Evaluation of Sample Pooling for Gene Sequencing of SARS-CoV-2: A Simulation Study

doi:10.21203/rs.3.rs-4022391/v1

Download PDF

Short Report

Evaluation of Sample Pooling for Gene Sequencing of SARS-CoV-2: A Simulation Study

https://doi.org/10.21203/rs.3.rs-4022391/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

COVID-19 continues to pose a significant public health threat, requiring ongoing epidemiological and genomic surveillance. Next-generation sequencing (NGS) is commonly utilized for monitoring viral evolution at high-cost. Real-time reverse transcription polymerase chain reaction(RT-qPCR) pooling test have been have been demonstrated to increase molecular testing throughput effectively during the pandemic, while pooling sequencing shows promise as a cost-effective tool for monitoring virus variants. A simulation study was conducted to evaluate the efficacy of sample pooling for viral gene sequencing of SARS-CoV-2. In this study, 72 original raw data of gene sequencing results with different lineages were collected and combined to create simulated samples based on five pooling strategies. A bioinformatics tool based on Freyja was utilized to analyze the variant composition of these 70 simulated pooled samples. The genetic composition of the pooled samples mostly recovered the variant compositions of the original samples, with discrepancies between the top X results (where X is the number of original samples included in the mixture)and the complete results (P < 0.05).Variations were observed in the calculation method for the top X results (P < 0.05) but not in the complete calculation approach (P > 0.05) across the five pooling strategy groups. Some original samples of suboptimal quality were not accurately identified. The present results indicated that sample pooling coupled with streamlined genotyping offers a promising approach for cost-effective gene sequencing of SARS-CoV-2, which will aid in genomic surveillance efforts against COVID-19.

Sample pooling

Gene sequencing

SARS-CoV-2

Simulation study

The impact of the COVID-19 pandemic on both human health and the social economy has been profound. With the continuous evolution and spread of the SARS-CoV-2 virus, ongoing genomic surveillance remains crucial [1]. Next-generation sequencing (NGS) is widely employed to monitor viral evolution. Notably, the sequences of SARS-CoV-2 shared on GISAID vary significantly across continents, with Europe contributing the most and Africa the least. Cost and sequencing throughput are likely limitations affecting the use of genomic surveillance in tracking COVID-19. These factors can influence our understanding of the prevalence of variants of concern or interest (VOCs/VOIs).

During the COVID-19 epidemic, the practice of pooling SARS-CoV-2 samples has been utilized for population screening due to its recognized benefits [2–3]. Pooling tests have increased the throughput of molecular testing and reduced costs [4], making them an economically advantageous option. However, the challenge lies in pooling sequencing, specifically in accurately estimating the composition of multiple SARS-CoV-2 lineages within samples containing mixed viral populations. Existing tools for SARS-CoV-2 lineage classification, such as pangolin or UShER, are primarily designed for clinical samples dominated by a single variant [5, 6]. Fortunately, the development and successful testing of Freyja, a tool capable of estimating the relative abundance of virus lineages within mixed samples like wastewater, presents a promising solution [7]. Given that patient swab samples often have higher viral loads and are less diluted compared to wastewater samples, it appears feasible to use Freyja to identify lineages in pooled sequencing of populations effectively.

This study aimed to assess the efficacy of lineage classification for SARS-CoV-2 in various simulated pooling samples. The objective was to evaluate the potential of pooled sequencing as a genomic surveillance tool for COVID-19 infectious disease.

This study included 72 original samples and 70 simulated pooling samples. Among these, 39 simulated samples precisely matched the original genotypic compositions, while 24 samples had one missing genotype and 7 samples had two missing genotypes as determined by the top X results. In contrast, 61 simulated samples exactly replicated the genetic compositions of the original samples, with 9 samples containing one missing genotype as determined by the complete results. The genetic compositions of the simulated pooling samples largely reflected the correct variant proportions of the original samples, with variations observed between different pooling strategies, as depicted in Fig. 1a.

Moreover, by categorizing consistent results as identified and missing results as unidentified, the statistical analysis revealed significant differences in the top X and complete ranking, with a chi-square value of 16.94 (P < 0.001). While there was variation in the top X calculation method among the five groups, the complete calculation method showed no significant differences in the pooling strategies employed(Table 1).

Table 1

The consistency of simulated pooling samples and original samples among various pooling strategies
		consistency
VariableOverall, N = 140		Identified N = 100(71%)^#	Unidentified N = 40(29%)^#	p-value
result-rank				< 0.001* ⁾
Top X	70 (50.00%)	39 (39.00%)	31 (77.50%)	(χ²=16.94)
complete	70 (50.00%)	61 (61.00%)	9 (22.50%)
Pooling strategy-top X				0.003**
Mix	5 (7.14%)	0 (0.00%)	5 (16.13%)
Change	30 (42.86%)	15 (38.46%)	15 (48.39%)
Genotype	15 (21.43%)	7 (17.95%)	8 (25.81%)
Subtype	8 (11.43%)	6 (15.38%)	2 (6.45%)
Recombinants	12 (17.14%)	11 (28.21%)	1 (3.23%)
Pooling strategy-complete				0.256**
Mix	5 (7.14%)	3 (4.92%)	2 (22.22%)
Change	30 (42.86%)	26 (42.62%)	4 (44.44%)
Genotype	15 (21.43%)	13 (21.31%)	2 (22.22%)
Subtype	8 (11.43%)	7 (11.48%)	1 (11.11%)
Recombinants	12 (17.14%)	12 (19.67%)	0 (0.00%)
#: n (%)
*: Pearson’s Chi-squared test
**: Fisher’s exact test

Regarding specific genotypes, the majority of original samples were accurately identified within mixed samples based on both the top X results and complete results. However, a few original samples were not correctly identified in either the top X results or the complete results. Examples include B.1.1.48 (mistaken for B.1.551 or missing), BA.2.76 (not consistently detected in most simulated samples), and XBB.1.16.2.1.1 (identified as XBB.1.16.2 or XBB.1.16.2.1). This is illustrated in Fig. 1b.

As the causal agent of coronavirus disease 2019 (COVID-19), SARS-CoV-2 is characterized by a high mutation rate, leading to the rapid emergence and global dissemination of new viral variants [8]. Consequently, ongoing real-time genomic surveillance is essential to track the spread of the disease and monitor viral evolution. In regions where sequencing capacity is limited and demand is high, pooled sequencing offers a cost-effective strategy to obtain substantial genomic data, thus enhancing our understanding of the virus’s evolutionary dynamics.

The bioinformatics analysis tool Freyja was utilized to estimate the relative frequencies of SARS-CoV-2 variants in pooled samples, employing a statistical model that incorporates a predefined set of genomic polymorphisms specific to these variants [7]. This approach has demonstrated efficacy in monitoring sewage samples across several countries [9, 10]. In this study, the software, based on Freyja, effectively identified the correct variant proportions in the majority of simulated samples. Nonetheless, there are notable considerations regarding the implementation of pooled sequencing in genomic surveillance that warrant further attention.

Not all simulated samples were accurately recovered, primarily due to original samples lacking definitive mutations and exhibiting lower extraction quality. For instance, the simulated samples harboring B.1.1.48 (assessed as low-quality by nextclade) and BA.2.76 (with less than 96% coverage) led to 36.8% unidentified cases among top X results and 80% among complete results. Therefore, ensuring the quality of original samples before pooling is essential, as those of poor quality can compromise identification accuracy. Additionally, observations indicated that samples with lower viral loads could be overlooked in the final pooled real-time reverse transcription polymerase chain reaction (RT-qPCR) result [11], emphasizing the importance of pooling samples with similar nucleic acid concentrations.

The software utilized for analyzing genomic composition ranked the abundance, revealing differences between the top X results and the complete results. Upon simulating various pooling scenarios, statistical discrepancies were noted among different pooling strategies, particularly when focusing on the top X results. Optimal consistency between the simulated samples and original samples was observed in the recombinants group, while the mix group showed the lowest level of consistency. These findings suggest that the complexity of the pooled samples significantly impacts the accuracy of identification. The models of pooling tests in RT-qPCR [12, 13] and wastewater-based epidemiological monitoring [14] serve as valuable references for the practical implementation of pooled sequencing for SARS-CoV-2.

While the overall results demonstrated improved identification compared to the top X approach, there were instances of inaccurate genotyping due to the presence of low-abundance mutation mixtures. The freyja method incorporated a bootstrap technique to calculate standard errors for predicting variant compositions. However, determining the optimal cutoff value for genomic composition results that strikes a balance between sensitivity and specificity remained a challenging task.

The study commenced by sporadically conducting gene sequencing on a pooled sample of routine RT-qPCR tests, successfully recovered the genotypes present in the individual samples. Due to the impracticality of pooling numerous samples into diverse groups, a simulated study was conducted to assess the feasibility of pooled sequencing. An additional constraint is that pooled sequencing can solely determine the viral lineage compositions and abundance, necessitating individual identification when new variants emerge, akin to pooling tests in RT-qPCR [15].

In summary, this study utilized simulated mixed samples to assess the feasibility of pooled sequencing with analysis using the Freyja tool. The findings demonstrated the successful recovery of the gene composition of the original samples. Therefore, pooling sequencing presents itself as a promising tool that can enhance genomic surveillance efforts in combating COVID-19 in a cost-effective manner.

Original samples

The original raw data in fastq.gz format from gene sequencing procedures carried out using Illumina NGS platforms were collected in our laboratory between 2020 and 2024. This dataset encompasses samples from individuals and vaccine materials. Subsequently, samples meeting our criteria of at least 96% coverage and a depth exceeding 1000X for the entire genome of SARS-CoV-2 were identified as genotype-determined specimens.

The study established various genotype groups at different hierarchical levels: four in the first-level classification, six in the second-level classification, and thirty-eight in the third-level classification. A total of 72 samples representing different subtypes were included in the study (ESM_1.pdf). For each sample, 500,000 raw data reads were randomly extracted using the Subsample tool of CLC Workbench 23.0 software (Qiagen, Dusseldorf, Germany) to create original samples.

Simulation of sample pooling

Five groups of mixtures were created to simulate sample pooling for gene sequencing using the Create Sequence tool within the CLC Workbench 23.0 software [Qiagen, Düsseldorf, Germany). The details of the simulated samples and mixtures are presented in ESM_2.pdf.

A mixture of five groups was formed based on the second-level classification, excluding Recombinants. One random sample from each group was selected and mixed equally, labeled as “Mix.”

A mixture of transition between two groups, comprising movement from the original* stage to the BA* stage, and from the BA* stage to the XBB* stage. Random samples were selected from each of the two groups at varying ratios of 1:4, 3:2, and 4:1, respectively, and combined as a mixture labeled as “Change.”

The samples were classified into three specific groups based on the first-level classification, excluding Recombinants. Five random samples were selected from each group and combined, then labeled as “Genotype.”

The samples were grouped into eight categories based on prevalent lineages, with Delta\XBB.1.5\XBB.1.9.1\XBB.1.9.2\XBB.1.16\BA.5.2\BA.5.2.48&49\JN being the most common. Each group was combined to create one simulated sample labeled as “Subtype.”

Three recombinants of XBF/XBL/XBC, along with three simulated recombinants of XBG/XBC/XDD, were identified and labeled as “Recombinants”.

Data analysis

The combined data from simulated samples underwent genotype and abundance analysis using the bioinformatics software module within the Pathogenic Microbial Analysis System. Specifically, the analysis was conducted utilizing the SARS-CoV-2 Analysis Module for Environmental Samples (V1.0.6, MicroFuture, Beijing, China). This system is underpinned by the freyja Algorithm and makes use of the genotyping tool nextclade (https://clades.nextstrain.org/).

The study assessed the lineage abundances, ranked genotype composition by abundance, and examined the genetic consistency between simulated pooled samples and the original samples. This evaluation was conducted based on the top X results (X representing the number of original samples in the mixture) and the complete results.

The categorical variables were analyzed using either the Chi-square test or Fisher’s exact probability method in R software Version 4.2.1. Statistical significance was defined as a p-value < 0.05.

Author contributions

All authors contributed to the study conception and design. Gene sequencing and original data preparation were performed by Xun He, Yuzhen Zhou, Wenjun Xie, Danyun Shen, Zhiqun He, Ruidan Li, Weixuan Liu. The study design, data collection and analysis were performed by Heng Chen, Yue Cheng, Liang Wang ,Xuejun Zhang and Xiang Zhao. The first draft of the manuscript was written by Heng Chen and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Acknowledgements

The colleagues involved in the epidemiological investigation, sampling, laboratory testing, and data analysis, as well as the patients who consented to donate their swab samples for detection, are acknowledged.

Fundings

This study was supported by the Key Technology Research and Development Program of the Science and Technology Department of Sichuan Province [2021YFS0064], the Youth Innovation Research Project by the Sichuan Medical Association [Q22017], and the Science Foundation of the Health and Family Planning Commission of Chengdu, Sichuan, China [22056].

Competing Interests and Declarations

The authors have no relevant financial or non-financial interests to disclose.

Markov, P. V, Ghafari, M, Beer, M, Lythgoe, K, Simmonds, P, Stilianakis, N. I, Katzourakis, A. (2023). The evolution of SARS-CoV-2. Nature reviews. Microbiology, 21(6), 361–379. https://doi.org/10.1038/s41579-023-00878-2
Chong, B. S. W, Tran, T, Druce, J, Ballard, S. A, Simpson, J. A, Catton, M. (2020). Sample pooling is a viable strategy for SARS-CoV-2 detection in low-prevalence settings. Pathology, 52(7), 796–800. https://doi.org/10.1016/j.pathol.2020.09.005
Augenblick, N, Kolstad, J, Obermeyer, Z, Wang, A. (2022). Pooled testing efficiency increases with test frequency. Proceedings of the National Academy of Sciences of the United States of America, 119(2), e2105180119. https://doi.org/10.1073/pnas.2105180119
Grobe, N, Cherif, A, Wang, X, Dong, Z, Kotanko, P. (2021). Sample pooling: burden or solution?. Clinical microbiology and infection : the official publication of the European Society of Clinical Microbiology and Infectious Diseases, 27(9), 1212–1220. https://doi.org/10.1016/j.cmi.2021.04.007
Rambaut, A, Holmes, E. C, O'Toole, Á, Hill, V, McCrone, J. T, Ruis, C, du Plessis, L, Pybus, O. G. (2020). A dynamic nomenclature proposal for SARS-CoV-2 lineages to assist genomic epidemiology. Nature microbiology, 5(11), 1403–1407. https://doi.org/10.1038/s41564-020-0770-5
Turakhia, Y, Thornlow, B, Hinrichs, A. S, De Maio, N, Gozashti, L, Lanfear, R, Haussler, D, Corbett-Detig, R. (2021). Ultrafast Sample placement on Existing tRees (UShER) enables real-time phylogenetics for the SARS-CoV-2 pandemic. Nature genetics, 53(6), 809–816. https://doi.org/10.1038/s41588-021-00862-7
Karthikeyan, S, Levy, J. I, De Hoff, P, et al. (2022). Wastewater sequencing reveals early cryptic SARS-CoV-2 variant transmission. Nature, 609(7925), 101–108. https://doi.org/10.1038/s41586-022-05049-6
Flores-Vega, V. R, Monroy-Molina, J. V, Jiménez-Hernández, L. E, Torres, A. G, Santos-Preciado, J. I, Rosales-Reyes, R. (2022). SARS-CoV-2: Evolution and Emergence of New Viral Variants. Viruses, 14(4), 653. https://doi.org/10.3390/v14040653
Khan, M, Li, L, Haak, L, Payen, S. H, Carine, M, Adhikari, K, Uppal, T, Hartley, P. D, Vasquez-Gross, H, Petereit, J, Verma, S. C, Pagilla, K. (2023). Significance of wastewater surveillance in detecting the prevalence of SARS-CoV-2 variants and other respiratory viruses in the community - A multi-site evaluation. One health (Amsterdam, Netherlands), 16, 100536. https://doi.org/10.1016/j.onehlt.2023.100536
Reis, A. C, Pinto, D, Monteiro, S, Santos, R, Martins, J. V, Sousa, A, Páscoa, R, Lourinho, R, Cunha, M. V. (2024). Systematic SARS-CoV-2 S-gene sequencing in wastewater samples enables early lineage detection and uncovers rare mutations in Portugal. The Science of the total environment, 921, 170961. Advance online publication. https://doi.org/10.1016/j.scitotenv.2024.170961
Mulu, A, Alemayehu, D. H, Alemu, F, Tefera, D. A, Wolde, S, Aseffa, G, Seyoum, T, Habtamu, M, Abdissa, A, Bayih, A. G, Beyene, G. T. (2021). Evaluation of sample pooling for screening of SARS CoV-2. PloS one, 16(2), e0247767. https://doi.org/10.1371/journal.pone.0247767
Hanel, R, Thurner, S. (2020). Boosting test-efficiency by pooled testing for SARS-CoV-2-Formula for optimal pool size. PloS one, 15(11), e0240652. https://doi.org/10.1371/journal.pone.0240652
Perchetti, G. A, Sullivan, K. W, Pepper, G, Huang, M. L, Breit, N, Mathias, P, Jerome, K. R, Greninger, A. L. (2020). Pooling of SARS-CoV-2 samples to increase molecular testing throughput. Journal of clinical virology : the official publication of the Pan American Society for Clinical Virology, 131, 104570. https://doi.org/10.1016/j.jcv.2020.104570
Valieris, R, Drummond, R. D, Defelicibus, A, Dias-Neto, E, Rosales, R. A, Tojal da Silva, I. (2022). A mixture model for determining SARS-Cov-2 variant composition in pooled samples. Bioinformatics (Oxford, England), 38(7), 1809–1815. https://doi.org/10.1093/bioinformatics/btac047
Hogan, C. A, Sahoo, M. K, Pinsky, B. A. (2020). Sample Pooling as a Strategy to Detect Community Transmission of SARS-CoV-2. JAMA, 323(19), 1967–1969. https://doi.org/10.1001/jama.2020.5445

No competing interests reported.

Download PDF

Version 1

posted

You are reading this latest preprint version

Evaluation of Sample Pooling for Gene Sequencing of SARS-CoV-2: A Simulation Study

Status:

Version 1

Abstract

Figures

Introduction

Results

Discussion

Materials and Methods

Original samples

Simulation of sample pooling

Data analysis

Declarations

Author contributions

Competing Interests and Declarations

The authors have no relevant financial or non-financial interests to disclose.

References

Additional Declarations

Supplementary Files

Status:

Version 1