MONIKA

Multi-Omic Network Inference & Knockout Analysis

Performs multiplex network inference in a biological setting. Each network layer is inferred using a weighted graphical LASSO with prior incorporation from the STRING Database. Target nodes and biological pathways are identified via network diffusion analysis.

The tool has been tested on colorectal cancer datasets.

Installation

To make installation easy, it is recommended to use a conda environment (miniconda or anaconda >= 24.9.1).

Once the repository is downloaded, edit the environment.yml file, change prefix to where you want to install the environment (at the bottom of the file, prefix: path/to/anaconda3/envs/monika)

Then, simply run:

conda env create -f environment.yml

conda activate monika

pip install pymnet

conda install -c conda-forge rpy2

Data

The multi-omic CRC data used for testing is included in the repository. It is sourced from: https://www.linkedomics.org/data_download/TCGA-COADREAD/. It contains patient-coupled samples for both transcriptomics, proteomics and RPPA (Reverse-Phase Protein Array).

Full Pipeline Run

Run the scripts in the following order, with default parameter settings, to infer networks and determine critical genes via diffusion analysis.

omics_data_processing.py

• Matches protein and RNA data
• Winsorizes outliers
• Checks for normality of variables and transforms them

network_inference.py

• Executes piglasso.py for generating edge count distributions
• Infers omics networks from edge counts (RNA + protein)
  • See "Running on HPC" below to run this step in under 10 minutes

Note: You might see the following warnings from R, but the program still runs fine:

R[write to console]: In addition: R[write to console]: Warning message:

R[write to console]: In (function (package, help, pos = 2, lib.loc = NULL, character.only = FALSE, : R[write to console]:

R[write to console]: library ‘/usr/lib/R/site-library’ contains no packages

Inference Results in results/net_results:

• omics_networks_info.txt Gives info on the inferred omics network layers

network_diffusion.py

• Performs knockout analysis

Diffusion Results in results/diff_results:

• NODE_KNOCKOUTS_RESULTS_symmetricTrue_low_dens.csv is a spreadsheet containing results on the effect of knockouts on the network, as well as investigating potential increases in similarity between cms123 and cmsALL
• Top_vs_bottom_gene_GDD.png: Plots of GDD for critical genes
• diffusion_animation.gif: Diffusion GIF for visualisation

Running on HPC

The most time-consuming step by far is network_inference.py. If you want to infer the network within minutes, rather than several hours, just upload the MONIKA folder to your HPC environment (NOTE: upload the folder AFTER running omics_data_processing.py but BEFORE running network_inference.py. This ensures that the processed data is on the cluster).

Once in the cluster, you'll want to create a conda environment here too. If you have an account at SURF, it becomes even simpler:

create the conda environment with:

conda create --name monika python=3.9 numpy networkx tqdm pyparsing rpy2 -c conda-forge

The scripts to run on HPC are already in the folder src/hpc_scripts. To get the edge counts from piglasso.py, just execute the job script called runpig.sh (sbatch runpig.sh). The results from this are stored in results/net_results. Just send them back to the same location (MONIKA/results/net_results) on your local machine. Once you have the files, run network_inference.py to continue the pipeline.

Transferring files back, example: scp -r user@snellius.surf.nl:MONIKA/results/net_results path/to/local/MONIKA/results/

Note: You will have to adjust some file paths in the runpig.sh job script.