sspa-in-metabolomics

This repository contains the code for the manuscript "Performance evaluation and applicability of single-sample pathway analysis methods to metabolomics data".

Data

Su et al. (2020) and Lloyd-Price et al (2019) metabolomics data can be found in the datasets folder.

Simulations and benchmarking

Code for the benchmarking analysis is in the Benchmarking folder. Simulations are expected to be run in parallel using a high-performance computing cluster. The code to run the simulations and calculate performance metrics is in the following files:

• Benchmarking/HPC_avg_norm_ranking_parallel.py
• Benchmarking/HPC_performance_metrics_parallel_effectsize.py
• Benchmarking/HPC_performance_metrics_parallel_noise.py

To run these files the following files are required in the same directory:

• Benchmarking/helper_functs.py: code to calculate performance metrics
• Benchmarking/methods.py: contains single-sample pathway analysis methods
• Benchmarking/process_met_data.py: processes metabolomics data prior to simulation
• Benchmarking/process_pathways.py: processes Reactome pathways into correct format
• Benchmarking/simulate_met_data.py: creates simulated metabolomics data

IBD application

Code to create the IBD pathway-based correlation network is in the Applications folder in the IBD_pathway_network.ipynb notebook. This creates a .graphml file which can be imported into Cytoscape.

Additional code for:

• Comparing the ARI achieved using metabolites vs pathways for clustering
• PCA plots
• Random forest model based on kPCA pathway scores

can be found in the sspa_IBD_applications.ipynb notebook.

Cite

If you use this code please cite Wieder, Lai, and Ebbels (2022) [manuscript in preparation].