DiseaseInteractiveNetwork

Discovering comorbid diseases using an inter-disease interactivity network based on biobank-scale PheWAS data

Requirements

This implementation is conducted by the following packages (to be installed independently)

• pandas
• numpy
• sklearn
• scipy

Data Description

1. Used/Converted datasets of constructing DDN

• UK Biobank PheWAS summary statistics were obtained from https://www.leelabsg.org/resources

./data_bioinfo/disease_snp.npz  # 427 diseases by 39382 SNPs
./data_bioinfo/full_phecode_list.csv # description of 427 diseases with PheCode
./data_bioinfo/snp_list_rsid.csv # list of used SNPs in this analysis

2. Generated results with proposed method: Predicted scores from graph-based SSL, Relative risk & Phi-correlations obtained from EHR data

./data_bioinfo/predicted_results_for_427_diseases.zip 

3. Generated Relative risk & Phi-correlations for 427 diseases

./data_bioinfo/EHR_comorbidity_UKBB/ **.pkl 

4. Supplemenatry Data

• Supplementary Text
• Supplementary Data S1: Description for disease nodes and their associations
• Supplementary Data S2: List of synergistic and antagonistic association

Implemenations of comorbidity scoring

• The following codes expalains simple examples for predicting direct/inverse comorbidity
• Hyperparameters
  • idx_ : index disease of interest (index according to full_phecode_list.csv
  • mu_ : graph-based SSL parametere (please refer the methods sections in the manuscript)

main.py 

4. The inter-disease interactive network was implmeneted in https://hdpm.biomedinfolab.com/ddn/signedDDN