Roguet Adélaïde, Esen Özcan C, Eren A Murat, Newton Ryan J, McLellan Sandra L
School of Freshwater Sciences, University of Wisconsin-Milwaukee, Milwaukee, Wisconsin, USA.
Department of Medicine, University of Chicago, Chicago, Illinois, USA.
mSystems. 2020 Mar 17;5(2):e00869-19. doi: 10.1128/mSystems.00869-19.
Sewage overflows, agricultural runoff, and stormwater discharges introduce fecal pollution into surface waters. Distinguishing these sources is critical for evaluating water quality and formulating remediation strategies. With the falling costs of sequencing, microbial community-based water quality assessment tools are under development. However, their application is limited by the need to build reference libraries, which requires extensive sampling of sources and bioinformatic expertise. Here, we introduce FORest Enteric Source IdentifiCation (FORENSIC; https://forensic.sfs.uwm.edu/), an online, library-independent source tracking platform based on random forest classification and 16S rRNA gene amplicon sequences to identify in environmental samples common fecal contamination sources, including humans, domestic pets, and agricultural animals. FORENSIC relies on a broad reference signature database of and , two predominant bacterial groups that have coevolved with their hosts. As a result, these groups demonstrate cohesive and reliable assemblage patterns within mammalian species or among species sharing the same diet/physiology. We created a scalable and extensible platform that we tested for global applicability using samples collected in distant geographic locations. This Web application offers a fast and intuitive approach for fecal source identification, particularly in sewage-contaminated waters. FORENSIC is an online platform to identify sources of fecal pollution without the need to create reference libraries. FORENSIC is based on the ability of random forest classification to extract cohesive source microbial signatures to create classifiers despite individual variability and to detect the signatures in environmental samples. We primarily focused on defining sewage signals, which are associated with a high human health risk in polluted waters. To test for fecal contamination sources, the platform only requires paired-end reads targeting the V4 or V6 regions of the 16S rRNA gene. We demonstrated that we could use V4V5 reads trimmed to the V4 positions to generate the reference signature. The systematic workflow we describe to create and validate the signatures could be applied to many disciplines. With the increasing gap between advancing technology and practical applications, this platform makes sequence-based water quality assessments accessible to the public health and water resource communities.
污水溢流、农业径流和雨水排放将粪便污染引入地表水。区分这些污染源对于评估水质和制定修复策略至关重要。随着测序成本的下降，基于微生物群落的水质评估工具正在开发中。然而，它们的应用受到构建参考文库的需求的限制，这需要对污染源进行广泛采样并具备生物信息学专业知识。在此，我们介绍了粪便源识别在线平台FORest Enteric Source IdentifiCation（FORENSIC；https://forensic.sfs.uwm.edu/），该平台基于随机森林分类和16S rRNA基因扩增子序列，无需参考文库即可识别环境样本中常见的粪便污染源，包括人类、家养宠物和农业动物。FORENSIC依赖于两个主要细菌类群的广泛参考特征数据库，这两个类群与其宿主共同进化。因此，这些类群在哺乳动物物种内部或具有相同饮食/生理特征的物种之间表现出凝聚且可靠的组合模式。我们创建了一个可扩展的平台，并使用在遥远地理位置收集的样本对其全球适用性进行了测试。这个网络应用程序为粪便源识别提供了一种快速且直观的方法，特别是在受污水污染的水域中。FORENSIC是一个无需创建参考文库即可识别粪便污染源的在线平台。FORENSIC基于随机森林分类提取凝聚性源微生物特征以创建分类器的能力，尽管存在个体差异，仍能在环境样本中检测到这些特征。我们主要专注于定义污水信号，其在受污染水体中与高人类健康风险相关。为了测试粪便污染源，该平台仅需要靶向16S rRNA基因V4或V6区域的双端读数。我们证明可以使用修剪到V4位置的V4V5读数来生成参考特征。我们描述的创建和验证特征的系统工作流程可应用于许多学科。随着先进技术与实际应用之间差距的不断扩大，该平台使公共卫生和水资源社区能够进行基于序列的水质评估。
