MR-GNN: Multi-Resolution and Dual Graph Neural Network for Predicting
Structured Entity Interactions
Nuo Xu
1
, Pinghui Wang
2,1∗†
, Long Chen
1
, Jing Tao
1
and Junzhou Zhao
1∗
1
MOE NEKEY Lab, Xi’an Jiaotong University, China
2
Shenzhen Research School, Xi’an Jiaotong University, China
nxu@sei.xjtu.edu.cn, {phwang, jtao}@mail.xjtu.edu.cn, chenlongche@stu.edu.cn,
junzhouzhao@gmail.com
Abstract
Predicting interactions between structured entities
lies at the core of numerous tasks such as drug reg-
imen and new material design. In recent years,
graph neural networks have become attractive.
They represent structured entities as graphs, and
then extract features from each individual graph us-
ing graph convolution operations. However, these
methods have some limitations: i) their networks
only extract features from a fix-sized subgraph
structure (i.e., a fix-sized receptive field) of each
node, and ignore features in substructures of differ-
ent sizes, and ii) features are extracted by consider-
ing each entity independently, which may not effec-
tively reflect the interaction between two entities.
To resolve these problems, we present MR-GNN,
an end-to-end graph neural network with the fol-
lowing features: i) it uses a multi-resolution based
architecture to extract node features from differ-
ent neighborhoods of each node, and, ii) it uses
dual graph-state long short-term memory networks
(LSTMs) to summarize local features of each graph
and extracts the interaction features between pair-
wise graphs. Experiments conducted on real-world
datasets show that MR-GNN improves the predic-
tion of state-of-the-art methods.
1 Introduction
A large variety of applications require understanding the in-
teractions between structured entities. For example, when
one medicine is taken together with another, each medicine’s
intended efficacy may be altered substantially (see Fig. 1).
Understanding their interactions is important to minimize the
side effects and maximize the synergistic benefits
[
Ryu et
al., 2018
]
. In chemistry, understanding what chemical reac-
tions will occur between two chemicals is helpful in design-
ing new materials with desired properties
[
Kwon and Yoon,
2017
]
. Despite its importance, examining all interactions by
performing clinical or laboratory experiments is impractical
∗
Corresponding Authors
†
Nuo Xu and Pinghui Wang contributed equally to this work.
Molecular formula
a combined medication scheme
Whether Allopurinol would increase
the risk of a hypersensitivity reaction
to Amoxicillin ?
Bad scheme Not bad scheme
No
Allopurinol
Amoxicillin
Interaction
prediction
Represent as graphs
Yes
N
O
N
NH
N
H
OH
N
H
O
HO
N
S
O
O
NH
2
Figure 1: Overview of graph-based framework. We transform two
drugs Allopurinol and Amoxicillin into graphs, where nodes repre-
sent atoms and edges refer to chemical bonds between atoms, and
predict interactions between them. When there exists an adverse re-
action between them, they cannot be taken together.
due to the potential harms to patients and also highly time
and monetary costs.
Recently, machine learning methods have been proposed to
address this problem, and they are demonstrated to be effec-
tive in many tasks
[
Duvenaud et al., 2015; Li et al., 2017;
Tian et al., 2016; Ryu et al., 2018
]
. These methods use
features extracted from entities to train a classifier to pre-
dict entity interactions. However, features have to be care-
fully provided by domain experts
[
Ryu et al., 2018; Tian
et al., 2016
]
, and it is labor-intensive. To automate feature
extraction, graph convolution neural networks (GCNs) have
been proposed
[
Alex et al., 2017; Kwon and Yoon, 2017;
Zitnik et al., 2018
]
. GCNs represent structured entities as
graphs, and use graph convolution operators to extract fea-
tures. One of the state-of-the-art GCN models, proposed by
Alex et al.
[
2017
]
, extracts features from the 3-hop neigh-
borhood of each node. We thus say that their model uses a
fix-sized receptive field (RF). However, using a fix-sized RF
to extract features may have limitations, which can be illus-
trated by the following example.
Example 1. Figure 2 shows two weak acids, i.e., Hydro-
quinone and Acetic acid. They are weak acids due to the
existence of substructures phenolic hydroxyl (ArOH) and car-
boxyl (COOH), respectively. Representing these two chemi-
cal compounds as graphs, we need a three-hop neighborhood
to accurately extract ArOH from Hydroquinone, and a two-
hop neighborhood to accurately extract COOH from Acetic
acid. While using a fix-sized neighborhood will result in that
either incomplete substructures being extracted (i.e., RF is
arXiv:1905.09558v1 [cs.LG] 23 May 2019