Abstract: This talk will present SpatialAgent, a multimodal llm agent, designed to enhance key stages in spatial transcriptomics: (1) experiment design, (2) multi-scale data annotation, and (3) interpretation and hypothesis generation. These tasks are critical for advancing autonomous intelligence in the field of spatial biology, which holds significant therapeutic potential for many cancers and neurodegenerative diseases like Alzheimer's. By collecting and evaluating the performance of approximately 15 domain scientists across these tasks, the authors also developed a framework that optimizes agent performance through human collaboration.

Bio: Hanchen Wang is a postdoc with Jure Leskovec at Stanford CS and Aviv Regev at Genentech. His leading work is published in Nature, Nature Biotechnology, Nature Machine Intelligence & NeurIPS. He also published and co-organized events on "AI for Science" in NeurIPS, ICML & ICLR. He had a CS Ph.D from Cambridge with Joan Lasenby from Trinity College; a Physics BS from Nanjing with Xinran Wang, where he was the commencement speaker. He secured university admission at age 15 while at Suzhou High School. He also spent time in finance, tech & biotech industries. He's a national athlete in 400m.