Biological Applications of Random Walk with Restart on Multilayer Networks

Anthony Baptista, Galadriel Brière, Anaïs Baudot

Node prioritization in Leukemia

Run MultiXrank using the following command line:

python ~/ApplicationsMultiXrank/Leukemia/run_mxr.py

Visualize top 20 proritized genes and drugs in Cytoscape with file: Leukemia/multiXrank_results/top20_cyto.cys

⚠️ We extended this use-case to include regulatory and metabolic networks. A tutorial for creating the networks and running MutliXrank on this use-case is available in the form of a Jupyter-Notebook is available on this GitHub: Leukemia_2

Node prioritization in Epilepsy and Nicotine Dependence using the Hetionet network

Learn more about Hetionet:

• Systematic integration of biomedical knowledge prioritizes drugs for repurposing
  Daniel S Himmelstein, Antoine Lizee, Christine Hessler, Leo Brueggeman, Sabrina L Chen, Dexter Hadley, Ari Green, Pouya Khankhanian, Sergio E Baranzini
  eLife. 2017. DOI: 10.7554/eLife.26726

Run MultiXrank using the following commands:

1 - Build Hetionet network

python ~/ApplicationsMultiXrank/Hetionet/HetionetDB_to_MultiXrankDB/hetionet_to_multixrank.py

2 - Run MultiXrank

Epilepsy

python ~/ApplicationsMultiXrank/Hetionet/Epilepsy/run_mxr.py

Nicotine Dependence

python ~/ApplicationsMultiXrank/Hetionet/NicotineDependence/run_mxr.py

3 - Run downstream analysis of MultiXrank scores

Epilepsy

python ~/ApplicationsMultiXrank/Hetionet/Epilepsy/downstream_analysis/give_name.py

python ~/ApplicationsMultiXrank/Hetionet/Epilepsy/downstream_analysis/check_results.py

python ~/ApplicationsMultiXrank/Hetionet/Epilepsy/downstream_analysis/code_pie_tot.py

Nicotine Dependence

python ~/ApplicationsMultiXrank/Hetionet/NicotineDependence/downstream_analysis/give_name.py

Suppervised prediction of gene-disease associations

1 - Create the training set

1914 positive G-D associations (from DisGeNET, v2.0, 2014) and 3828 negative G-D associations sampled randomly

python ~/ApplicationsMultiXrank/GeneDiseaseAssociations/1_gene_disease_multiplexes/training_set.py

2 - Run MXR for all associations in the training set and store the results in sparse matrices

python ~/ApplicationsMultiXrank/GeneDiseaseAssociations/1_gene_disease_multiplexes/generate_rwr.py

python ~/ApplicationsMultiXrank/GeneDiseaseAssociations/1_gene_disease_multiplexes/make_sparse_matrices.py

3 - Train classifiers

python ~/ApplicationsMultiXrank/GeneDiseaseAssociations/1_gene_disease_multiplexes/train_models.py

4 - Compare DisGeNET v2.0 (2014) and DisGeNET v7.0 (2020) associations and generate the test set

python ~/ApplicationsMultiXrank/GeneDiseaseAssociations/1_gene_disease_multiplexes/test_2020associations/compare_2014_2020_associations.py

7218 positive G-D associations (from DisGeNET, v7.0, 2020) and 7218 negative G-D associations sampled randomly

python ~/ApplicationsMultiXrank/GeneDiseaseAssociations/1_gene_disease_multiplexes/test_2020associations/make_test_set.py

5 - Predict 2020 associations from MXR scores

python ~/ApplicationsMultiXrank/GeneDiseaseAssociations/1_gene_disease_multiplexes/test_2020associations/generate_rwr.py

python ~/ApplicationsMultiXrank/GeneDiseaseAssociations/1_gene_disease_multiplexes/test_2020associations/make_sparse_matrices.py

python ~/ApplicationsMultiXrank/GeneDiseaseAssociations/1_gene_disease_multiplexes/test_2020associations/predict_2020associations.py

You can also run the pipeline for a 3 layer network containing a drug interaction layer by repeating the same steps in GeneDiseaseAssociations/2_gene_disease_drug_multiplexes.

Diffusion Profiles in Hematopoietic Cells for Immune Disease Comparison

1 - Define seed immune diseases

python ~/ApplicationsMultiXrank/Comorbidity/immune_diseases_set.py

Going from the list of immune diseases contained in autoimmune_disease.txt, keep as seeds autoimmune diseases that appear in the disease layer diseases_monoplex_no_self_loop.tsv. Note that MultiXrank do not consider seeds that only present self-loops in the multiplex network. Hence, we only consider immune diseases that are connected to other diseases in the disease layer.

2 - Run MultiXrank for each immune disease taken as seed and each of the hematopoietic cell-specific network

python ~/ApplicationsMultiXrank/Comorbidity/run_MXR.py

3 - Recover the tree lineage of hematopoietic cells from MultiXrank output scores

python ~/ApplicationsMultiXrank/Comorbidity/tree_lineage_analysis.py

3 - Analyse immune diseases based on MultiXrank output scores

python ~/ApplicationsMultiXrank/Comorbidity/diseases_analysis.py