Tewari Jeevan, Dahl Benjamin W, Saucerman Jeffrey J
co-equal contributors.
Department of Biomedical Engineering, University of Virginia.
bioRxiv. 2025 Jul 29:2025.07.28.667217. doi: 10.1101/2025.07.28.667217.
Computational models of signaling networks provide frameworks for predicting how molecular cues guide cell decisions. But they are typically limited by manual curation from incomplete literature. Here, we test whether general-purpose large language models (LLMs) generate accurate models of signaling networks. We find that general purpose LLMs generate 24-58% of the reactions of literature-curated networks for cardiomyocyte hypertrophy, myofibroblast activation, and mechano-signaling, and predicting network responses to perturbations with accuracies of 5-26%. While current general-purpose LLMs generate signaling networks with limited accuracy, this study provides a pipeline and benchmarks to guide future improvements.
信号网络的计算模型为预测分子信号如何引导细胞决策提供了框架。但它们通常受到来自不完整文献的人工整理的限制。在这里，我们测试通用大语言模型（LLMs）是否能生成准确的信号网络模型。我们发现，通用大语言模型生成了文献整理网络中24%-58%的心肌细胞肥大、肌成纤维细胞激活和机械信号反应，并以5%-26%的准确率预测网络对扰动的反应。虽然目前的通用大语言模型生成信号网络的准确性有限，但这项研究提供了一个流程和基准，以指导未来的改进。
