Poster presented at the Thirty-Second International Joint Conference on Artificial Intelligence, 2023, Macao, SAR. https://doi.org/10.24963/ijcai.2023/554

1. SemiGNN-PPI: Self-Ensembling Multi-Graph Neural Network for
Efficient and Generalizable Protein-Protein Interaction Prediction
The 32nd International Joint Conference on Artificial Intelligence
19th-25th August 2023, Macao, S.A.R
Ziyuan Zhao1,2, Peisheng Qian1, Xulei Yang1 , Zeng Zeng3 ,
Cuntai Guan2 , Wai Leong Tam4 , Xiaoli Li1,2
1 I2R, A*STAR 2 NTU 3 SHU 4 GIS, A*STAR
Paper ID: 2877
Introduction
 Protein-protein Interactions (PPIs) are central to various cellular
functions and processes. However, PPI prediction is hampered by
 Domain Shift: DL models trained on one domain can suffer
tremendous performance degradation when evaluated on another
domain.
 Label Scarcity: Many interactions need to be annotated from
experimental data.
 Our method: We propose a simple yet effective Self-ensembling multi-
Graph Neural Network-based PPI prediction (SemiGNN-PPI) framework
for improving model efficiency and generalization.
SemiGNN-PPI Framework
This work is supported by Competitive Research
Programme “NRF-CRP22-2019-0003”, National
Research Foundation Singapore, and partially
supported by the Agency for Science, Technology
and Research (A*STAR) core funding.
[1] Zhao, Z., Zhou, F., Xu, K., Zeng, Z., Guan, C., & Zhou, S. K. LE-UDA: Label-efficient
unsupervised domain adaptation for medical image segmentation. IEEE Transactions on Medical
Imaging 2023.
[2] Zhao, Z., Zhou, F., Zeng, Z., Guan, C., & Zhou, S. Meta-hallucinator: Towards few-shot cross-
modality cardiac image segmentation. MICCAI 2022.
[3] Lv, G., Hu, Z., Bi, Y., & Zhang, S. Learning unknown from correlations: Graph neural network for
inter-novel-protein interaction prediction. In IJCAI International joint conference on artificial
intelligence, 2021
[4] Zhao, Z., Qian, P., Yang, X., Zeng, Z., Guan, C., Tam, W. L., & Li, X. SemiGNN-PPI: Self-
Ensembling Multi-Graph Neural Network for Efficient and Generalizable Protein-Protein Interaction
Prediction. In IJCAI International joint conference on artificial intelligence, 2023
References Acknowledgments Contact
For more information, please
contact: zhaoz@i2r.a-star.edu.sg
or friend me via LinkedIn or
ResearchGate.
https://jacobzhaoziyuan.github.io/
Results
 Multi-Graph Encoding
 Protein-Graph Encoding
We use proteins as nodes and PPIs as edges to build the PPI graph
and use graph neural networks (GNNs) to aggregate
representations from neighboring proteins in the PPI graph.
 Label-Graph Encoding
We model the interdependencies between different PPI types (labels)
using a graph and learn inter-dependent classifiers with Graph
Convolutional Network (GCN).
 Self-ensemble Graph Learning
 We adopt the mean teaching architecture for unsupervised learning
with two graph data augmentation methods:
 Edge Manipulation (EM)
To improve the robustness against connectivity variations, we
randomly replace a certain percentage of edges in the input to the
student and teacher models, since some edges (PPIs) between
different nodes (proteins) may be unidentified or wrong in
experimental procedures.
 Node Manipulation (NM)
To improve the robustness against attribute missing, we randomly
remove node features, mask them with zeros and feed them into
the student and teacher models respectively, to expect the model to
effectively learn the features even in the presence of missing
attribute information.
Edge Manipulation Node Manipulation
 Graph Consistency Constraint
 Edge matching: We construct the student embedding graph and the
teacher embedding graph by calculating all pairwise Pearson’s
correlation coefficient (PCC) between nodes in the same batch and
enforce consistent instance-wise correlations.
 Node matching: We formulate the edge embedding graph by
calculating all pairwise PCC between student encoding and teacher
encoding in the same batch to explicitly align encoding of the same
protein from the teacher and the student network.
 Dataset
 STRING, SHS148k, and SHS27k. The PPIs are annotated with 7
types: activation, binding, catalysis, expression, inhibition, post-
translational modification (ptmod), and reaction. Each PPI is labeled
with at least one of them.
 Comparison with Other Methods
 Label Scarcity
 Domain Shift
Performance comparison on trainset-
heterologous testsets.
DG: domain generalization
IDA: inductive domain adaptation
TDA: transductive domain adaptation