Yeoh Jing Wui, Ng Kai Boon Ivan, Teh Ai Ying, Zhang JingYun, Chee Wai Kit David, Poh Chueh Loo
Department of Biomedical Engineering, Faculty of Engineering , National University of Singapore , Singapore 119077.
NUS Synthetic Biology for Clinical and Technological Innovation (SynCTI), Life Sciences Institute , National University of Singapore , Singapore 119077.
ACS Synth Biol. 2019 Jul 19;8(7):1484-1497. doi: 10.1021/acssynbio.8b00523. Epub 2019 May 7.
Constructing a complex functional gene circuit composed of different modular biological parts to achieve the desired performance remains challenging without a proper understanding of how the individual module behaves. To address this, mathematical models serve as an important tool toward better interpretation by quantifying the performance of the overall gene circuit, providing insights, and guiding the experimental designs. As different gene circuits might require exclusively different mathematical representations in the form of ordinary differential equations to capture their transient dynamic behaviors, a recurring challenge in model development is the selection of the appropriate model. Here, we developed an automated biomodel selection system (BMSS) which includes a library of pre-established models with intuitive or unintuitive features derived from a vast array of expression profiles. Selection of models is built upon the Akaike information criteria (AIC). We tested the automated platform using characterization data of routinely used inducible systems, constitutive expression systems, and several different logic gate systems (NOT, AND, and OR gates). The BMSS achieved a good agreement for all the different characterization data sets and managed to select the most appropriate model accordingly. To enable exchange and reproducibility of gene circuit design models, the BMSS platform also generates Synthetic Biology Open Language (SBOL)-compliant gene circuit diagrams and Systems Biology Markup Language (SBML) output files. All aspects of the algorithm were programmed in a modular manner to ease the efforts on model extensions or customizations by users. Taken together, the BMSS which is implemented in Python supports users in deriving the best mathematical model candidate in a fast, efficient, and automated way using part/circuit characterization data.
在没有正确理解单个模块行为的情况下，构建由不同模块化生物部件组成的复杂功能基因电路以实现所需性能仍然具有挑战性。为了解决这个问题，数学模型作为一种重要工具，通过量化整个基因电路的性能、提供见解和指导实验设计，有助于更好地进行解释。由于不同的基因电路可能需要以常微分方程的形式采用完全不同的数学表示来捕捉其瞬态动态行为，因此模型开发中一个反复出现的挑战是选择合适的模型。在这里，我们开发了一种自动生物模型选择系统（BMSS），它包括一个预先建立的模型库，这些模型具有从大量表达谱中得出的直观或非直观特征。模型选择基于赤池信息准则（AIC）。我们使用常规使用的诱导系统、组成型表达系统以及几种不同逻辑门系统（非门、与门和或门）的表征数据对该自动化平台进行了测试。BMSS对所有不同的表征数据集都达成了良好的一致性，并相应地成功选择了最合适的模型。为了实现基因电路设计模型的交换和可重复性，BMSS平台还生成符合合成生物学开放语言（SBOL）的基因电路图和系统生物学标记语言（SBML）输出文件。算法的所有方面都以模块化方式进行编程，以方便用户进行模型扩展或定制。综上所述，用Python实现的BMSS支持用户使用部件/电路表征数据以快速、高效和自动化的方式推导出最佳数学模型候选。
