EUVIMEDEUVIMED entdecken
VergleichsstudieOpen Access · CC BY 4.0

Graph Neural Networks vs. Traditional QSAR: A Comprehensive Comparison for Multi-Label Molecular Odor Prediction

Tengteng Wen iD, Xianfa Cai, Jincheng Li

Molecules · 30.11.2025 · 30 (23) · 4605

DOI: 10.3390/molecules30234605 ↗

Kurz gefasst

Vergleichsstudie in der Zeitschrift Molecules, online erschienen am 30.11.2025. Verfasst von Tengteng Wen und 2 weiteren Autor:innen. Beteiligte Einrichtung: Guangdong College of Medical Information Engineering, Pharmaceutical University, Guangzhou 510006, China. Hauptthemen laut MEDLINE: Quantitative Structure-Activity Relationship, Neural Networks, Computer, Odorants. Die Arbeit wurde bisher 14-mal zitiert (Crossref) und nennt 26 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Guangzhou Science, Technology and Innovation Commission.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Molecular odor prediction represents a fundamental challenge in computational chemistry with significant applications in fragrance design, food science, and chemical safety assessment. While traditional Quantitative Structure–Activity Relationship (QSAR) methods rely on hand-crafted molecular descriptors, recent advances in graph neural networks (GNNs) enable direct end-to-end learning from molecular graph structures. However, systematic comparison between these approaches for multi-label odor prediction remains limited. This study presents a comprehensive evaluation of traditional QSAR methods compared with modern GNN approaches for multi-label molecular odor prediction. Using the GoodScent dataset containing 3304 molecules with six high-frequency odor types (fruity, green, sweet, floral, woody, herbal), we systematically evaluate 23 model configurations across traditional machine learning algorithms (Random Forest, SVM, GBDT, MLP, XGBoost, LightGBM) with three feature-processing strategies and three GNN architectures (GCN, GAT, NNConv). The results demonstrate that GNN models achieve significantly superior performance, with GCN achieving the highest macro F1-score of 0.5193 compared to 0.4766 for the best traditional method (MLP with basic preprocessing), representing a 24.1% relative improvement. Critically, we discover that threshold optimization is essential for multi-label chemical classification. These findings establish GNNs as the preferred approach for molecular property prediction tasks and provide crucial insights for handling class imbalance in chemical informatics applications.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Quantitative Structure-Activity RelationshipNeural Networks, ComputerOdorantsMachine LearningAlgorithmsGraph Neural Networks

Stichwörter der Autor:innen

graph neural network (GNN)molecular odor predictionmulti-label classificationthreshold optimization

Autor:innen und Institutionen

  • Guangdong College of Medical Information Engineering, Pharmaceutical University, Guangzhou 510006, China.

ORCID-Profile: Tengteng Wen

Förderung

  • Guangzhou Science, Technology and Innovation Commission

Zitieren

APA

Wen, T., Cai, X., & Li, J. (2025). Graph Neural Networks vs. Traditional QSAR: A Comprehensive Comparison for Multi-Label Molecular Odor Prediction. Molecules, 30(23), 4605. https://doi.org/10.3390/molecules30234605

Vancouver

Wen T, Cai X, Li J. Graph Neural Networks vs. Traditional QSAR: A Comprehensive Comparison for Multi-Label Molecular Odor Prediction. Molecules. 2025;30(23):4605. doi:10.3390/molecules30234605

BibTeX
@article{Wen2025,
  author = {Wen, Tengteng and Cai, Xianfa and Li, Jincheng},
  title = {Graph Neural Networks vs. Traditional QSAR: A Comprehensive Comparison for Multi-Label Molecular Odor Prediction},
  journal = {Molecules},
  year = {2025},
  volume = {30},
  number = {23},
  pages = {4605},
  doi = {10.3390/molecules30234605}
}
Kostenlos starten →