The rapid identification and isolation of infected individuals remains a key strategy for controlling the spread of SARS-CoV-2. Frequent testing of populations to detect infection early in asymptomatic or presymptomatic individuals can be a powerful tool for intercepting transmission, especially when the viral prevalence is low. However, RT-PCR testing—the gold standard of SARS-CoV-2 diagnosis—is expensive, making regular testing of every individual unfeasible. Sample pooling is one approach to lowering costs. By combining samples and testing them in groups the number of tests required is reduced, substantially lowering costs. Here we report on the implementation of pooling strategies using 3-d and 4-d hypercubes to test a professional sports team in South Africa. We have shown that infected samples can be reliably detected in groups of 27 and 81, with minimal loss of assay sensitivity for samples with individual Ct values of up to 32. We report on the automation of sample pooling, using a liquid-handling robot and an automated web interface to identify positive samples. We conclude that hypercube pooling allows for the reliable RT-PCR detection of SARS-CoV-2 infection, at significantly lower costs than lateral flow antigen (LFA) tests.

Introduction

A novel coronavirus, SARS-CoV-2, emerged at the end of 2019 in the city of Wuhan, China. The highly transmissible nature of SARS-CoV-21 has resulted in a pandemic which continues to persist. South Africa (SA) and other African countries are currently facing a resurgence or “third wave” of infections, which in some countries is more severe than previously experienced2,3,4. Although there are approved vaccines against SARS-CoV-2, these are not yet available in amounts sufficient to control the pandemic. To make matters worse, there are new SARS-CoV-2 variants (which are typically identified using genomics surveillance tools5,6) that have been identified in SA7, which may be associated with higher transmissibility and hence a more rapid spread of the virus. There is therefore an urgent need for more efficient population screening and the isolation of infected individuals to reduce the transmission of SARS-CoV-2.

Presymptomatic or asymptomatic individuals, who are infectious viral carriers8, are the hidden drivers of the pandemic. They represent an estimated proportion ranging from 18 to 81% of infections9, thereby posing a major challenge to the containment of SARS-CoV-2. If such individuals can be efficiently detected through frequent, repeated population testing at scale and thereby enabled to isolate before they infect others, the spread of the virus can be prevented8. Therefore, efficient and affordable, high throughput SARS-CoV-2 testing is highly desirable as a means of controlling the pandemic.

Reverse-transcription real-time polymerase chain reaction (RT-PCR) testing is the gold-standard technology used for SARS-CoV-2 diagnosis. This test can cost up to US$ 56 (ZAR 850) per test, making high-throughput RT-PCR testing of every individual impractical. Sample pooling offers an attractive solution. By combining samples and testing them together, instead of performing individual tests, one can significantly reduce the number of tests and the associated labour and consumable costs. This method was first proposed by Dorfman10 in 1943. Hypercube pooling, developed by Mutesa et al.11, requires even fewer tests and yields greater cost savings. For example, at viral prevalences p < 0.05%, hypercube pooling yields a 100-fold cost-reduction as opposed to a 22-fold for Dorfman’s algorithm. Hypercube pooling therefore offers a highly affordable means of testing large numbers of samples. We also describe a quantitative cost comparison with lateral flow antigen (LFA) tests, finding hypercube pooling to have a significant cost advantage at low viral prevalence (Supplementary Information—Cost comparison of hypercube-based pooled tests vs. lateral flow antigen tests).

Hypercube pooling has been eloquently explained in the literature11. Briefly, the samples to be tested are divided into equally sized subsamples which are pooled together according to a mathematical algorithm. Here we report on pooled testing methods which can uniquely identify infected samples among groups of 27 and 81 samples (using 3- and 4-dimensional hypercubes, respectively), in far fewer than 27 or 81 tests. If the group is negative, then all individual samples in that group are deemed negative. If a group is positive, then its sub-samples are recombined in the form of slices within the hypercube corresponding to different overlapping sub-pools. Groups of 27 or 81 samples are sub-pooled into 9 and 12 slices respectively, each consisting of 9 or 27 subsamples respectively. Each individual sample is then represented in 3 or 4 different slices respectively, and the test results for the slices can be used to infer which sample is positive, based on its consistent detection within each of the hypercube slices, without an individual test ever being required. The slicing patterns for these groups are shown in the Supplementary Information Tables S1 and S2 and the complete workflows are presented in Fig. 1.