# Drug Interaction Vectors Neural Network: DrIVeNN

arXiv: https://arxiv.org/abs/2308.13891 · HTML: https://mlsanity.com/paper/2308.13891/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 9 of its 32 references that are in the map.
- First posted on arXiv 26 Aug 2023.

## Builds on (most cited first, 9 of 9)

- [Neural Message Passing for Quantum Chemistry](https://mlsanity.com/paper/1704.01212/) (2017)
- [Distributed Representations of Words and Phrases and their Compositionality](https://mlsanity.com/paper/1310.4546/) (2013)
- [UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction](https://mlsanity.com/paper/1802.03426/) (2018)
- [DeepWalk: Online Learning of Social Representations](https://mlsanity.com/paper/1403.6652/) (2014)
- [Complex Embeddings for Simple Link Prediction](https://mlsanity.com/paper/1606.06357/) (2016)
- [Hyperband: A Novel Bandit-Based Approach to Hyperparameter Optimization](https://mlsanity.com/paper/1603.06560/) (2016)
- [Modeling polypharmacy side effects with graph convolutional networks](https://mlsanity.com/paper/1802.00543/) (2018)
- [DGL-LifeSci: An Open-Source Toolkit for Deep Learning on Graphs in Life Science](https://mlsanity.com/paper/2106.14232/) (2021)
- [Drug-Drug Interaction Prediction Based on Knowledge Graph Embeddings and Convolutional-LSTM Network](https://mlsanity.com/paper/1908.01288/) (2019)

## Abstract (arXiv)

Polypharmacy, the concurrent use of multiple drugs to treat a single condition, is common in patients managing multiple or complex conditions. However, as more drugs are added to the treatment plan, the risk of adverse drug events (ADEs) rises rapidly. Many serious ADEs associated with polypharmacy only become known after the drugs are in use. It is impractical to test every possible drug combination during clinical trials. This issue is particularly prevalent among older adults with cardiovascular disease (CVD) where polypharmacy and ADEs are commonly observed. In this research, our primary objective was to identify key drug features to build and evaluate a model for modeling polypharmacy ADEs. Our secondary objective was to assess our model on a domain-specific case study. We developed a two-layer neural network that incorporated drug features such as molecular structure, drug-protein interactions, and mono drug side effects (DrIVeNN). We assessed DrIVeNN using publicly available side effect databases and determined Principal Component Analysis (PCA) with a variance threshold of 0.95 as the most effective feature selection method. DrIVeNN performed moderately better than state-of-the-art models like RESCAL, DEDICOM, DeepWalk, Decagon, DeepDDI, KGDDI, and KGNN in terms of AUROC for the drug-drug interaction prediction task. We also conducted a domain-specific case study centered on the treatment of cardiovascular disease (CVD). When the best performing model architecture was applied to the CVD treatment cohort, there was a significant increase in performance from the general model. We observed an average AUROC for CVD drug pair prediction increasing from 0.826 (general model) to 0.975 (CVD specific model). Our findings indicate the strong potential of domain-specific models for improving the accuracy of drug-drug interaction predictions.