columns (not in a particular order): P_value, logOR (or logHR), OR (or HR), and gene. Other columns in the file will be ignored by the code.
• The gene column can contain either gene Ensembl IDs or gene symbols, but it should be in the same format as the gene column in the atlas data. An example custom sumstats file is at sample_data/sample_sumstats/AD_meta_sumstats_all.csv

Individual analyses

• All relevant code in python_main_cell_type_spec_method.

• Run the auto-script version of the .py code that accepts relevant input arguments. Other arguments can be found in the relavant bash files in the bash_scripts folder.

• 4 versions of code:

  • elastic_kfold_ver2: Run ElasticNet using k-fold cross-validation, find the best performing model, and obtain cell-tissue coefficients.
  • lasso_stability_analyses: Run a custom-made algorithm to find a suitable lambda value for LASSO, then run LASSO stability analysis across 2000 subsamples of the original dataset, then obtain the most stable cell-tissues.
  • tree_based_methods: Run random forests and obtain cell-tissue importance scores.
  • univar_association_testing: Run univariate regression for each cell-tissue, similar to univar_regression, but with more parameters included (beta, z-scores, FDR adjusted, ...) and options to add covariates such as mean gene expression, principal components, ...

Pipeline analysis

Pipelines are designed to run the aforementioned individual scripts in a more systematic way. One such pipeline design is to run univariate regression first, identify the most significant features by FDR correction, then run multivariate methods using these features and rank them based on coefficient magnitudes / permutation importance scores. This is implemented in pipelines/pipeline_univar_to_multivar.py.

The single input argument passed into the pipeline is a yml file containing all information needed to run the pipeline, such as input/output paths, disease types, and hyperparameters of individual methods. An example yml file is at univar_multivar_sample.yml.

To run a pipeline, launch a lightweight interactive job, then run:

python pipelines/pipeline_univar_to_multivar.py --yml_file pipeline_yml/univar_multivar_sample.yml