The integration of Artificial Intelligence (AI) into pharmaceutical research has accelerated drug discovery by streamlining target identification, compound screening, and candidate optimization. Yet, developing effective AI-driven systems remains challenging due to high- dimensional data and the need for expert-driven model tuning. Automated Machine Learning (AutoML) mitigates these issues by automating key stages of the ML pipeline data preprocessing, model selection, hyperparameter tuning, and evaluation thereby enhancing scalability and reducing dependence on domain expertise. AutoML's ability to handle diverse datasets, from omics to lipid nanoparticle (LNP) optimization enabling faster translation of candidates into clinically viable therapeutics. Techniques such as transfer learning, graph neural networks, and transformer-based models enrich molecular representations and improve predictive performance. Advances like DNN-VS further bolster tasks such as virtual screening and bioactivity prediction. Nonetheless, challenges persist in achieving model interpretability and integrating multimodal data. Future efforts should focus on adaptive, domain-aware AutoML strategies to overcome these limitations.
Sena Aydin (Wed,) studied this question.