Despite initial forays into clinical settings, single-cell technologies do not yet routinely inform medical decision-making. Here, we identify and categorize barriers hindering the clinical deployment of single-cell omics. We articulate a framework to identify patient subpopulations that stand to benefit from such biomarkers and outline the requirements to derive actionable clinical readouts.

Significant progress has been made in recent years in applying large language models and multimodal artificial intelligence to health and medicine, transforming diagnostics, patient interactions, and medical forecasting, although challenges like privacy, regulation, and system integration remain before widespread clinical adoption.

Cell Death Discovery, Published online: 10 July 2025; doi:10.1038/s41420-025-02615-5The transcriptomic fingerprint of cancer response to Tumor Treating Fields (TTFields)

36氪获悉，华泰证券认为，多模态大模型和应用发展的奇点将至。技术进步方面：原生多模态模型架构得到业界认可，OpenAI和Google的原生多模态模型已经在性能、延时、部署上展现出了优势。商业化进展方面：全球维度看，相当一部分AI应用公司的商业化产品依赖多模态能力。国内维度看，国内公司在视频生成赛道已经实现了较成熟的全球化和商业化之路。与市场不同的观点在于，更早认识到原生多模态架构将成为主流；AI商业化不能仅聚焦在二级公司，更要关注全球维度一级公司的进展。