Machine Learning Approach Reveals the Assembly of Activated Sludge Microbiome with Different Carbon Sources during Microcosm Startup

A metagenome-derived artificial intelligence modeling framework advances the predictive diagnosis and interpretation of petroleum-polluted groundwater

Groundwater (GW) quality monitoring is vital for sustainable water resource management. The present study introduced a metagenome-derived machine learning (ML) model aimed at enhancing the predictive understanding and diagnostic interpretation of GW pollution associated with petroleum. In this framework, taxonomic and metabolic profiles derived from GW metagenomes were combined for use as the input dataset. By employing strategies that optimized data integration, model selection, and parameter tuning, we achieved a significant increase in diagnostic accuracy for petroleum-polluted GW. Explanatory artificial intelligence techniques identified petroleum degradation pathways and Rhodocyclaceae as strong predictors of a pollution diagnosis. Metagenomic analysis corroborated the presence of gene operons encoding aminobenzoate and xylene biodegradation within the de novo assembled genome of Rhodocyclaceae. Our genome-centric metagenomic analysis thus clarified the ecological interactions associated with microbiomes in breaking down petroleum contaminants, validating the ML-based diagnostic results. This metagenome-derived ML framework not only enhances the predictive diagnosis of petroleum pollution but also offers interpretable insights into the interaction between microbiomes and petroleum. The proposed ML framework demonstrates great promise for use as a science-based strategy for the on-site monitoring and remediation of GW pollution.

The ASM2d model with two-step nitrification can better simulate biological nutrient removal systems enriched with complete ammonia oxidizing bacteria (comammox Nitrospira)

The discovery of comammox Nitrospira, a complete ammonia-oxidizing microorganism belonging to the genus Nitrospira, has brought new insights into the nitrification process in wastewater treatment plants (WWTPs). The applicability of Activated Sludge Model No. 2 d with one-step nitrification (ASM2d-OSN) or two-step nitrification (ASM2d-TSN) for the simulation of the biological nutrient removal (BNR) processes of a full-scale WWTP in the presence of comammox Nitrospira was studied. Microbial analysis and kinetic parameter measurements showed comammox Nitrospira was enriched in the BNR system operated under low dissolved oxygen (DO) and long sludge retention time (SRT). The relative abundance of Nitrospira under the conditions of stage I (DO = 0.5 mg/L, SRT = 60 d) was about twice of that under stage II conditions (DO = 4.0 mg/L, SRT = 26 d), and the copy number of the comammox amoA gene for stage I was 33 times higher than that for stage II. Compared to the ASM2d-OSN model, the ASM2d-TSN model simulated the performance of the WWTP under stage I conditions better, and the Theil inequality coefficient values of all the tested water quality parameters were lower than using ASM2d-OSN. These results indicate that an ASM2d model with two-step nitrification is a better choice for the simulation of WWTPs with the presence of comammox.

Machine learning reveals the complex ecological interplay of microbiome in a full-scale membrane bioreactor wastewater treatment plant

Membrane bioreactor (MBR) systems are one of the most widely used wastewater treatment processes for various municipal and industrial waste streams. The present study aimed to advance the understanding of ecologically important keystone taxa that play an important role in full-scale MBR systems. A machine-learning (ML) modeling framework based on microbiome data was developed to successfully predict, with an average accuracy of >91.6%, the operational characteristics of three representative full-scale wastewater systems: an MBR, a conventional activated sludge system, and a sequencing batch reactor. ML-based feature-importance analysis identified Ferruginibacter as a keystone organism in the MBR system. The phylogeny and known ecophysiology of members of Ferruginibacter supported their role in metabolizing complex organic polymers (e.g., extracellular polymeric substances) in MBR systems characterized by high concentrations of mixed liquor suspended solids and a high solid retention time. ML regression modeling also revealed temporal patterns of Ferruginibacter in response to water temperature. ML modeling was thus successfully employed in the present study to investigate complex/non-linear relationships between keystone taxa and environmental conditions that cannot be detected using conventional approaches. Overall, our microbiome-data-enabled ML modeling approach represents a methodological advance for identifying keystone taxa and their complex ecological interactions, which has implications for the sustainable and predictive management of MBR systems.

Effects of biochar addition on the fate of ciprofloxacin and its associated antibiotic tolerance in an activated sludge microbiome

This study investigated the effects of adding biochar (BC) on the fate of ciprofloxacin (CIP) and its related antibiotic tolerance (AT) in activated sludge. Three activated sludge reactors were established with different types of BC, derived from apple, pear, and mulberry tree, respectively, and one reactor with no BC. All reactors were exposed to an environmentally relevant level of CIP that acted as a definitive selective pressure significantly promoting AT to four representative antibiotics (CIP, ampicillin, tetracycline, and polymyxin B) by up to two orders of magnitude. While CIP removal was negligible in the reactor without BC, the BC-dosed reactors effectively removed CIP (70-95% removals) through primarily adsorption by BC and biodegradation/biosorption by biomass. The AT in the BC-added reactors was suppressed by 10-99%, compared to that without BC. The BC addition played a key role in sequestering CIP, thereby decreasing the selective pressure that enabled the proactive prevention of AT increase. 16S rRNA gene sequencing analysis showed that the BC addition alleviated the CIP-mediated toxicity to community diversity and organisms related to phosphorous removal. Machine learning modeling with random forest and support vector models using AS microbiome data collectively pinpointed Achromobacter selected by CIP and strongly associated with the AT increase in activated sludge. The identification of Achromobacter as an important AT bacteria revealed by the machine learning modeling with multiple models was also validated with a linear Pearson's correlation analysis. Overall, our study highlighted Achromobacter as a potential useful sentinel for monitoring AT occurring in the environment and suggested BC as a promising additive in wastewater treatment to improve micropollutant removal, mitigate potential AT propagation, and maintain community diversity against toxic antibiotic loadings.

Machine-learning insights into nitrate-reducing communities in a full-scale municipal wastewater treatment plant

This study carried out machine-learning (ML) modeling using activated sludge microbiome data to predict the operational characteristics of biological unit processes (i.e., anaerobic, anoxic, and aerobic) in a full-scale municipal wastewater treatment plant. An ML application pipeline with optimization strategies (e.g., model selection, input data preprocessing, and hyperparameter tuning) could significantly improve prediction performance. Comparative analysis of the ML prediction performance suggested that linear models (support vector machine and logistic regression) had a high prediction performance (93% accuracy), comparable to that of non-linear models such as random forest. Feature importance analysis using the linear ML models identified the microbial taxa that were specifically associated with anoxic processes, many of which (e.g., Ferruginibacter) were found to have ecologically important genomic and phenotypic characteristics (e.g., for nitrate reduction). Time-series microbial community dynamics demonstrated that the taxa identified using ML were frequently occurring and dominating in the anoxic process over time, thus representing the core nitrate-reducing community. Despite the general dominance of the core community over time, the analysis further revealed successional seasonal patterns of distinct sub-groups, indicating differences in the functional contribution of sub-groups by season to the overall nitrate-reducing potential of the system. Overall, the results of this study suggest that ML modeling holds great promise for the predictive identification and understanding of key microbial players governing the functioning and stability of biological wastewater systems.