Department of Chemistry Seminar - From Generative AI to Scientific Agents: Towards AI-Native Drug Discovery and Development
Supporting the below United Nations Sustainable Development Goals:支持以下聯合國可持續發展目標：支持以下联合国可持续发展目标：
Speaker: Prof. Yang ZHANG
Institution(s): Tianjin University, Tianjin, China
Shenzhen Loop Area Institute, Shenzhen, China
Hosted by: Prof. Haibin SU
Abstract
Artificial intelligence (AI) is transforming drug discovery and development, with advances in generative modeling, biomolecular design, and experimental data analysis. However, translating these capabilities into effective research workflows remains challenging, as molecular discovery, experimental validation, and bioprocess development require the integration of heterogeneous data, domain-specific knowledge, computational models, and experimental systems. Scientific agents offer an emerging approach to bridging these capabilities through coordinated reasoning, tool use, and iterative experimental feedback.
This seminar examines the progression from individual AI models to integrated scientific agents across three interconnected research areas. In biomolecular discovery, the GRAPE-LM framework( Nature Biotechnology), combines nucleic acid language modeling with single-round functional screening to enable the generative design and experimental validation of RNA aptamers. In intelligent experimentation, AI-driven microscopy, laboratory automation, and multi-agent coordination extend computational intelligence toward active data acquisition, instrument control, and experimental feedback. In biopharmaceutical process development, scientific agents integrate process knowledge, experimental data, and domain-specific modeling approaches to support evidence-based analysis, research planning, and optimization in filtration and chromatography.
These advances provide a foundation for AI-native research systems that connect molecular design, experimental workflows, and bioprocess development within a unified computational framework. By integrating specialized scientific agents, laboratory instruments, research data, and human expertise, such systems aim to enable traceable execution, continuous knowledge accumulation, and adaptive research workflows. This emerging paradigm represents a shift from task-specific AI applications toward coordinated, iterative, and increasingly autonomous scientific research in drug discovery and development.