Kimyasal Bileşiklerin Tat Özelliklerinin GNN Tabanlı Modellerle Sınıflandırılması

Yıl 2025, Cilt: 6 Sayı: 1, 28 - 34, 30.06.2025

Ulvi İsgandarli , Eyyüp Gülbandılar

https://doi.org/10.53608/estudambilisim.1700516

Öz

Özet: Bu çalışma, kimyasal bileşiklerin tat özelliklerinin (acı veya tatlı) moleküler yapılarından tahmin edilmesini ele almaktadır. Bu amaçla, bileşiklerin kanonik SMILES gösterimlerinden moleküler grafikler oluşturulmuş ve bu grafikler üzerinde GraphSAGE, Graph Convolutional Network (GCN) ve Graph Attention Network (GAT) olmak üzere üç farklı Graph Neural Network (GNN) mimarisi uygulanmıştır. Modeller, Kaggle platformundan elde edilen bir veri seti üzerinde eğitilmiş ve doğruluk, kesinlik, duyarlılık, özgüllük ve F1 skoru gibi metrikler kullanılarak değerlendirilmiştir. Sonuçlar, GNN tabanlı yaklaşımların tat sınıflandırma görevinde etkili olduğunu ve özellikle GraphSAGE modelinin %90.09 doğruluk ve 0.8772 F1 skoru ile en iyi performansı sergilediğini ortaya koymuştur. Bu bulgular, GNN tabanlı modellerin özellikle tat ile ilişkili bileşiklerin hızlı taranması ve değerlendirilmesi gibi görevlerde ilaç keşfi ve gıda bilimi gibi alanlara katkı sağlayabileceğini göstermektedir.

Anahtar Kelimeler

Kimyasal Bileşikler , Tat Sınıflandırması , Grafik Sinir Ağları , Grafik Evrişimli Ağlar , Grafik Dikkat Ağları

Classification of Taste Properties of Chemical Compounds with GNN-Based Models

Öz

This study focuses on predicting the taste properties (bitter or sweet) of chemical compounds from their molecular structures. To achieve this, molecular graphs were generated from the canonical SMILES representations of the compounds, and three different Graph Neural Network (GNN) architectures GraphSAGE, Graph Convolutional Network (GCN), and Graph Attention Network (GAT) were applied to these graphs. The models were trained on a dataset obtained from the Kaggle platform and evaluated using metrics such as accuracy, precision, sensitivity, specifity and F1 score. The results demonstrated that GNN-based approaches are effective in taste classification tasks, with the GraphSAGE model showing the best performance, achieving an accuracy of 90.09% and an F1 score of 0.8772. These findings suggest that GNN-based models can contribute to fields such as drug discovery and food science, especially in tasks like the rapid screening and evaluation of compounds related to taste.

Anahtar Kelimeler

Chemical Compounds , Taste Classification , Graph Neural Networks , Graph Convolutional Networks , Graph Attention Networks

Kaynakça

• Schieberle, P., & Hofmann, T. (2011). Mapping the combinatorial code of food flavors by means of molecular sensory science approach. In Chemical, Sensory and Technological Properties (pp. 413-438). Informa UK Limited.
• Schwartz, C., Chabanet, C., Lange, C., Issanchou, S., & Nicklaus, S. (2011). The role of taste in food acceptance at the beginning of complementary feeding. Physiology & behavior, 104(4), 646-652.
• Mennella, J. A., Spector, A. C., Reed, D. R., & Coldwell, S. E. (2013). The bad taste of medicines: overview of basic research on bitter taste. Clinical therapeutics, 35(8), 1225-1246.
• Weininger, D. (1988). SMILES, a chemical language and information system. 1. Introduction to methodology and encoding rules. Journal of chemical information and computer sciences, 28(1), 31-36.
• O’Boyle, N. M. (2012). Towards a Universal SMILES representation-A standard method to generate canonical SMILES based on the InChI. Journal of cheminformatics, 4, 1-14.
• Wang, Y., Li, Z., & Barati Farimani, A. (2023). Graph neural networks for molecules. In Machine learning in molecular sciences (pp. 21-66). Cham: Springer International Publishing.
• Song, R., Liu, K., He, Q., He, F., & Han, W. (2024). Exploring bitter and sweet: the application of large language models in molecular taste prediction. Journal of Chemical Information and Modeling, 64(10), 4102-4111.
• Song, Y., Chang, S., Tian, J., Pan, W., Feng, L., & Ji, H. (2023). A comprehensive comparative analysis of deep learning based feature representations for molecular taste prediction. Foods, 12(18), 3386.
• Bo, W., Qin, D., Zheng, X., Wang, Y., Ding, B., Li, Y., & Liang, G. (2022). Prediction of bitterant and sweetener using structure-taste relationship models based on an artificial neural network. Food Research International, 153, 110974.
• Pallante, L., Cannariato, M., Vezzulli, F., Malavolta, M., Lambri, M., & Deriu, M. A. (2023). Machine learning aided molecular modelling of taste to identify food fingerprints. Chemical Engineering Transactions, 102(2023), 283-288.
• Ramanathan, V., & DN, S. S. (2025). Predicting Molecular Taste: Multi-Label and Multi-Class Classification. bioRxiv, 2025-05.
• Androutsos, L., Pallante, L., Bompotas, A., Stojceski, F., Grasso, G., Piga, D., ... & Mavroudi, S. (2024). Predicting multiple taste sensations with a multiobjective machine learning method. npj Science of Food, 8(1), 47.
• Landrum, G. (2013). Rdkit documentation. Release, 1(1-79), 4.
• Farmer, R. (2022). Classify the bitter or sweet taste of compounds [Veri seti]. Kaggle. https://doi.org/10.34740/KAGGLE/DSV/4234193
• Fey, M., & Lenssen, J. E. (2019). Fast graph representation learning with PyTorch Geometric. arXiv. https://arxiv.org/abs/1903.02428
• Hamilton, W., Ying, Z., & Leskovec, J. (2017). Inductive representation learning on large graphs. Advances in neural information processing systems, 30.
• Veličković, P., Cucurull, G., Casanova, A., Romero, A., Lio, P., & Bengio, Y. (2017). Graph attention networks. arXiv preprint arXiv:1710.10903.
• Kipf, T. N., & Welling, M. (2016). Semi-supervised classification with graph convolutional networks. arXiv preprint arXiv:1609.02907.
• Yacouby, R., & Axman, D. (2020). Probabilistic extension of precision, recall, and f1 score for more thorough evaluation of classification models. In Proceedings of the first workshop on evaluation and comparison of NLP systems (pp. 79-91).
• Glaros, A. G., & Kline, R. B. (1988). Understanding the accuracy of tests with cutting scores: The sensitivity, specificity, and predictive value model. Journal of clinical psychology, 44(6), 1013-1023.
• Oh, J., Cho, K., & Bruna, J. (2019). Advancing graphsage with a data-driven node sampling. arXiv preprint arXiv:1904.12935.
• Zitnik, M., Agrawal, M., & Leskovec, J. (2018). Modeling polypharmacy side effects with graph convolutional networks. Bioinformatics, 34(13), i457-i466.
• Hu, W., Liu, B., Gomes, J., Zitnik, M., Liang, P., Pande, V., & Leskovec, J. (2019). Strategies for pre-training graph neural networks. arXiv preprint arXiv:1905.12265.