Big Data on Small Organisms: Petascale Simulations of Data-Driven, Whole-Cell Microbial Models

该项目旨在开发下一代基因组规模的、数据驱动的微生物生物模型。由于高通量数据的可获得性、细胞组织及其对工业和人类健康的重要性，该项目将首先专注于研究最多的微生物--革兰氏阴性杆菌--大肠杆菌。该项目将利用过去15年来并行高通量分子图谱方面的进展所产生的多组学数据集，数据驱动的、集成的、多尺度模型的出现，其预测能力大幅提高，以及机器学习方面的新技术，特别是与深度学习有关的新技术。对微生物健康和细胞状态的准确预测可能会对我们检验与健康、社会或经济利益直接相关的假设的方式产生深远的影响。该奖项将通过本科课程、IGEM团队和其他倡议，支持多名本科生和研究生在生物系统的计算建模和高性能模拟方面的培训。该项目将支持从最广泛使用的微生物的最大标准化组学百科全书中产生知识，这将有利于分子和细胞生物学中下一代数据驱动预测方法的开发和培训。它将为评估最先进的多尺度模型提供计算资源，该模型具有从集体组学数据预测表型特征和环境条件的能力。这将是第一个以特定微生物(大肠杆菌)为目标的系统级模拟器，能够以从单个基因浓度到种群动态的分辨率来模拟细胞种群。为了实现这一点，必须在这种背景下采用和应用进程迁移、负载平衡和强伸缩技术，这些都是全细胞建模领域的新特征。所提出的高性能预测模拟将与假设生成和检验密切相关。这些模拟将解决与它们在复杂环境中的微生物培养的表型和表达谱有关的问题。在系统生物学的背景下，这些技术的集成具有变革的潜力，但前提是有能够处理这些任务的必要的计算基础设施。Blue Waters超级计算机凭借其独特的架构、大规模的模拟能力和专业的支持人员，为实现这一雄心勃勃的目标提供了理想的平台。

This project aims to develop the next-generation of genome-scale, data-driven models for microbial organisms. The project will first focus on the most-studied microbe, the gram-negative bacterium Escherichia coli, due to the availability of high-throughput data, cellular organization, its significance to industry and human health. The project will take advantage of the multi-omics datasets that resulted from advances in parallel high-throughput molecular profiling over the past fifteen years, the emergence of data-driven, integrative, multi-scale models with substantial improvement of their predictive power and new techniques in machine learning, especially those related to deep learning. Accurate prediction of microbial fitness and cellular state can have profound implications to the way we test hypotheses that are directly related to health, social or economic benefits. This award will support the training of multiple undergraduate and graduate students in computational modeling and high-performance simulations of biological systems through undergraduate courses, IGEM teams and other initiatives.This project will support the generation of knowledge from the largest normalized omics compendia for the most widely used microbe that will be a boon for the development and training of the next generation of data-driven predictive methods in molecular and cellular biology. It will provide the computational resources to evaluate a state-of-the-art multi-scale model with the capacity to predict phenotypic characteristics and environmental conditions from collective omics data. This will be the first systems-level simulator that targets a specific microbe (E. coli) and will be able to simulate populations of cells with a resolution ranging from individual gene concentrations to population dynamics. To achieve that, process migration, load-balancing and strong scaling techniques have to be adopted and applied in this context, which are all novel features for the area of whole-cell modeling. The proposed HPC simulations will be intimately related to hypothesis generation and testing. The simulations will address questions related to what their phenotype and expression profiles of microbial cultures are in complex environments. In the context of systems biology, integration of these techniques has the potential of being transformative, but only if the necessary computational infrastructure able to handle these tasks is available. The Blue Waters Supercomputer with its unique architecture, large-scale simulation capabilities and professional support staff provides the ideal platform to achieve this ambitious goal.