You are here:
Monitoring of Nitrification in Chloraminated Drinking Water Distribution Systems With Microbiome Bioindicators Using Supervised Machine Learning
Citation:
Gomez-Alvarez, V. AND R. Revetta. Monitoring of Nitrification in Chloraminated Drinking Water Distribution Systems With Microbiome Bioindicators Using Supervised Machine Learning. Frontiers in Microbiology. Frontiers, Lausanne, Switzerland, 11:2254-2267, (2020). https://doi.org/10.3389/fmicb.2020.571009
Impact/Purpose:
While disinfection strategies aim to mitigate the presence of microbes in drinking water distribution systems (DWDS), they do not completely eradicate their growth; therefore, a better understanding of the DWDS microbiome is needed to develop microbial control strategies. This research provides evidence of the feasibility of using the composition of bacterial biomarkers in BW to predict operational failures in the system (e.g. nitrification). This new information can be used to optimize current nitrification monitoring plans and ultimately helps to safeguard drinking water and human health.
Description:
Many water utilities in the US using chloramine as disinfectant treatment in their distribution systems have experienced nitrification episodes, which detrimentally impact the water quality. A chloraminated drinking water distribution system (DWDS) simulator was operated through four successive operational schemes, including two stable events (SS) and an episode of nitrification (SF), followed by a ‘chlorine burn’ (SR) by switching disinfectant from chloramine to free chlorine. We used 16S rRNA sequencing data to generate high-resolution taxonomic profiles of the bulk water (BW) microbiome for biomarker discovery. Principal coordinate analysis identified two major clusters (SS and SF; PERMANOVA, p < 0.0001) consistent with the effect of disturbance in the relative abundances of the core microbiome. Analysis of the BW microbiome (LEfSe analysis) revealed both an enrichment and depletion of various bacterial populations associated with nitrification. A supervised machine learning approach (naïve Bayes classifier) trained with biotic (biomass, OTUs, taxonomy) and abiotic (water quality) datasets was used to classify BW samples. Performance of each model was examined using the area under the curve (AUC) from the receiver-operating characteristic (ROC) and precision-recall (PR) curves. ROC-AUC and PR-AUC using biomass were 0.477 and 0.553, respectively; which is equivalent to a random classification. The AUCs gradually increased to 0.778 and 0.775 when genus-level membership (i.e. presence and absence) was used in the model, and increased significantly using structure (i.e. distribution) dataset (AUC = 1.000, p < 0.01). Community structure significantly improved the predictive ability of the model beyond that of membership only.
URLs/Downloads:
DOI: Monitoring of Nitrification in Chloraminated Drinking Water Distribution Systems With Microbiome Bioindicators Using Supervised Machine Learning
Free access through PubMed Central