Modeling polypharmacy side effects with graph convolutional networks
By Marinka Zitnik, Monica Agrawal, Jure Leskovec. 2018
1,392 citations recorded by the source corpus.
Motivation: The use of drug combinations, termed polypharmacy, is common to treat patients with complex diseases or co-existing conditions. However, a major consequence of polypharmacy is a much higher risk of adverse side effects for the patient. Polypharmacy side effects emerge because of drug-drug interactions, in which activity of one drug may change, favorably or unfavorably, if taken with another drug. The knowledge of drug interactions is often limited because these complex relationships are rare, and are usually not observed in relatively small clinical testing. Discovering polypharmacy side effects thus remains an important challenge with significant implications for patient mortality and morbidity. Results: Here, we present Decagon, an approach for modeling polypharmacy side effects. The approach constructs a multimodal graph of protein-protein interactions, drug-protein target interactions, and the polypharmacy side effects, which are represented as drug-drug interactions, where each side effect is an edge of a different type. Decagon is developed specifically to handle such multimodal graphs with a large number of edge types. Our approach develops a new graph convolutional neural network for multirelational link prediction in multimodal networks. Unlike approaches limited to predicting simple drug-drug interaction values, Decagon can predict the exact side effect, if any, through which a given drug combination manifests clinically. Decagon accurately predicts polypharmacy side effects, outperforming baselines by up to 69%. We find that it automatically learns representations of side effects indicative of co-occurrence of polypharmacy in patients. Furthermore, Decagon models particularly well polypharmacy side effects that have a strong molecular basis, while on predominantly non-molecular side effects, it achieves good performance because of effective sharing of model parameters across edge types. Decagon opens up opportunities to use large pharmacogenomic and patient population data to flag and prioritize polypharmacy side effects for follow-up analysis via formal pharmacological studies. Availability: Source code and preprocessed datasets are at: http://snap.stanford.edu/decagon. Contact: jure@cs.stanford.edu
Subfields: Artificial Intelligence, Computational Theory and Mathematics, Molecular Biology, Graph Learning, Medical & Health AI
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Jure Leskovec
- Yoshua Bengio
- Percy Liang
- Mohit Bansal
- William L. Hamilton
- Jianfeng Gao
- Zhiyuan Liu
- Luke S. Zettlemoyer
In Computational Theory and Mathematics
- Jure Leskovec
- Marinka Zitnik
- Matthias Fey
- Rex Ying
- Michael M. Bronstein
- William L. Hamilton
- Jian Tang
- Fei Wang
In Molecular Biology
- Jure Leskovec
- Marinka Zitnik
- David T. Jones
- Rex Ying
- Arne Elofsson
- Stephan Günnemann
- Joe G. Greener
- Petar Velickovic
In Graph Learning
- Jure Leskovec
- William L. Hamilton
- Stephan Günnemann
- Matthias Fey
- Marinka Zitnik
- Jian Tang
- Michael M. Bronstein
- Rex Ying
In Medical & Health AI
- William L. Hamilton
- Marinka Zitnik
- Hossein Azizpour
- Federico Baldassarre
- Kexin Huang
- Michelle M. Li
- Jure Leskovec
- Grgoire Montavon
Citations in the corpus
- Strategies for Pre-training Graph Neural Networks (2019, background): “ by local protein-protein interaction (PPI) graphs [57, 55]. Supervised learning of graphs, especially with Graph Neural Networks (GNNs) [26, 15, 59, 53], has shown promising results in these domains [64, 56, 13, 60]. Despite the promise, there remain two key challenges in applying GNNs to these sci”
- Graph Convolutional Neural Networks for Web-Scale Recommender Systems (2018, methodology): “Following on this work, a number of authors proposed improvements, extensions, and approximations of these spectral convolutions [6, 10, 11, 13, 18, 21, 24, 29, 31], leading to new state-of-the-art results on benchmarks such as node classification, link prediction, as well as recommender system task”
- Tensor Decompositions for temporal knowledge base completion (2020, background): “Link prediction in relational data has been the subject of interest, given the widespread availability of such data and the breadth of its use in bioinformatics (Zitnik et al., 2018), recommender systems (Koren et al., 2009) or Knowledge Base completion (Nickel et al., 2016a).”
- Everything is Connected: Graph Neural Networks (2023, background): “, 2020), or drugs and drugs—predicting adverse side-effects from polypharmacy (Zitnik et al., 2018; Deac et al., 2019).”
- Graph Representation Learning (2020, background): “…and has countless real-world applications: recommending content to users in social platforms [Ying et al., 2018a], predicting drug side-effects [Zitnik et al., 2018], or inferring new facts in a relational databases [Bordes et al., 2013]—all of these tasks can be viewed as special cases of…”
- Graph Representation Learning in Biomedicine and Healthcare (2021, positive): “…their target proteins, and interactions between drugs (such as synergistic effects, where the effects of the combination are different from the contributions of the effects of each drug) 226,227 can be constructed for every cell line, yielding a collection of cell-line-specific networks 226 (Fig.”
- A Gentle Introduction to Deep Learning for Graphs (2019, background): “, drug side-effect identification [147] and drug discovery.”
- Open Graph Benchmark: Datasets for Machine Learning on Graphs (2020, background): “[107] Marinka Zitnik, Monica Agrawal, and Jure Leskovec.”
- Pre-training Graph Neural Networks (2019, background): “Supervised learning of graphs, especially with Graph Neural Networks (GNNs) [26, 15, 59, 53], has shown promising results in these domains [64, 56, 13, 60].”
- Graph convolutional networks for computational drug development and discovery (2019, background): “[40] proposed a framework based on GCNs to identify DDI further on the polypharmacy side effects level, i.e. DDI with different types, providing additional guidance on drug combinational treatments.”
- Dual-Primal Graph Convolutional Networks (2018, background): “Geometric deep learning has been successfully employed in a broad spectrum of applications, ranging from computer graphics, vision [31, 6, 43], and medicine [37, 48] to chemistry [12, 14] and high energy physics [21].”
- Graph embedding on biomedical networks: methods, applications and evaluations (2019, positive): “Besides, graph-based autoencoder is introduced for DDIs prediction (Zitnik et al., 2018; Ma et al., 2018) whose intuition is similar to GAE (Kipf and Welling, 2016).”
- Aspect-Level Sentiment Analysis Via Convolution over Dependency Tree (2019, neutral): “Several AI applications such as link prediction (Schlichtkrull et al., 2018; Zitnik et al., 2018; Kong et al., 2019), semantic role labeling (Marcheggiani and Titov, 2017), and relation extraction (Zhang et al., 2018) have successfully exploited GCNs to improve representation learning.”
- Graph Neural Networks with Heterophily (2020, background): “…data, Graph Neural Network (GNN) models have gained wide research interest (Scarselli et al. 2008) and been adapted in applications including recommendation systems (Ying et al. 2018), bioinformatics (Zitnik, Agrawal, and Leskovec 2018; Yan et al. 2019), fraud detection (Dou et al. 2020), and more.”
- Embedding Logical Queries on Knowledge Graphs (2018, background): “Drug-drug interaction network contains 29 different types of edges (one for each type of polypharmacy side effects) and describes which drug pairs lead to which side effects [49].”
- Graph Neural Networks: A Review of Methods and Applications (2018, background): “[108] also suggests a GCN based model for polypharmacy side effects prediction.”
- AWB-GCN: A Graph Convolutional Network Accelerator with Runtime Workload Rebalancing (2019, background): “…number of real-world applications [8], including electric grid cascading failure analysis [32], prediction of chemical reactivity [33], prediction of synthesized material properties [34], polypharmacy side-effect modeling [35], accurate advertisement in E-commerce [36], and cybersecurity [37].”
- Higher-Order Explanations of Graph Neural Networks via Relevant Walks (2020, methodology): “Graph neural networks (GNNs) [1], [2] have been proposed as a method to learn from observations in general graph structures and have found use in an ever growing number of applications [3], [4], [5], [6], [7], [8].”