Machine Learning for Detecting Virus Infection Hotspots Via Wastewater-Based Epidemiology: The Case of SARS-CoV-2 RNA

Area of Science:

• Environmental science
• Public health
• Epidemiology

Background:

• Wastewater-based epidemiology (WBE) effectively monitors public health issues like disease and drug use.
• WBE enables cost-effective surveillance of wastewater-detectable pathogens, complementing traditional methods.
• Current research highlights gaps in integrating hydraulic modeling and machine learning with WBE for viral outbreak detection.

Purpose of the Study:

• To develop and evaluate a coupled hydraulic and machine learning model for pathogen source tracking in sewer networks.
• To assess the methodology's effectiveness in identifying disease hotspots, specifically for SARS-CoV-2.
• To explore the potential of WBE for rapid back-tracing of human-excreted biomarkers.

Main Methods:

• Loosely coupled a physically-based hydraulic model with a machine learning model for pathogen transport and source identification.
• Applied the developed methodology to a hypothetical sewer network to detect disease hotspots.
• Evaluated the model's performance under various conditions, including different sewer system properties and sampling procedures.

Main Results:

• The machine learning model demonstrated promising capabilities in recognizing disease hotspots within the sewer network.
• Model accuracy is highly sensitive to the time-resolution of monitoring data, sewer system characteristics (e.g., flow velocity), and sampling procedures.
• The study identified limitations related to the need for high-frequency, contaminant-specific sensor systems, which are not yet available.

Conclusions:

• The proposed methodology advances WBE by enabling rapid source tracking of pathogens from wastewater samples.
• This approach offers potential for real-time public health surveillance and early detection of disease outbreaks.
• Future advancements require the development of high-frequency sensor technologies to fully realize the potential of this integrated WBE approach.