Exploratory Analysis of Drug-to-Drug Side Effects Using Patient-Centered Data
This repository contains the code and data used in the paper:
“A Network Approach to Polypharmacy: An Exploratory Analysis of Drug-to-Drug Side Effects Using Patient-Centered Data”
Our work applies network analysis techniques to a large drug interaction graph to explore the reactivity and harmfulness of various medications, particularly those most commonly prescribed in the United States. We combine structural graph metrics with crowdsourced rankings of side effect severity to offer a patient-centered perspective on polypharmacy risks.
We built an undirected weighted multigraph where:
- Nodes represent drugs.
- Edges represent documented side effects due to interactions.
- Edge weights reflect perceived severity of side effects (fear score), matched from a separate study using MiniLM sentence embeddings.
Key metrics computed:
- Simple Degree Centrality – Unique drug interaction count
- Multi-edge Degree Centrality – Total number of interaction edges
- Weighted Degree – Harmfulness + frequency combined
- Averaged Weighted Degree – Mean harmfulness per interaction
- Edge Multiplicity – Number of side effects between specific drug pairs
- Closeness Centrality – Proxy for interaction predictability
-
BioSNAP Polypharmacy Dataset https://snap.stanford.edu/biodata/datasets/10017/10017-ChChSe-Decagon.html
-
SIDER / OFFSIDES / TWOSIDES https://nsides.io/
-
Perceived Harmfulness Rankings Gottlieb et al., JMIR, 2015
-
Top 300 US Prescribed Drugs (2022) https://clincalc.com/DrugStats/Top300Drugs.aspx
-
PubChem API https://pubchem.ncbi.nlm.nih.gov/