NSF2026: EAGER: Identifying microbes’ population-level environmental responses using Bayesian modeling

在地球科学理事会和综合活动办公室的国家科学基金会2026基金项目的支持下，杜克大学的达纳·亨特教授、马克·博苏克教授和詹姆斯·克拉克教授进行了一项研究，为塑造海洋微生物生产力和功能的因素以及在飓风等极端事件中这种变化提供了新的见解。这项研究的驱动力来自这样一个事实，即海洋微生物提供必要的生态系统服务，包括初级生产（光合作用）和有机物周转，这维持了所有海洋生物。也就是说，微生物群在多大程度上受到环境因素的影响仍不清楚，比如温度和初级生产力，这些因素可以被季节、干扰、全球变化和其他因素改变。这项研究结合了在北卡罗来纳州波弗特岛沿海地区的长期观测，并利用这些数据，利用佛罗伦萨飓风（2018年）和多里安飓风（2019年）前后进行的高频测量，捕捉微生物群及其环境的年度变化。研究飓风对海洋生物群落的影响是很重要的，因为飓风是一种多因素干扰，它将外来的淡水和陆地微生物引入一个稳定的系统，同时改变沿海海洋的盐度、营养物质和有机物。这项工作结合了现场观测和建模的信息，开发了新的方法，可以区分现场样品中经常同时出现的因素，例如大西洋沿岸夏季气温升高和初级产量增加。通过整合微生物组研究的多个方面，本工作加深了目前对沿海海洋微生物组系统及其功能的认识。它还提出了新的可验证的假设，以指导未来的研究。这项工作的更广泛影响包括对本科生、研究生和博士后的高级培训，以及将研究成果转化为面向K-12学生和公众的产品。其他影响包括为项目过程中开发的软件编写详细的用户手册和培训材料，以便将研究成果用于未来的微生物组研究和本科教育。本研究利用已建立的长达十年的微生物时间序列，Piver岛海岸观测站（PICO, Beaufort Inlet, NC USA）来改进微生物种群及其与不断变化的环境的关系的建模。PICO数据集拥有10年每周（或更频繁）的微生物群落SSU rRNA基因序列数据集，再加上样品、原位和环境参数，是沿海海洋微生物组最完整、最长期的数据集之一。这项工作将贝叶斯模型应用于PICO时间序列，以提高对微生物组对海洋条件反应的理解和预测。贝叶斯模型非常适合微生物系统，因为它们有能力处理稀疏数据集，捕捉对环境变化的非线性响应，并包括干扰的影响。这项研究整合了微生物组应用和贝叶斯模型gjamTime。这种组合有可能通过利用多元时间序列方法的进步来改变微生物生态，这些方法可以适应个体分类群与其环境之间随时间的依赖性。该项目的一个目标是使用Beaufort Inlet现场自然干扰（即飓风）的时间序列数据来测试模型预测，并探索各种关键环境参数，如温度（+3°C）和初级生产作为关键环境参数。类似的工作将在更广泛的海洋中进行。该研究的影响超出了目标沿海数据集，因为如果成功，该方法可以应用于其他不同的研究系统，如土壤和人类微生物组。它还可以用于解决有关环境过滤、干扰和随机性的问题，这些问题对于理解控制微生物对环境变化反应的因素和过程至关重要。该项目响应了NSF2026 Idea Machine获奖作品“变化中的地球中的全球微生物组”和“想象一个拥有清洁海洋的生活”。该奖项反映了NSF的法定使命，并通过基金会的智力价值和更广泛的影响审查标准进行了评估，认为值得支持。

With support from the Directorate for Geosciences and the NSF 2026 Fund Program in the Office of Integrated Activities, Professors Dana Hunt, Mark Borsuk, and James Clark at Duke University conduct research that provides new insights into the factors that shape microbial productivity and function in the oceans as well as how this change during extreme events such as hurricanes. The driver of this research comes from the fact that marine microbes provide essential ecosystem services, including primary production (photosynthesis) and organic matter turnover, that sustains all marine organisms. That said, it still remains unclear as to what extent microbiomes are shaped by environmental factors, such as temperature and primary productivity, that can be altered by season, disturbances, global change, and other factors. This research combines long-term observations at a coastal site at Beaufort Island, North Carolina and uses these data to capture annual changes in microbiomes and their environments using high frequency measurements that were taken before and after hurricanes Florence (2018) and Dorian (2019). Examining the impact of hurricanes on marine biomes is important because hurricanes are multi-factor disturbances that introduce both foreign freshwater and terrestrial microbes into a stable system while altering salinity, nutrients, and organic matter in the coastal ocean. This work combines information from field observations and modeling to develop new approaches that will allow the differentiation of factors that often co-occur in field samples, such as warmer temperatures and higher primary production that occur during the summer months in the coastal Atlantic Ocean. By integrating multiple aspects of microbiome research, this work deepens current understanding of the coastal ocean microbiome system and its functionality. It also develops new testable hypothesis to guide future research. Broader impacts of the work include advanced training for undergraduate, graduate, and postdoctoral students, as well as translating research results into products for K-12 students and the public. Additional impacts include the production of detailed user manuals and training materials for software developed in the course of the project to facilitate the use of research results for future microbiome research and undergraduate education.This research leverages an established decade-long microbial time-series, the Piver’s Island Coastal Observatory (PICO, Beaufort Inlet, NC USA) to improve the modeling of microbial populations and their relationship to changing environments. With 10 years of weekly (or more frequent) microbial community SSU rRNA gene sequence datasets, coupled with the suite of sample, in-situ, and environmental parameters, the PICO dataset is one of the most complete, long-term datasets for coastal ocean microbiomes. The work carried out uses the application of Bayesian modeling to the PICO time series to improve understanding and predictions of microbiome responses to ocean conditions. Bayesian models are well suited to microbial systems because they have the ability to handle sparse datasets, capture non-linear responses to environmental changes, and include impacts of disturbances. This research integrates microbiome applications and the Bayesian model gjamTime. This combination has the potential to transform microbial ecology by leveraging advances in multivariate time-series methods that accommodate the dependence among individual taxa and their environment over time. One goal of the project is to test model predictions using time-series data from natural disturbances (i.e., hurricanes) at the Beaufort Inlet site and explore various key environmental parameters such as temperature (+3 °C) and primary production as key environmental parameters. Similar work will be done more broadly for the ocean. Impacts of the research extend beyond the targeted coastal dataset as, if successful, the approach can be applied to other diverse study systems such as soil and human microbiomes. It can also be used to address questions about environmental filtering, disturbance and stochasticity, each of which is critical to understanding the factors and processes that govern microbial responses to environmental change.This project responds to the NSF2026 Idea Machine winning entries of "Global Microbiome in a Changing Planet" and "Imagine a Life with Clean Oceans"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.