Wastewater and neural networks help scientists project current and future rates of COVID-19

June 8, 2021
Monitoring wastewater can potentially stop coronavirus outbreaks before they get out of hand. Here, sewage samples are collected from the dorms at Utah State University. (AP Photo/Rick Bowmer)

Monitoring wastewater can potentially stop coronavirus outbreaks before they get out of hand. Here, sewage samples are collected from the dorms at Utah State University. (AP Photo/Rick Bowmer)

  • Twitter
  • Linkedin
  • Facebook

Instead of letting sewage and runoff water go down the drain, researchers from Stanford University and major universities in Australia have used wastewater to predict the community prevalence of COVID-19 and provide cities with an early warning system if a potential second wave is detected after the peak of an outbreak.

The team's pioneering data-driven approach to wastewater-based monitoring of COVID-19 has been detailed in a new study, published May 23 in Science of The Total Environment. Their project will soon expand to East Asia and India.

The COVID-19 pandemic continues to ravage countries around the world. The researchers noted that, as of February 2021, the World Health Organization reported 109 million infections and 2.4 million deaths. When this article was written in June 2021, the WHO reported over 171 million confirmed cases and over 3.7 million deaths.

"Before the outbreak of COVID, I got major funding for tracking pathogens in wastewater," lead author and University of Wollongong researcher Dr. Guangming Jiang told The Academic Times. "When COVID appeared, people knew that I was working on that topic and told me I should track the disease."

Though Jiang was hesitant at first to switch his focus from common pathogens to a new disease, a chance opportunity made him think otherwise.

"There was a quite rare opportunity here as the first major outbreak in Australia was on a cruise ship, which was docked in Wollongong, close to our university's campus," he said. Jiang and his team managed to obtain samples from the ship, jumpstarting their coronavirus research.

Jiang is interested in epidemiology, or the science of a disease's spread and prevalence over a wide geographical population. The novelty of COVID-19 as a disease was appealing to him, as many variables related to it are still untested. 

"There are lots of unknowns. We don't know much about the shedding dynamics of the virus or the behavior of the virus in the sewer system," Jiang said. 

Previous studies have shown virus shedding in the urine or stools of people infected with COVID-19, including those with asymptomatic infections. Wastewater is therefore a fitting medium to monitor the disease. Researchers at the University of California, Merced are tracking global efforts to monitor SARS-CoV-2 wastewater in a cleverly titled "COVID19Poops" project.

Though wastewater monitoring is not a new tool when it comes to the spread of coronavirus, previous models have had some major shortcomings: In the paper, the authors noted that such approaches have been "greatly limited due to the complexity and uncertainties associated with the process." Data-heavy models can flesh out many of these unknowns. 

All data were collected from seven recent papers that provided SARS-CoV-2 RNA concentrations in wastewater along with information on confirmed and active cases of COVID-19. Jiang and his colleagues compared these numbers from real clinical tests to several models' predictions to check the models' performance and accuracy.

Out of the three models tested in the study, one clearly came out on top.

An artificial neural network model, designed to simulate the way neural networks work inside human brains, was able to reasonably estimate the prevalence of COVID-19 cases. Notably, this model could forecast the total number of upcoming cases in two time frames – both at the initial stage and in the next two to four days following an outbreak's peak.

A different neural network, an adaptive neuro fuzzy inference system, was less robust and more complicated when the data sets were incomplete. Though not completely inapplicable – the authors mention that it is accurate in some controlled situations – it could not estimate upcoming coronavirus cases after the peak, Jiang said.

Last and possibly least, a multiple linear regression model was less accurate and robust in predicting COVID-19. The authors make a point to say that it is not recommended.

Jiang mentioned that all together, the predicted numbers were slightly higher than the clinically confirmed number of cases. Asymptomatic patients and local laws on coronavirus are the culprits for this variance, the authors said. For example, Japan only tested patients with symptoms, essential workers and travelers from overseas, while Australia had an open public testing policy.

"The more data we have and the higher quality of the data, the better the model will be," Jiang added. Eventually, his team will move the AI model to Google Cloud to speed up their computations. A faster model and more available data will also allow the team to focus on other viral diseases, such as influenza, he said.

Jiang is well on the way to his goal. Before the paper was even published, he received additional funding from the Australian Academy of Science to support the expansion of the project. 

"It will allow me to develop this model in a broader scenario and establish a network for the Asia-Pacific region," he said.

Jiang also thinks that monitoring wastewater may help improve access to COVID-19 vaccinations. 

"We are still arguing about if we can get to herd immunity," he said. "A wastewater-based epidemiology approach can give us an idea about vaccination coverage with real-time data. We could even shift resources to other people who are more needy."

The study, "Data-driven estimation of COVID-19 community prevalence through wastewater-based epidemiology," published May 23 in Science of The Total Environment, was authored by Xuan Li, Shuxin Zhang, Jiahua Shi, Muttucumaru Sivakumar and Guangming Jiang, University of Wollongong; Jagadeeshkumar Kulandaivelu, Queensland Urban Utilities; Jochen Mueller, The University of Queensland; Stephen Luby, Stanford University; Warish Ahmed, CSIRO Land and Water; and Lachlan Coin, The University of Melbourne.