Predicting COVID-19 case counts using SARS-CoV-2 genetic diversity from wastewater

Sana Naderi, Steven Gregory Sutclliffe, Gavin M. Douglas, Sukriye Celikkol Aydin, Inès Levade, Judith Fafard, Lila Naouelle Salhi, Fernando Sanchez-Quete, Sarah J. Reiling, Ju‐Ling Liu, Marc-Denis Rioux, David Dreifuss, Ivan Topolsky, Niko Beerenwinkel, Selena M. Sagan, Stephanie K. Loeb, Peter A. Vanrolleghem, Sarah Dorner, Dominic Frigon, Jiannis Ragoussis, B. Jesse ShapiromedRxiv 2025 · DOI

TL;DR

Researchers have found that tracking the genetic diversity of SARS-CoV-2 in wastewater—specifically the number of "segregating sites" (unique mutations)—can predict clinical case surges up to a week in advance. This sequencing-based approach proved more robust than traditional PCR-based viral concentration measurements, which can be "noisy" and failed to provide significant lead time in regions with less frequent sampling.

What Problem Does This Paper Address?

Wastewater-based epidemiology (WBE) became a cornerstone of public health during the COVID-19 pandemic. Traditionally, officials rely on RT-qPCR to measure the concentration of viral RNA in sewage. While cost-effective, this method has several "real-world" flaws:

  1. Environmental Noise: Heavy rain, snowmelt, or chemical degradation in sewers can dilute or destroy viral signals.
  2. Biological Variability: People shed virus at different rates and for different durations (sometimes weeks after recovery), blurring the line between new infections and old ones.
  3. Primer Mismatches: As the virus evolves, mutations can occur exactly where PCR primers are supposed to bind, leading to "amplicon dropouts" and false negatives.

The authors sought a more robust "leading indicator"—a metric that doesn't just tell us how much virus is there, but whether the viral population is actively expanding.

Methods

The team pivoted from measuring quantity to measuring diversity. They used Whole-Genome Sequencing (WGS) to analyze the "pool" of viral RNA in wastewater.

The key metric they used is the number of segregating sites: the count of genomic positions where multiple versions (alleles) of the virus exist in a single sample. In population genetics, an explosion of rare, new mutations is a classic signature of a population undergoing rapid expansion.

To prove this diversity predicts future cases, they used:

  • Lagged Spearman Correlations: To see if diversity today correlates with cases $X$ days from now.
  • Granger Causality Tests: A statistical tool used to determine if one time series is actually useful in forecasting another.
  • Computational Deconvolution: Using the tool Freyja to identify which specific lineages (like Omicron BA.5 or XBB.1.5) were present.

Datasets

The study validated its approach across two distinct geographic and temporal settings:

  1. Quebec, Canada: 10 sampling locations (ranging from urban Montreal to rural Sept-Îles) tracked for about a year (March 2022 – March 2023), totaling 2,574 sequenced samples.
  2. Switzerland: An independent dataset from 6 locations (including Zurich and Geneva) sampled almost daily from February to December 2022.

Key Insights and Ideas

  • Diversity as a Proxy for Growth: The "aha moment" here is that as an outbreak starts, many different people are infected with slightly different versions of the virus. This creates a spike in genetic diversity before the total viral load reaches its peak.
  • Variant-Agnostic Surveillance: Unlike methods that look for specific "Variants of Concern" (like Delta or Omicron), counting segregating sites doesn't require a reference database. It can detect a population expansion even if the variant causing it is brand new and uncharacterized.
  • Robustness to "Amplicon Dropout": Because WGS covers the entire 30,000-base genome, losing a few segments due to mutations (which would break a PCR test) doesn't significantly change the overall diversity count.
  • Clearing the "Shedding" Noise: The authors note that once a virus stops replicating in a patient, it stops producing new mutations. Therefore, diversity metrics may be less affected by "old" viral RNA being shed by recovered patients compared to simple concentration metrics.

Key Results

  • Lead Time: In Quebec, the number of segregating sites predicted case counts 5 to 8 days in advance in 6 out of 10 locations.
  • PCR vs. Sequencing: In the Quebec dataset, standard PCR-based viral concentrations were not significantly predictive of future cases at any location.
  • Consistency: In the Swiss dataset (which had much denser, daily sampling), both methods worked well, predicting cases in 5 out of 6 locations. This suggests that while PCR works under "ideal" daily sampling conditions, sequencing-based diversity is far more reliable when sampling is less frequent (e.g., twice a week).
  • Statistical Strength: A linear mixed model showed that diversity metrics were consistently more strongly associated with case counts than viral load, with PCR estimates being roughly 4.86% less accurate in their correlations.

Limitations and Future Work

  • Cost and Speed: WGS is currently more expensive and slower than PCR. For this to be a real-time public health tool, sequencing turnaround times need to drop.
  • Sampling Frequency: The authors found that the signal is strongest when samples are taken at least twice a week. In very sparse datasets (like Quebec City in this study), the predictive power vanished.
  • Generalization: While this worked for SARS-CoV-2, it remains to be seen if the same "diversity-as-growth" signal can be used for other pathogens like Influenza, RSV, or even antibiotic-resistant bacteria.

The study concludes that while PCR is a great "quick look" tool, genetic diversity provides a more robust, "weather-proof" forecast for public health planning.

Summary updated

Abstract

Abstract Wastewater monitoring is a promising, cost-effective method for pathogen surveillance, particularly when individual patient testing is limited. A typical approach is to estimate viral concentrations using PCR-based quantification of viral nucleic acids. This approach has been successfully used during the ongoing COVID-19 pandemic to track, and even predict, waves of infection in a community. Although simple in principle, PCR-based quantification can be noisy and susceptible to false-negatives when mutations interrupt primer binding sites. Here we compare PCR-based quantification to whole-genome sequencing of SARS-CoV-2 to predict the number of positive COVID-19 cases in a local region. To do so, we calculate a simple population genetic metric as a proxy for population expansion: the number of polymorphic nucleotide sites in the viral genome, referred to as the number of segregating sites. We show that the number of segregating sites is predictive of COVID-19 case counts from up to 5-8 days prior, in 6 out of 10 wastewater sampling locations across the province of Quebec, Canada tracked over a mean period of ∼280 days. By contrast, PCR-based viral concentrations were not significantly predictive at any sampling location in Quebec. In an independent Swiss wastewater dataset sampled over a similar duration, both segregating sites and PCR-based quantification were predictive of future case counts in 5 out of 6 locations. Together, these results suggest that PCR-based approaches may be more sensitive to region-specific differences in sampling frequency and quality, whereas sequencing-based approaches may be more robust leading indicators of infection counts. Importance Wastewater monitoring of pathogens is an increasingly important part of the epidemiological surveillance toolkit. It typically involves quantifying pathogens in wastewater by PCR, an approach that is relatively simple but can be noisy and prone to false-negatives. Here we explore an alternative approach, based on quantifying viral genetic diversity by whole-genome sequencing SARS-CoV-2 from wastewater. Using wastewater samples from Quebec and Switzerland, we show that a simple metric of genetic diversity (which tracks with population expansion and does not require any pre-defined database of viral variants) is predictive of COVID-19 case counts about a week in advance, in a majority of sampled locations. While sequencing may be more costly and labor-intensive than PCR, it provides a robust and informative tool for wastewater-based epidemiology.
BibTeX
@misc{naderi2025predicting,
  author = {Sana Naderi and Steven Gregory Sutclliffe and Gavin M. Douglas and Sukriye Celikkol Aydin and Inès Levade and Judith Fafard and Lila Naouelle Salhi and Fernando Sanchez-Quete and Sarah J. Reiling and Ju‐Ling Liu and Marc-Denis Rioux and David Dreifuss and Ivan Topolsky and Niko Beerenwinkel and Selena M. Sagan and Stephanie K. Loeb and Peter A. Vanrolleghem and Sarah Dorner and Dominic Frigon and Jiannis Ragoussis and B. Jesse Shapiro},
  title = {{Predicting COVID-19 case counts using SARS-CoV-2 genetic diversity from wastewater}},
  year = {2025},
  doi = {10.1101/2025.11.07.25339787},
  howpublished = {medRxiv},
}

Loading related content...

Figure