EUVIMEDOpdag EUVIMED
Open access

Improved Method for Predicting GPCR – GPCR Interaction Pairs

Aoi Fukushima, Arina Ouma, Hiroki Teruse, Hiroyuki Toh, Wataru Nemoto

Proteins: Structure, Function, and Bioinformatics · 2026-09 · 94 (9) · 1576-1596

DOI: 10.1002/prot.70138 ↗

At a glance

Journal article in Proteins: Structure, Function, and Bioinformatics, published online on 2026-04-10. Written by Aoi Fukushima and 4 co-authors. Institution: Master's Programs in Life Science and Engineering, Graduate School of Science and Engineering, Tokyo Denki University (TDU), Saitama, Japan. Major MEDLINE topics: Receptors, G-Protein-Coupled, Computational Biology. 85 references. The full text is freely available in PubMed Central. Funded by: Japan Society for the Promotion of Science.

AbstractSubjectsBibliografiske oplysningerCiteRelated studies

Abstract (original language)

G protein–coupled receptors (GPCRs) can form oligomers, which activate distinct signaling pathways compared to monomeric GPCRs. Oligomerization influences GPCR trafficking, ligand affinity, and signal transduction, and has been implicated in diseases such as schizophrenia and hypertension. Understanding GPCR oligomerization is essential for uncovering disease mechanisms and developing new therapeutic strategies. Previously, we developed a GPCR–GPCR interaction pair predictor (GGIP) utilizing an SVM algorithm. Features for the predictor were generated by quantifying amino acid properties, assigning scores, and averaging these values across the target sequences. In this study, we aimed to enhance the predictive accuracy of GGIP. We evaluated four methods by combining two feature generation techniques with two prediction algorithms. We tested the sequence segmentation‐based method from our previous work and automatic feature generation using amino acid sequences with an autoencoder while evaluating both the SVM and gradient‐boosting decision tree (GBDT) as prediction algorithms. Combining segmentation‐based feature generation with GBDT yielded the highest performance, achieving an AUROC of over 0.98. Some features could be identified as a basis for predicting that a pair of GPCRs would interact, based on amino acid properties and their arrangement in the three‐dimensional structure. Integrating our improved method with disease‐related gene expression variation data revealed a significant association between GPCR interaction pairs and the presence of disease‐related differentially expressed genes (DEGs). Specifically, around 90% of experimentally determined interaction pairs contained at least one protomer gene classified as a disease‐related DEG, suggesting that GPCRs forming interaction pairs are more likely to be associated with disease‐related gene expression changes. Among these pairs, we identified the interaction between mGluR2 and 5‐HT2AR, which has been postulated to be linked to schizophrenia. Although this association was not registered in the database, we were able to confirm it through published literature. Given the significant association with disease‐related DEGs, this approach is critical for identifying disease‐associated GPCR interaction pairs and guiding future therapeutic developments.

Subjects

MeSH terms (MEDLINE)

Receptors, G-Protein-CoupledComputational BiologySupport Vector MachineHumansPrediction AlgorithmsBoosting Machine Learning AlgorithmsProtein BindingAutoencoderAmino Acid SequenceAlgorithms

Author keywords

SHAPautoencoderbioinformaticsgradient‐boosting decision treeprotein–protein interaction

Substances

Receptors, G-Protein-Coupled

Authors and institutions

  • Master's Programs in Life Science and Engineering, Graduate School of Science and Engineering, Tokyo Denki University (TDU), Saitama, Japan.
  • Division of Life Science, Department of Science and Engineering, School of Science and Engineering, Tokyo Denki University (TDU), Saitama, Japan.
  • School of Biological and Environmental Sciences, Kwansei Gakuin University, Sanda-shi, Hyogo, Japan.

Funding

  • Japan Society for the Promotion of Science

Cite

APA

Fukushima, A., Ouma, A., Teruse, H., Toh, H., & Nemoto, W. (2026). Improved Method for Predicting GPCR – GPCR Interaction Pairs. Proteins: Structure, Function, and Bioinformatics, 94(9), 1576-1596. https://doi.org/10.1002/prot.70138

Vancouver

Fukushima A, Ouma A, Teruse H, Toh H, Nemoto W. Improved Method for Predicting GPCR – GPCR Interaction Pairs. Proteins: Structure, Function, and Bioinformatics. 2026;94(9):1576-1596. doi:10.1002/prot.70138

BibTeX
@article{Fukushima2026,
  author = {Fukushima, Aoi and Ouma, Arina and Teruse, Hiroki and Toh, Hiroyuki and Nemoto, Wataru},
  title = {Improved Method for Predicting GPCR – GPCR Interaction Pairs},
  journal = {Proteins: Structure, Function, and Bioinformatics},
  year = {2026},
  volume = {94},
  number = {9},
  pages = {1576--1596},
  doi = {10.1002/prot.70138}
}
Start for free →