The numbers of diagnostic tests for SARS-CoV-2 are exploding.* These tests detect either viral RNA or antibodies to virus particles. But, which is best? That, of course, depends on what question is being asked.
To help communicate how the different kinds of tests work, what they measure, and when they are best used, we've prepared an infographic.
The Active and Post Infection Biomolecule Profile Defines the Appropriate Test
The first step in understanding which test is best is understanding the viral infection life cycle. Technically viruses are not living. They infect cells and turn their host's molecular machinery into virus production factories. Infected cells produce large numbers of viral particles (virions) with each new virus infecting another cell, repeating the process, and rapidly increasing the numbers in a relatively short period of time. Over a few days, the numbers of infectious virions increases substantially. Later, the infection subsides and the number of virions decreases. This decrease is related to either the hosts' immune system fighting off the virus or due to the virus going "dormant" within the host. If neither happens, the host can die.
What the viral life cycle means from a diagnostic testing perspective, is that certain kinds of biomolecules such as viral nucleic acids (RNA or DNA) or the host's antibodies are abundant at different times. For this discussion abundant means we can detect them. As Molecular tests detect nucleic acids and Serologic Tests detect antibodies, they are best suited for times when those respective biomolecules are available.
A SARS Coronavirus infection (redrawn from (1)) is shown in the infographic (right). Remember the first SARS (Severe Acute Respiratory Syndrome, ca. 2002-2004)? It was an earlier episode of what we are now experiencing, but with a more severe infection from a less infectious corona virus. Hence, it did not require the extreme counter measures that are now needed to slow the current pandemic. A benefit of that first SARS experience, however, was that it produced data that we can use to make predictive models of the current SARS-CoV-2 infections. As observed in the graph, the first biomolecules are viral RNA (vRNA - corona viruses are RNA viruses). As the virus replicates, the number of virions (each with a vRNA molecule) increases and their RNA becomes more detectable. Later, as the concentration of vRNA drops, and the number of antibody molecules increases. The first antibodies are the pentameric IgM antibodies. Later, the cells producing IgM switch to making IgG. Thus, vRNA is a biomarker for detecting active infections, and IgM/IgG are biomarkers for late active and post infection monitoring.
Timelines are indicated in the biomolecule profile. Active infections can last for 21-28 days; perhaps longer in some people. This is why people who become sick should quarantine for at least 14 days, as it can take up to two weeks to show symptoms and then another two weeks for recovery. After 7-14 days, host antibodies are produced, and are present much longer. IgM falls off after a few weeks to months and IgG can persist for months to years. But, while we have some data for 2002-2004 SARS coronavirus the persistence of antibodies specific to SARS-CoV-2 proteins is unknown (question marks in the graphic). The only way to know how long antibodies persist is to monitor COVID-19 patients for a long time. Some of this science is underway (2-4).
The biomolecule profile tells us that molecular tests can only be used to detect active infections and serologic tests are best for post infection monitoring. So how do they work? And, how do we ensure the results are meaningful? The following sections describe the classic forms of the tests. It is worth noting that the number of companies and organizations developing tests is rapidly increasing,* and they are developing many variations on the basic themes. These variations seek to improve test parameters that affect test cost, quality, and time. But, the fundamental ways in which the tests work, and sources of test error, have not changed since they were first developed between ~30 and 50 years ago.**
Why are Perfect Tests Impossible?
The goal of any test is 100% perfection so that every person who has an infection tests positive, and every person who does not have an infection, tests negative. Perfection is not achievable in the real word, and the dreaded terms sensitivity and specificity are used to define a test's imperfection.
Sensitivity describes the smallest number of molecules (the test threshold) that you need to say a test result is positive. When relatively more molecules are needed for detection, false negative results occur. These would be individuals that have an infection, but their number of molecules result in a signal that is below the detection threshold. Thus, their infection cannot be detected.
Specificity is the converse and defines how well the test can discriminate between different molecules, or signal from noise. Using the above detection threshold as an example, if the assay has too many false negative results, we may "lower the bar" so that the assay detects a smaller number of molecules. The assay is made more sensitive because it detects lower signals. Now, some the previously identified false negatives become true positives, but some of the true negatives become false positives. In this way sensitivity and specificity are interrelated.
Sensitivity and specificity values are typically expressed as ratios, converted to percent, of true (positive or negative) results to the total results. So a test that has 95% sensitivity means 5% of those testing negative either are, or have been infected. Likewise a specificity of 95% means 5% of those testing positive either are not, or have not been infected. Which measure is more important, sensitivity or specificity? That depends of the goal of the diagnostic assay.
Common molecular tests use PCR (polymerase chain reaction) to detect DNA molecules. In PCR, synthetic DNA primers that complement DNA sequences flanking a region of interest, are combined with a heat-stable DNA polymerase enzyme, the four DNA building blocks (2-deoxyadenosine triphosphate, 2-deoxycytosine triphosphate, 2-deoxyguanosine triphosphate, and 2-deoxythymidine triphosphate; together dNTPs), and a detector. Repeated cycles of high and low temperatures are used to bind primers, synthesize DNA, and melt the pieces apart. If the DNA region of interest is present, each cycle doubles the amount of that DNA. The amount of DNA produced can be measure using a detector (also known as quantitative, qPCR). The simplest is a dye that binds double stranded DNA (dsDNA) and becomes fluorescent. The fluorescence intensity increases with increasing amounts of dsDNA. But coronaviruses contain RNA, and PCR amplifies DNA. To overcome this issue the RNA is first converted to DNA using an enzyme (RT, reverse transcriptase) that makes DNA copies from RNA. Hence to detect vRNA, RT-PCR is used.
An advantage of molecular testing is that it is non-invasive. Viral particles used in tests are obtained from swabs that have been wiped on some surface. While having a swab up your nose or down your throat may seem invasive, medically speaking it is not. Because molecular testing detects viral RNA it can also be used to test surfaces where viruses might be, or tell us where viruses may have been. In either case, it is important to note that molecular tests only detect the nucleic acids, they cannot tell us if those nucleic acids are in infectious virions.
Molecular Testing Considerations
Factors that affect the sensitivity of molecular testing include contamination, non-specific sequence hybridization, PCR artifacts, and having a detection threshold that is too high. PCR-based assays can be extremely sensitive and contamination can be a big issue. Just ask the CDC (Centers for Disease Control (5)). Assay threshold is specified by the assay design and is set by which is more acceptable, false positives or false negatives.
In the case of COVID-19, the consequence of a false positive is quarantine. Not so bad when one considers each active infection can produce more infections. False negatives on the other hand mean that infected people do not know they are infected and can go on to infect more people. In the case of non-specific sequence hybridization where PCR primers match desired and undesired targets, false positives are reduced by assay design and good controls that include DNA from other kinds of common infections. Similarly PCR artifacts, where primers bind other primers and result in unintended DNA amplification, can be controlled with tests that do not contain any added DNA (known as blanks that of just assay buffer). Blank samples also control for contamination.
As specificity is the yin to the yang of sensitivity, false negatives are affected by the assay's sensitivity threshold, RNA degradation, DNA sequence variation, and when tests are conducted. RNA degradation is the opposite of contamination. That is, testable material is not present. Degradation is avoided with proper sample handling and treating positive control samples in the same way as test samples. DNA sequence variation, where mutations change the sequence of an expected primer binding site can reduce or "block" amplification. This kind of error is controlled for by designing assays that target multiple locations, so if one region fails other regions can still be amplified. The last issue, when tests are conducted, is important. If infected individuals are tested too early, before viral replication, their infections can be missed. As we are not yet to point of mass testing, where healthy and infected individuals are tested and monitored, this kind of error remains unknown. Finally, positive controls that give expected results tell us that a test is working properly.
Serological testing measures antibodies that bind to viral proteins. Antibodies are found in blood so serological testing, unlike molecular testing, is technically invasive becuase a needle must be used to puncture skin and draw blood. Also known as immunoassays, serologic tests, like molecular tests, follow a common format. In this example the format is an ELISA (Enzyme Linked Immunosorbent Assay). There are many variations on this theme too.
In the general approach, a solid surface is coated with antigenic material. The material can be intact (non-infective) virions, purified viral proteins, or even synthetic peptides. The solid surface is first incubated with blood (serum) from individuals. If anti-viral antibodies are present, they bind to the antigens. After a period of time, the surface is washed to remove unbound material. Next, another antibody, produced in a different organism (with goats and rabbits being favorites), that binds to human antibodies, is added. This antibody has an enzyme reporter attached to it, hence "Enzyme Linked." After another incubation period, excess material is washed away and a chemical (substrate of the attached enzyme) is added. The attached enzyme converts the substrate into a colored product and the color intensity, as read by a spectrophotometer, defines positive results.
Serologic Testing Considerations
Compared to molecular tests, serologic tests are fairly simple. Add sample, wash, add anti-antibody-enzyme, wash, and detect. What could go wrong? Plenty. In terms of sensitivity, false negatives come from the usual culprit of detection thresholds that are set too high. As above, detection thresholds are defined when the assay is developed. Thresholds trade the ability to detect every person who has been infected with the possibility that some people, who test positive, may not have been infected. In the case of "immunity cards," it's probably not a good idea to hand out cards to those who have never been infected, so we'd want high specificity with our false positive rate being as low as possible. Even then, the presentence of anti-SARS-CoV-2 antibodies is not a guarantee of immunity, and further limits the usefulness of immunity cards.
Contamination is controlled by strong manufacturing quality control (GMP: Good Manufacturing Practices) and test operation (having well qualified operators and standard operating procedures [SOPs]). SOPs also minimize the likelihood of false positive and false negative results that arise from under or over washing, or over/under incubation times. Antibodies, like DNA primers, often bind to things we do not want them to bind to. This is even more likely if care is not taken. For example, other coronaviruses may have similar protein spikes. As some of these cause common colds, they could cause false positives in assays if the proper controls are not used.
A larger challenge is testing at the right time in the infection cycle. As observed in the biomolecule profile, antibodies develop after a person has been infected. This development is highly variable with respect to the onset and strength of the immune response in each individual. Thus far, we have little data on COVID-19 immune responses in large populations so false negatives due to assay timing, how long antibodies persist, and assay thresholds is expected.
In short, antibody tests can have more variability than molecular tests .
Molecular and serologic tests are being used to identify SARS-CoV-2 / COVID-19 infections. Molecular (RT-PCR) tests only measure active infections and serologic (immunoassay) tests are best for late stage and post infection monitoring. No test is perfect and both methods have several kinds of issues that give false positive and false negative results. False negative results lower a test's sensitivity and false positive results lower a test's specificity. Tests are designed to ask certain questions and trade sensitivity for specificity and vise versa. Errors related to contamination and assay operation can be avoided with good GMPs and strict SOPs. Assay controls and expected outcomes provide additional verification to ensure the highest sensitivities and specificities can be met.
And, one more thing, antibodies do not equal immunity. That's a topic for another time.
* To learn more about companies working on COVID-19, visit https://biotech-careers.org/company-core-activity/covid-19
** First ELISA (1971), https://en.wikipedia.org/wiki/ELISA;
** First RT-PCR (1988), Search of PubMed with "reverse transcription pcr" gives: Genomic amplification with transcript sequencing. Stoflet ES, Koeberl DD, Sarkar G, Sommer SS. Science. 1988 Jan 29;239(4839):491-4.
** First qRT-PCR (2000), What we're discussing, https://en.wikipedia.org/wiki/Reverse_transcription_polymerase_chain_rea....
 Longitudinal Profile of Immunoglobulin G (IgG), IgM, and IgA Antibodies against the Severe Acute Respiratory Syndrome (SARS) Coronavirus Nucleocapsid Protein in Patients with Pneumonia Due to the SARS Coronavirus. Woo, Patrick & Lau, Susanna & Wong, Beatrice & Chan, Kwok-Hung & Chu, Chung-Ming & Tsoi, Hoi-wah & Huang, Yi & Peiris, Joseph S & Yuen, Kwok-yung. (2004). Clinical and diagnostic laboratory immunology. 11. 665-8. 10.1128/CDLI.11.4.665-668.2004.
 Comparative Performance of SARS-CoV-2 Detection Assays using Seven Different Primer/Probe Sets and One Assay Kit. Nalla AK, Casto AM, Huang MW, Perchetti GA, Sampoleo R, Shrestha L, Wei Y, Zhu H, Jerome KR, Greninger AL. J Clin Microbiol. 2020 Apr 8. pii: JCM.00557-20. doi: 10.1128/JCM.00557-20.
 Temporal dynamics in viral shedding and transmissibility of COVID-19. He X, Lau EHY, Wu P, Deng X, Wang J, Hao X, Lau YC, Wong JY, Guan Y, Tan X, Mo X, Chen Y, Liao B, Chen W, Hu F, Zhang Q, Zhong M, Wu Y, Zhao L, Zhang F, Cowling BJ, Li F, Leung GM. Nat Med. 2020 Apr 15. doi: 10.1038/s41591-020-0869-5. [Epub ahead of print]
 Virological assessment of hospitalized patients with COVID-2019. Wölfel R, Corman VM, Guggemos W, Seilmaier M, Zange S, Müller MA, Niemeyer D, Jones TC, Vollmar P, Rothe C, Hoelscher M, Bleicker T, Brünink S, Schneider J, Ehmann R, Zwirglmaier K, Drosten C, Wendtner C. Nature. 2020 Apr 1. doi: 10.1038/s41586-020-2196-x. [Epub ahead of print]
 Contamination at CDC lab delayed rollout of coronavirus tests - The Washington Post (04/18/2020)
 Coronavirus Antibody Tests: Can You Trust the Results? (04/24/2020)
Acknowledgements: This work was supported in part by the "START Immuno Biotech" NSF grant (DUE 1700441, Shoreline Community College). Dr. Sandra Porter (Digital World Biology) provided helpful comments and edits.