This repository contains the programming code we developed for the paper submitted to PLoS ONE. The code implements the methodology of the paper and reproduce all the figures of the paper.

The code is written in AWK, Python and R. To use the code you will need:

• Python 3 [https://www.python.org] with packages: community, networkx, pandas, pubchempy, requests
• R [https://www.r-project.org] with the packages: caret, data.table, dplyr, ggplot2, igraph, lsa, plotROC, ranger, ROSE, tidyverse
• XMLStarlet [https://xmlstar.sourceforge.net]
• MySQL [https://www.mysql.com] or MariaDB [https://mariadb.org] server

The Makefile file executes appropriate download scripts, Python and R files which process data. Before you run the code you should register on Drugbank Web site [https://www.drugbank.ca/]. This enables you to download appropriate files from Drugbank. In addition you should install UMLS and SemMedDB. Instructions to install UMLS can be found in the following link [https://www.nlm.nih.gov/research/umls/quickstart.html]. Next, go to SemMedDB site [https://skr3.nlm.nih.gov/SemMedDB/], download PREDICATION file and run sql script. Note that you should also insert appropriate credentials into Makefile (only in drugbank, ndfrt, and semmeddb directories).

First download the code from GitHub repository using the command

git clone https://github.com/akastrin/ddi-prediction.git

or by downloading a zipped version and extracting it as usual.

There are two directories in the home folder (ddi-prediction), namely df and figures. db directory contains the following subdirectories: drugbank, kegg, ndfrt, semmeddb, and twosides. To reproduce the results of the study you should execute Makefile file in each database directory one after another. For example:

cd db/drugbank/
make

To reproduce the figures presented in the paper run the following code from the command line:

cd figures/
Rscript net_plot.R
Rscript roc_plot.R
Rscript var_imp_plot.R
Rscript hist_plot.R
Rscript hclust_plot.R