How does antibody testing for COVID-19 work?
An overview of the principle of the antibody testing used to diagnose COVID-19 and how artificial intelligence (AI) can help largely in designing these tools.
The science behind immunoassay is extremely simple. When humans are infected with a virus, specific antibodies that bind to virus proteins are produced. The principle of immunoassay is to detect these antibodies and produce a signal that can be read through an instrument (e.g. a plate reader).
In other words, viral protein is linked to the plastic support, and then the clinical specimen (e.g serum, swab, ..) is added. If antibodies against the virus are present in the specimen, they will bind to the immobilized viral antigen. The bound antibodies are then detected by using a second antibody that binds to the first antibody. This secondary antibody is linked with an indicator (e.g HRP, fluorescent) which usually needs a plate reader to detect/read it.
Although this method is indirect and less sensitive than polymerase chain reaction (PCR), it is extremely simple and doesn’t need prior preparations for the clinical samples (e.g. PCR needs RNA extraction which is time-consuming). The clinical samples can be applied immediately without fractionation. These kits usually produce results within 30–60 mins and therefore can test a considerable number of people. The problem is the need for special instruments (the plate reader) to read the signal.
Can we optimize the immunoassay to create point-of-care (POC) testing?
The short answer is yes.
We can use the paper-based point-of-care (POC) immunoassays which are disposable POC serology tests that are similar to pregnancy tests. These are simply composed of three major components, i.e., paper as the substrate, antibodies as the detection element, and reporter molecules as the signal-transforming element. They are also called lateral flow immunoassays (LFIA). The interesting part is that these tests can be read visually (with naked eyes), or provide data when combined with reader technology.
For example, in the last 2–3 weeks, both Biomerica and Bioeasy have released their COVID-19 Rapid POC CE-IVD Test.
Briefly, there are three lines in the kit (IgM, IgG, control).
Why are there three lines rather than two (positive and negative)?
The answer is different antibodies can be used to help a doctor tell the difference between a new and past infection. A sample can be positive if there are IgM (these are the first antibodies to appear in response to the virus which implies a recent infection) or IgG (this implies an infection in its later course). Both IgM and IgG antibodies can also present in positive samples.
The nose and the back of the throat are the two sites where the virus is replicating, so the sample is taken from these sites and applied in the sample pad. Then, the sample interacts with labeled COVID-19 antigen with gold nanoparticles (explained later), then moves through the membrane by capillary force, and finally accumulates in the test (IgG and IgM lines) and control zones because of the binding to capture antibodies.
In a positive test, the sample/conjugate complex moves to the first line which contains an immobilized antibody that recognizes human IgM. Only human IgM antibody/COVID-19 antigen/gold nanoparticle complexes will produce a visible colored line. The results are produced within 10 minutes.
Different labels directly affect the detection results. Gold nanoparticles (AuNPs) are the nanomaterial used in many clinical diagnosis tools because of their ability to form conjugates with biomolecules (e.g., proteins) and due to their stability and intense color.
How artificial intelligence can help in optimizing these kits?
AI plays an extremely important role in optimizing the labels (e.g.AuNPs). “Whether the labels can be tightly combined with the detectors is a critical criterion to evaluate their quality. They should also be able to retain their properties upon conjugation and be detectable when their concentrations are low. Its biological activity is also not expected to change in terms of sensitivity and reproducibility during movement and over a short time”.
Therefore, scientists use different mathematical models and algorithms for lateral flow immunoassays (LFIA) to optimize the test results.
However, these kits are not accurate enough to detect an ongoing infection since it doesn’t detect the virus directly. However, these tests will be extremely needed to identify the proportion of the population who have contracted SARS-CoV-2 or even people who are asymptomatic and potentially those with immunity to the virus.
