TL;DR
This study benchmarks four major next-generation sequencing (NGS) platforms—Illumina, Element Aviti, and two Oxford Nanopore (ONT) flow cells—to determine their effectiveness in tracking SARS-CoV-2 variants in wastewater. The researchers found that while newer and portable technologies have higher error rates, they all reliably track variant trends, with ONT sequencing providing actionable results in as little as five hours.
What Problem Does This Paper Address?
Wastewater-based epidemiology (WBE) became a cornerstone of public health during the COVID-19 pandemic, offering a non-invasive, cost-effective way to monitor virus prevalence at the population level. However, the "gold standard" for sequencing these samples—Illumina—is often centralized, expensive, and slow.
Public health officials need to know if alternative platforms, like the highly accurate Element Aviti or the portable and rapid Oxford Nanopore Technologies (ONT), can deliver the same quality of data. The challenge is that wastewater is a "noisy" matrix; it contains degraded RNA from multiple circulating variants and various inhibitory substances. This paper asks: how much does your choice of sequencer actually affect your ability to estimate which variants are in the water and at what concentration?
Methods
The authors conducted a systematic head-to-head comparison using two experimental setups:
- Surveillance Samples: 42 real-world samples collected over one week from six wastewater treatment plants (WWTPs) across Switzerland.
- Spike-in Experiments: A controlled dilution series where a known clinical isolate (Omicron BA.1) was spiked into a "wild-type" wastewater background to test the sensitivity and accuracy of each platform.
All samples were processed using the ARTIC v4.1 tiling-amplicon protocol, which creates 400bp DNA chunks covering the viral genome. For bioinformatic analysis, they used V-pipe and a tool called LolliPop to "deconvolve" the data—essentially a mathematical way to look at a soup of mutations and calculate the relative abundance of specific lineages (like BA.5 or BQ.1.1).
Datasets
- Swiss Surveillance Set: 42 daily composite samples from WWTPs in Altenrhein, Chur, Laupen, Lugano, Geneva, and Zurich (January 2023).
- BA.1 Spike-in Set: 12 dilution steps (in triplicate) of clinical Omicron BA.1 into SARS-CoV-2-free wastewater.
- Reference Data: Clinical isolate sequences and the NC_045512.2 reference genome.
- Public Availability: The raw sequencing data is hosted on the European Nucleotide Archive (PRJEB44932).
Key Insights and Ideas
- The New Contender: The Element Aviti platform performed exceptionally well, showing the highest concordance with Illumina. In fact, its mean sequencing error rate (0.0032) was slightly lower than Illumina’s (0.0038), making it a formidable alternative for high-precision surveillance.
- Noise vs. Signal: ONT platforms (MinION and Flongle) had significantly higher error rates (0.1% to 0.2%). This leads to "noise"—the detection of spurious variants at low levels. However, because the deconvolution algorithms look at many signature mutations simultaneously, they are remarkably tolerant of these errors when estimating the overall trend.
- The 5-Hour Window: One of the most practical findings is that ONT sequencing runtime can be slashed. The authors found that data collected in the first five hours of a run provided variant estimates nearly identical to those from a full 72-hour run. This suggests a path toward near-real-time genomic surveillance.
- Coverage Consistency: All technologies showed similar "drops" in coverage at specific points in the genome (amplicons 14 and 73), proving that these biases stem from the PCR amplification step rather than the sequencing hardware itself.
Key Results
- Accuracy: Aviti and Illumina were nearly identical ($R^2 = 0.935$). MinION also showed strong correlation with Illumina ($R^2 = 0.920$), while the smaller Flongle flow cell was less precise ($R^2 = 0.674$).
- Error Rates: Nanopore technologies showed a much larger spread in error rates. Flongle, in particular, struggled with "undefined mutations" due to its higher noise floor.
- Variant Underestimation: Initially, Nanopore seemed to underestimate the BA.1 variant. The researchers proved this wasn't a systematic bias but rather a result of the "noise" taking up a small percentage of the total abundance. When they re-normalized the data to ignore spurious variants, the underestimation disappeared.
Limitations and Future Work
- Short Timeframe: The surveillance study only covered one week. The authors suggest that longer time series would allow for "kernel smoothing," which would likely further reduce the divergence between Nanopore and Illumina data.
- Read Length: The study used 400bp amplicons, which doesn't take advantage of Nanopore’s ability to sequence very long fragments. Long-read sequencing could theoretically allow researchers to see multiple mutations on a single strand of RNA, making variant identification even more certain.
- Wastewater Complexity: Extracting long, high-quality RNA from wastewater remains a major technical hurdle that limits the shift toward long-read methods.
- Computational Load: While sequencing is getting faster, the authors note that bioinformatic pipelines like V-pipe still require significant high-performance computing resources, which remains a bottleneck for local, real-time implementation.
Summary updated