eMB: Collaborative Research: ML/AI-assisted environmental scale microbial nonlinear metabolic models

我们的世界在许多方面由单细胞生物群落主导，例如细菌、古生菌和非丝状真菌。这些群落是所有地球化学循环中的关键参与者；它们存在于包括人类在内的所有多细胞生物体中，在这些生物中，它们除了可能产生有害影响外，还具有基本的有益作用。它们在工程系统中也普遍存在，同样，它们可能是有益的(例如废水处理)，也可能是有害的(例如饮用水分配)。在很大程度上，微生物群落与环境的相互作用是通过它们吸收的化学物质、它们用这些输入制造的产品以及它们排泄的副产品来代谢的，所以了解这些代谢能力对于了解微生物如何影响环境是至关重要的。基因测序的进展使确定微生物“机械”(酶)变得越来越容易；研究人员在预测这种“机械”如何结合成“装配线”(代谢途径)方面正变得越来越熟练。有了这些知识，下一步就是了解这些“装配线”如何适应他们所处的环境，形成一种决定整个微生物群落功能的大规模“分配系统”。在大范围内，这成为一个具有挑战性的计算问题，而现有的方法是不够的。该项目旨在通过将机器学习工具引入算法中的关键瓶颈来加速这些计算。该项目是坦普尔大学、蒙大拿州立大学和加州大学圣地亚哥分校的合作项目，提供了宝贵的教育、培训和外展机会。这项提议资助的活动将集中在开发基于机器学习和人工智能辅助优化的计算方法，这些方法足够有效，从而有可能将复杂的细胞尺度的微生物行为模型(代谢和基因表达模型，所谓的ME模式)嵌入环境尺度的微生物群落活动的PDE模型。为支持这一努力，将对几种特定生物(金黄色葡萄球菌、表皮葡萄球菌、枯草杆菌)的ME模型进行改造，以用于特定环境规模的模型(基本生物膜群落和建成环境下的空气下群落)。作为补充，将为这些系统建立环境尺度的基于连续介质力学(偏微分方程)的模型，并对其进行调整，以用于新的计算方法。人工智能还将应用于这一宏观尺度，试图在大范围内识别关键的新陈代谢过程。这一奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

Our world is dominated in many respects by communities of single celled organisms, e.g., bacteria, archaea and non-filamentous fungi. Such communities are key players in all geochemical cycles; they are present in all multicellular organisms, including humans, where in addition to possibly harmful effects, they also have essential beneficial roles. They are ubiquitous in engineered systems as well where, again, they can be beneficial (e.g., waste water treatment) or harmful (e.g., drinking water distribution). In large part, microbial communities interactions with their environments are metabolic through chemicals they take in, products they make with these inputs, and byproducts they excrete, so understanding these metabolic capabilities is essential for understanding how microbes effect their surroundings. Advances in genetic sequencing are making it easier and easier to determine microbial "machinery" (enzymes); researchers are becoming increasingly adept in predicting how this "machinery" combines into "assemblage lines" (metabolic pathways). Armed with this knowledge, the next step is to understand how these "assembly lines" fit into their environment into a sort of large scale "distribution system" that determines overall microbial community function. At large scale, this becomes a challenging computational problem, and current methods are not adequate. This project aims to accelerate these computations by introducing machine learning tools into key bottlenecks in the algorithms. The project is a collaboration between Temple University, Montana State University, and the University of California, San Diego and offers valuable educational, training, and outreach opportunities. Activity funded by this proposal would center on development of computational methods, based on machine learning and artificial intelligence assisted optimization, that are sufficiently efficient so as to make it possible to embed complex cell-scaled models of microbial behavior (metabolic and gene expression models, so-called ME modes) into environmental scale PDE-based models of microbial community activity. In support of this effort, ME models of several specific organisms (S. aureus, S. epidermidis, B. subtilis) will be adapted for use in specific environmental-scale models (basic biofilm communities and built-environment subaerial communities). In complement, environmental-scale continuum-mechanics-based (partial differential equation) models for these systems will be constructed and adapted for use with the new computational methods. AI will also be applied at this macroscale to attempt to identify key metabolic processes at the large scale.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.