Annu Int Conf IEEE Eng Med Biol Soc. 2021 Nov;2021:4458-4464. doi: 10.1109/EMBC46164.2021.9630181.
Computational modeling is crucial for understanding and analyzing complex systems. In biology, model creation is a human dependent task that requires reading hundreds of papers and conducting wet lab experiments, which would take days or months. To overcome this hurdle, we propose a novel automated method, that utilizes the knowledge published in literature to suggest model extensions by selecting most relevant and useful information in few seconds. In particular, our novel approach organizes the events extracted from the literature as a collaboration graph with additional metric that relies on the event occurrence frequency in literature. Additionally, we show that common graph centrality metrics vary in the assessment of the extracted events. We have demonstrated the reliability of the proposed method using three different selected models, namely, T cell differentiation, T cell large granular lymphocyte, and pancreatic cancer cell. Our proposed method was able to find high percent of the desired new events with an average recall of 82%.
计算建模对于理解和分析复杂系统至关重要。在生物学中，模型创建是一项依赖于人的任务，需要阅读数百篇论文并进行湿实验室实验，这可能需要数天或数月的时间。为了克服这一障碍，我们提出了一种新颖的自动化方法，该方法利用文献中发表的知识，通过在几秒钟内选择最相关和最有用的信息来建议模型扩展。具体来说，我们的新方法将从文献中提取的事件组织成一个协作图，并使用额外的基于文献中事件发生频率的度量标准。此外，我们还表明，常见的图中心性度量在提取事件的评估中存在差异。我们使用三个不同的选定模型（即 T 细胞分化、T 细胞大颗粒淋巴细胞和胰腺癌细胞）展示了所提出方法的可靠性。我们的方法能够找到高比例的所需新事件，平均召回率为 82%。
