As clinical testing for SARS-CoV-2 wanes, wastewater surveillance emerges as a crucial method for monitoring the emergence of new variants of concern (VoCs). Wastewater samples, representing a pooled sample from a community, contain a diverse range of circulating SARS-CoV-2 variants, including potentially novel genotypes. However, challenges exist in detecting VoCs in wastewater due to factors such as uneven genome coverage, RNA degradation, and the need for sufficient sequencing depth to identify variants. Current methods often require high variant representation in the sample, hindering early detection and often ignore insertion and deletion (indel) information. Existing approaches also suffer from biases introduced by the underlying database of known SARS-CoV-2 genomes. This study addresses these limitations by introducing QualD, a computational pipeline designed to analyze SARS-CoV-2 wastewater sequencing data and infer the presence of VoCs using both single nucleotide variants (SNVs) and indels. The researchers aim to demonstrate QualD's improved performance in terms of both accuracy and timeliness compared to existing tools, specifically Freyja, using both real and simulated data. The early and accurate detection of VoCs in wastewater is essential for timely public health responses and effective pandemic management.

The existing literature highlights the value of wastewater monitoring for SARS-CoV-2 surveillance, offering early detection and broad population coverage. However, current approaches for VoC detection in wastewater face challenges. Many methods require high sequencing depth and breadth of coverage, making early detection difficult. Most methods disregard indel information and rely solely on SNVs, limiting sensitivity. Furthermore, all methods relying on databases of known genomes are susceptible to biases in both false positive and false negative calls. Previous studies have explored the use of wastewater sequencing for SARS-CoV-2 surveillance, with varying success in early detection and accuracy depending on methodologies. While tools like Freyja and CAJAC have made advancements, limitations remain in terms of sensitivity, precision and the integration of indel information. The need for a more robust and sensitive tool for early and accurate VoC detection in wastewater is clear.

The study employed a multi-faceted methodology. First, 2637 wastewater samples were collected weekly from 39 wastewater treatment plants (WWTPs) in Houston, Texas, between February 23, 2021 and May 31, 2022. These samples underwent RNA extraction and sequencing using various protocols, detailed in Supplementary Table 2. A mutation database was constructed using GISAID data, identifying quasi-unique mutations for each lineage. QualD, the developed pipeline, leverages both SNVs and indel data to infer VoC presence. The process involves identifying quasi-unique mutations (mutations present in >50% of genomes within a lineage, but not in other lineages), aggregating these mutations at a chosen hierarchical level, and then using the presence or absence of these mutations in the wastewater samples to predict VoC presence. QualD's performance was compared with Freyja, a state-of-the-art tool, using both real Houston wastewater data and simulated data. Three simulation protocols were used: (a) random SNV dropout, (b) coverage template-based SNV resampling, and (c) coverage template-based read resampling. These protocols simulated different levels of sequencing data degradation and coverage variation. The precision, recall, and F1 score were used to evaluate the performance of both QualD and Freyja across the different simulation scenarios and real data. The researchers also analyzed the multiple sequence alignment (MSA) of SARS-CoV-2 Omicron variant genomes to support their findings regarding early detection of the Omicron variant.

QualD demonstrated superior performance compared to Freyja in both real and simulated data. In the Houston wastewater data, QualD detected the Delta VoC two weeks earlier than the first sequenced clinical sample in Texas, and detected Omicron two weeks earlier than Freyja. Simulated experiments showed that QualD maintained high sensitivity even with substantial SNV dropout (as low as 10%), while Freyja required at least 50% of SNVs to be retained for reliable detection. QualD consistently exhibited higher precision than Freyja across all simulated scenarios for all VoCs (Alpha, Delta, Gamma, and Omicron), as shown in Table 1. The heatmap visualization of Omicron variant calls in the Houston wastewater data (Figure 1d) showed that a significant portion of samples contained the 9 bp deletion (N:DEL31-33), a stable mutation characteristic of the Omicron variant. This highlights QualD's ability to leverage indel information for improved detection. The robustness of QualD's detection was further supported by the stability of the coverage for the N:DEL31-33 deletion in the real data, with over 61% of samples having at least 10 reads covering the flanking bases. The inclusion of indel information and aggregated variant calls proved crucial for QualD's high precision.

The findings demonstrate the significant improvement that QualD offers for early and accurate detection of SARS-CoV-2 VoCs in wastewater. QualD's superior sensitivity and precision compared to Freyja, particularly in the face of degraded or incomplete sequencing data, underscores its value for wastewater-based surveillance. The ability to leverage indel information significantly enhances its ability to detect variants at an early stage, before they become widely prevalent in clinical samples. The results demonstrate the potential of QualD for robust public health surveillance, offering timely insights into emerging VoCs and informing preventative measures. The integration of QualD alongside other tools like Freyja and COJAC could provide a comprehensive surveillance system, combining high sensitivity, abundance estimation, and high specificity confirmation.

QualD presents a significant advancement in wastewater-based surveillance of SARS-CoV-2 VoCs. Its ability to provide early, accurate, and robust detection, even with limited data, makes it a valuable tool for public health monitoring. Future directions include expanding QualD's framework to other pathogens and developing more sophisticated simulation datasets to better reflect the complexities of wastewater sequencing data.

While QualD demonstrates superior performance, limitations exist. The accuracy of the tool depends on the quality of the wastewater sequencing data and the completeness of the mutation database. The simulation protocols, while comprehensive, may not perfectly capture all the complexities of real-world wastewater samples. Further refinement of the simulation methods and comprehensive testing with diverse wastewater samples from different geographical locations would enhance the generalizability of the findings.