Generating novel molecular structures with pharmacological activity remains a fundamental challenge in drug discovery. Traditional drug development approaches, heavily reliant on manual labor, are characterized by time-consuming processes, substantial costs, and limited exploration efficiency. To address these limitations, we propose a Graph-based Molecular Generation System (GAMS) that integrates decoder architecture with graph structures to effectively capture molecular structural features, physicochemical properties, and pharmacological characteristics, enabling rapid generation and design of drug-like molecules. Our innovation lies in the development of a dynamic property-guided decoding strategy that periodically incorporates drug property predictions during the generation process, enhancing the drug-likeness of generated molecules. Experimental results demonstrate that GAMS outperforms some existing methods in terms of quantitative drug-likeness scores (QED) and the proportion of drug-like molecules, validating its effectiveness in molecular design and optimization.

Jindong Sha. GAMS: Graph-Based Autoregressive Molecular Generation System for Drug Discovery. Frontiers in Medical Science Research (2025), Vol. 7, Issue 6: 32-40. https://doi.org/10.25236/FMSR.2025.070604.

[1] J. A. DiMasi, H. G. Grabowski, and R. W. Hansen, “Innovation in the pharmaceutical industry: New estimates of r&d costs,” Journal of health economics, vol. 47, pp. 20–33, 2016.

[2] N. Berdigaliyev and M. Aljofan, “An overview of drug discovery and development,” Future medicinal chemistry, vol. 12, no. 10, pp. 939–947, 2020.

[3] D. A. Pereira and J. A. Williams, “Origin and evolution of high throughput screening,” British journal of pharmacology, vol. 152, no. 1, pp. 53–61, 2007.

[4] A. Lavecchia, “Deep learning in drug discovery: Opportunities, challenges and future prospects,” Drug discovery today, vol. 24, no. 10, pp. 2017–2032, 2019.

[5] R. Gupta, D. Srivastava, M. Sahu, S. Tiwari, R. K. Ambasta, and P. Kumar, “Artificial intelligence to deep learning: Machine intelligence approach for drug discovery,” Molecular diversity, vol. 25, pp. 1315–1360, 2021.

[6] J. Arús-Pous et al., “Randomized SMILES strings improve the quality of molecular generative models,” Journal of cheminformatics, vol. 11, pp. 1–13, 2019.

[7] J. Arús-Pous et al., “SMILES-based deep generative scaffold decorator for de-novo drug design,” Journal of cheminformatics, vol. 12, pp. 1–18, 2020.

[8] C.-K. Wu, X.-C. Zhang, Z.-J. Yang, A.-P. Lu, T.-J. Hou, and D.-S. Cao, “Learning to SMILES: BAN-based strategies to improve latent representation learning from molecules,” Briefings in bioinformatics, vol. 22, no. 6, p. bbab327, 2021.

[9] B. Winter, C. Winter, J. Schilling, and A. Bardow, “A smile is all you need: Predicting limiting activity coefficients from SMILES with natural language processing,” Digital Discovery, vol. 1, no. 6, pp. 859–869, 2022.