Scholarlink

<p>Large language models (LLMs) have demonstrated impressive capabilities in summarization, reasoning, and content generation, yet their inability to directly interpret large-scale omics data has limited their utility in data-driven hypothesis generation—particularly in mechanism discovery that demands the integration and interpretation of multimodal datasets, heterogeneous models, and deep domain expertise. Conversely, traditional computational algorithms excel at quantitative analysis of omics data but often rely heavily on labor-intensive, expert-driven interpretation to extract biologically meaningful insights. Here, we introduce (<i>cointelligent single-cell spatial cell‒cell communication</i>: iS2C2), a novel cointelligent platform that synergizes mathematically rigorous computational algorithms with the contextual reasoning capabilities of LLMs to automatically generate biologically interpretable hypotheses from single-cell RNA-seq and spatial transcriptomics data. The iS2C2 platform incorporates a transparent and reproducible cell–cell communication analysis pipeline built upon mathematically rigorous algorithms designed to enhance interpretability for integration with LLMs that contextualize algorithmic outputs or predictions using domain-specific knowledge and literature-derived evidence. When applied to Alzheimer’s disease and cancer datasets, iS2C2 generated accurate, reproducible, and expert-validated hypotheses, unveiling previously unrecognized signaling pathways and mechanistic insights in disease microenvironments. This cointelligent approach bridges the gap between structured computational analysis and generative reasoning, heralding a paradigm shift toward fully automated, interpretable biological discovery and advancing the frontiers of next-generation precision medicine and systems biology.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

iS2C2: a cointelligent platform for mechanistic discovery of disease cellular crosstalk

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月11日

查看全文
  • Jianting Sheng,
  • Ju Young Ahn,
  • Li Yang,
  • Zhihao Wan,
  • Shaohua Qi,
  • Xiaohui Yu,
  • Zhan Xu,
  • Yuliang Cao,
  • Matthew Vasquez,
  • Amna Irfan,
  • Yuanyuan Zhu,
  • Hong Zhao,
  • Zheng Yin,
  • Ying Zhu,
  • Yunfeng Ding,
  • Alireza Faridar,
  • Lin Wang,
  • Fengshuo Liu,
  • Hongxia Wang,
  • Zhigang Ji,
  • Dongxue Mao,
  • Michael Chan,
  • Daniel Kermany,
  • Wenjuan Dong,
  • Doo Yeon Kim,
  • Xiang H.-F. Zhang,
  • Stephen T. C. Wong

摘要

Large language models (LLMs) have demonstrated impressive capabilities in summarization, reasoning, and content generation, yet their inability to directly interpret large-scale omics data has limited their utility in data-driven hypothesis generation—particularly in mechanism discovery that demands the integration and interpretation of multimodal datasets, heterogeneous models, and deep domain expertise. Conversely, traditional computational algorithms excel at quantitative analysis of omics data but often rely heavily on labor-intensive, expert-driven interpretation to extract biologically meaningful insights. Here, we introduce (cointelligent single-cell spatial cell‒cell communication: iS2C2), a novel cointelligent platform that synergizes mathematically rigorous computational algorithms with the contextual reasoning capabilities of LLMs to automatically generate biologically interpretable hypotheses from single-cell RNA-seq and spatial transcriptomics data. The iS2C2 platform incorporates a transparent and reproducible cell–cell communication analysis pipeline built upon mathematically rigorous algorithms designed to enhance interpretability for integration with LLMs that contextualize algorithmic outputs or predictions using domain-specific knowledge and literature-derived evidence. When applied to Alzheimer’s disease and cancer datasets, iS2C2 generated accurate, reproducible, and expert-validated hypotheses, unveiling previously unrecognized signaling pathways and mechanistic insights in disease microenvironments. This cointelligent approach bridges the gap between structured computational analysis and generative reasoning, heralding a paradigm shift toward fully automated, interpretable biological discovery and advancing the frontiers of next-generation precision medicine and systems biology.