SARS-CoV-2 is a coronavirus from the beta coronavirus 2B lineage ^[1][37]. Coronaviruses are a family of RNA viruses that cause disease in mammals and birds. The human diseases that are primarily associated with respiratory tract infections range from the common cold to more severe diseases such as MERS, SARS, and COVID-19. All members of this family have characteristic projections from their surface called spikes, which give them a unique shape under the electron microscope resembling a stellar corona; thus, the name. Like other coronaviruses, SARS-CoV-2 has an envelope which protects its genome, a positive-sense single-stranded RNA (+ssRNA) with a size of around 29.9 kb ^[2][38].

SARS-CoV-2 has four structural proteins: S (spike protein), E (envelope protein), M (membrane protein), and N (nucleocapsid protein). It also contains 16 non-structural proteins named NSP1 through 16. The non-structural proteins perform a variety of functions essential for viral replication, infection, and life cycle. Some of the most well-known NSPs are NSP12 (RNA-dependent RNA polymerase) and NSP5 (main protease). The N protein is responsible for RNA binding and genome packaging, and, thus, is essential for viral replication. It contains an RNA binding domain as well as a dimerization domain; however, a large part of the protein is predicted to be intrinsically disordered ^[3][39].

The membrane is composed of M, E and S proteins in addition to lipids. The ratio of the E:S:M proteins in coronaviruses is around 1:20:300 ^[4][40]. The M proteins are the major structural proteins in SARS-CoV-2 and the most abundant of the three. The SARS-CoV-2 M protein is a 25-30 kDa O-glycosylated protein with three major domains: an N-terminal ectodomain, a transmembrane domain which passes the membrane three times, and a C-terminal endodomain ^[5][41]. The M protein is essential in all stages of the viral life cycle, from assembly to budding to infection.

The E proteins are the least abundant proteins on the surface of the virus, with nearly 20 copies present, and have different sequences across species ^[5][41]. This small protein of 8.4–12 kDa forms a pentameric ion channel in the membrane through their single-pass alpha helix ^[5][41]. The major role of the E protein is in assembly, trafficking, and morphogenesis ^[6][42].

The S protein is a glycosylated homotrimer which is in charge of binding to host receptors and causing infection ^[5][41]. Each monomer is around 150–200 kDa with around 1273 amino acids ^[5][41]. In SARS-CoV-2, the spike protein is known to bind to the receptor angiotensin-converting enzyme II (ACE2) ^[7][43]. The S protein is subdivided into two components: S1 (the head, including the receptor binding domain, or RBD, in addition to an N-terminal domain) and S2 (the stem, which contains multiple sub-domains, including a fusion peptide, a central helix, two heptad repeats, a transmembrane domain, and a cytosolic tail). S1 is in charge of binding to the host cell receptors while S2 is in charge of fusing the viral and host membranes. The high glycosylation level of the S protein is shown to play a role in folding as well as evading the host’s immune system ^[8][44]. The S protein can be in a closed or open state. The open state is necessary for binding to the receptor.

The RBD domain is responsible for binding to the ACE2 and starting infection. The core of this domain mostly consists of B sheets which are stabilized through several disulfide bonds. The core is highly conserved. The external subdomain is mostly dominated by loops which are stabilized through a disulfide bond. The interaction between the RBD domain and ACE2 is highly dominated by hydrophilic interactions and hydrogen bonding across the interface.

One hallmark of RNA viruses including SARS-CoV-2 is their rapid mutation rates ^[9][46]. Many mutations are silent, in that they do not change amino-acid residues and, thus, result in no functional consequences. However, there are many examples of mutations that can affect the structure and function of viral proteins and confer evolutionary advantages in terms of increased pathogenesis, faster transmission, or higher survival rates. Over the 2 years since the beginning of the COVID-19 pandemic, many new variants have developed, some of which have higher transmission rates or increased pathogenicity.

The three-electrode setup is the most commonly used setup for EIS measurements (as it is for other voltametric and amperometric measurements). This setup is composed of a working electrode (WE), a counter (auxiliary) electrode (CE), and a reference electrode (RE). Surface modifications and impedance analysis is carried out on the WE. Novel WE modification methods are the main component of interest in many electrochemical studies. The RE is a low-impedance electrode designed to establish a stable reference potential in the electrochemical system. Without a stable reference potential, we cannot be confident that changes in the impedance signal are due to WE surface changes and not due to drift. Most REs are developed using a combination of Ag and AgCl materials and function as quasi-reference electrodes given the presence of chloride in all biofluids. The CE is a low-resistance electrode (typically Pt or Au) which allows current flow between the CE and WE and enables impedance measurements.

The large majority of reviewed papers utilized a three-electrode system ^{[10][11][12][13][14][15][16][17][18][19][20][21][22][23][24][25][26][27][28]}[9,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65]. However, in EIS, because impedance is the main parameter, it is possible to operate with a two-electrode system ^{[29][30][31][32]}[66,67,68,69]. In theory, two-electrode systems with a WE and a combination RE/CE are inherently less stable with repeated measurements than their three-electrode counterparts ^[33][17]. A dedicated CE is important for providing a low-resistance path for current to flow to (1) prevent signal attenuation due to the high-resistance RE and (2) protect the RE from high currents, which can damage it and its ability to establish a stable reference potential ^[33][17]. However, the two-electrode systems reviewed demonstrated limits of detection (LOD) comparable to the best of their three-electrode counterparts, reaching past the femtomolar and femtogram/milliliter ranges ^{[29][30][31][32]}[66,67,68,69]. This impressive performance is likely due to extreme measurement ranges (Soares et al.), which can artificially improve LOD and the use of highly specific, small aptamers for binding (Ramanathan et al.) (the impact of these two factors will be discussed later) ^[29][30][66,67]. Of the four two-electrode papers, only Xue et al. measured repeatability within the same electrode, demonstrating an acceptable 4.7% (n = 3) relative standard deviation (RSD) ^[32][69].

Carbon-based electrodes are desirable because of their low cost and versatility in functionalization. Abrego-Martinez et al., Wu et al., Lorenzen et al., and Brazaca et al. used electrodeposition to add AuNPs to the surface of their carbon electrodes, while Hussein et al. added WO₃ by electroplating ^{[10][11][17][20][27]}[9,48,54,57,64]. Soares et al. added carboxymethyl chitosan to their carbon WE and Li et al. and Torres et al. added glutaraldehyde to their carbon electrodes ^[19][29][34][56,66,70]. Other groups opted to use the popular electroactive polymer poly(3,4-ethylenedioxythiophene) polystyrene sulfonate (PEDOT:PSS) on top of existing carbon and Au electrode material ^[20][32][35][57,69,71].

Electrodes have taken a variety of different form factors. They are typically two-dimensional, flat electrodes made of different materials. This form factor was popularized by commercially available screen-printed electrodes such as the Dropsens electrodes manufactured by Metrohm. Some groups opted to take advantage of this existing manufacturing technology and built their sensors by modifying the surface of existing commercially available sensors ^{[11][16][17][18][24][25][27][28][30][31]}[48,53,54,55,61,62,64,65,67,68].

Tepeli et al. (Dropsens), Soto and Orozco (Dropsens), Wu et al. (Zensor), Hussein et al. (Gamry), Abrego-Martinez et al. (Biodevice Technology), and Sharif et al. (Dropsens) purchased fully integrated electrodes ^{[11][16][17][24][25][27]}[48,53,54,61,62,64]. Soares et al., Ramanathan et al., Rashed et al. and Lasserre et al. utilized specially commissioned single or interdigitated electrodes ^{[18][29][30][31]}[55,66,67,68].

Others opted to manufacture electrodes themselves, and low-cost materials such as paper and thin films stood out from other materials due to their flexibility and versatile chemical properties (although Ashur et al. used conventioanl silicon lithography) ^{[10][12][13][15][19][21][22][23][32][34][35]}[9,49,50,52,56,58,59,60,69,70,71]. Researchers developed and patterned their electrodes onto these low-cost substrates with various methods. Pola et al. and Ali et al. used aerosol jet printing and Salahandish et al. used a flatbed microprinter to deposit electrode material ^[12][23][35][49,60,71]. Xue et al. used nanoscale soft printing to deposit Au wire, while Li et al. grew hydrothermal ZnO wire onto paper ^[19][32][56,69]. Nicoliche et al. used oven pyrolization to generate single-carbon WEs ^[21][58]. Perdomo et al. and Brazaca et al. used screen printing to deposit conductive ink onto their substrates ^[10][22][9,59]. Wax printing was also used to define microfluidic channels in paper-based devices ^[15][26][52,63]. Lorenzen et al. used a steel mesh as a sensing-platform substrate ^[20][57].

While all these substrates are relatively low-cost, two different paradigms of fabrication methodology emerge: one driven by cost-efficiency and scalability (i.e., wax and screen printing), and one focused on precision manufacturing (i.e., photolithography and hydrothermal growth); they tend to be inversely related. However, both paradigms seem to enable well-controlled batch electrode fabrication, which allows electrode miniaturization, lower required sample volumes, and reduced cost compared to qRT-PCR.

m ^[36][72]. The resolution of screen-printed electrodes depends on the contact angle of the specific ink and substrate, as well as the pseudoplasticity of the ink ^[37][73]. In lithography, the wavelength of the exposure system and the type of photoresistance are the two determining factors for resolution ^[38][74]. By varying wavelength, feature sizes from less than 1 nm to higher than 100 nm can be achieved: ion beam lithography results in feature sizes less than 1 nm, electron beam lithography systems can achieve feature sizes of less than 10 nm, and optical lithography can be used for feature sizes higher than 100 nm ^[38][74].

New form factors have surfaced more recently due, in large part, to the adaptability of EIS to different substrates. Rashed et al. modified a well-plate electrode with an RBD protein to detect anti-SARS-CoV-2 antibodies, and Ali et al. built a PDMS-molded microfluidic flow channel on top of 3D microprinted reduced graphene oxide structures ^[12][31][49,68]. Perhaps most interestingly, Xue et al. developed an impedimetric face-mask-integrated sensor for detecting the S protein in exhaled breath aerosols ^[32][69].

Perhaps the most critical step in fabricating an EIS sensor is functionalizing it with the appropriate capture element. Before doing so, a cleaning step is sometimes used to remove impurities from the surface of the electrode, which may hinder functionalization.

Au has traditionally been employed as an electrode material due to its inertness and excellent thermodynamic stability ^[39][75]. Various Au cleaning procedures, such as chemical and electrochemical cleaning, have been reported in the prior literature ^[40][76]. To eliminate environmental pollutants from the Au surface, an oxidizing substance such as sulfuric acid can be used to chemically clean the surface ^[41][77]. Electrochemical cleaning can also be performed, where the electrodes are placed in acid solution and voltammetric cycling is performed ^[42][78]. These cleaned electrodes are then ready for immobilization.

One of the most popular tools for immobilization is thiol-Au chemistry. This typically forms a monolayer onto the surface of the electrode which can uniformly orient the capture elements. Ashur et al., Soto and Orozco, Lasserre et al., and Abrego-Martinez et al. deposited thiolated antibodies, peptides, and aptamers directly onto Au surfaces ^{[11][13][18][24]}[48,50,55,61]. The thiol-Au chemistry was also used in conjunction with the popular N-ethyl-N’-(3-(dimethylamino)propyl)carbodiimide/N-hydroxysuccinimide (EDC/NHS) chemistry used for activating carboxylic acids to form amide bonds with primary amines. In these WEs, researchers first add a thiolated acid to an Au electrode, where the thiolated end interacts with Au to form an orderly monolayer which exposes an aligned carboxylic acid for EDC/NHS chemistry to take place ^{[10][12][25][27][30]}[9,49,62,64,67].

The EDC/NHS chemistry has also been used apart from acid monolayers. Perdomo et al. electrodeposited para-aminobenzoic acid on a carbon electrode to utilize EDC/NHS chemistry ^[22][59]. Soares et al., Pola et al., and Zaccariotto et al. used EDC/NHS chemistry on a carbon electrode with already-exposed carboxyl groups to immobilize capture elements ^[23][28][29][60,65,66].

There are many other methods for capture-element immobilization. Wu et al. and Xue et al. used the tried-and-tested biotin-streptavidin chemistry to functionalize their electrodes ^[27][32][64,69]. In addition, outside of specific linking chemistries, several groups also used nonspecific adhesion by simply drop casting the capture elements onto predominantly carbon electrodes (Rashed’s group drop casted on an Au electrode nonspecifically, and Ramanathan et al. drop casted onto diamond nanopowder) ^{[20][21][30][32][35]}[57,58,67,69,71]. Ehsan et al. immobilize a monolayer of 1-pyrenebutanoic acid succinimidyl ester via their Van der Waals forces, which results in the same exposed O-pyrene group as the end result of EDC/NHS coupling ^[15][52]. Avelino et al., Torres et al., and Li et al. immobilized their capture elements in glutaraldehyde matrixes ^[14][19][34][51,56,70]. Hussein et al. and Sharif et al. captured their analytes with a molecularly-imprinted polymer, thus not requiring specific linking chemistries ^[16][17][53,54].

After surface modification, the electrodes are typically blocked with a passivating molecule to prevent nonspecific adhesion or interaction with activated carboxyl groups. It was most common to use a low percent (0.005–1%) of bovine serum albumin (BSA), since it is large and neutrally charged at physiological pH ^{[14][15][18][20][21][22][27][28][34][35]}[51,52,55,57,58,59,64,65,70,71]. Alternative molecules that were used include ethanolamine, mercaptohexanol, 3% milk, and proprietary blocking solution ^{[11][19][23][24][30][31]}[48,56,60,61,67,68].

The design and size of the electrodes is closely related to the sample size that can be added to the electrode. Often, fully integrated sensors (where the WE, RE, and CE are all integrated onto one surface) will require between 1–10 μL of sample, with only three groups using samples sizes of 50 μL ^{[10][11][12][13][14][16][17][21][24][30][31][34][35]}[9,48,49,50,51,53,54,58,61,67,68,70,71]. Perdomo et al. were able to use only 0.3 μL of sample to operate their sensor (however, they used a more standard 50 μL of measurement buffer containing a redox marker) ^[22][59]. The interdigitated electrodes made by Soares et al., Ramanathan et al., and Rashed et al. required 250, 20, and 50 μL samples, respectively ^[29][30][31][66,67,68].

Further reduction in sample volumes makes testing easier on patients and makes frequent and widespread testing more feasible. In addition, this does seem possible; groups did not optimize sample volume, but simply reported the sample volume used. At some point, there are diminishing returns in reducing sample volume, however, so a minimal-but-reasonable amount of sample should be used (i.e., 10 μL is approximately a pinprick of blood—further reduction in sample volumes may not have practical advantage).

In order for these electrodes to provide actionable information, they need to be used in conjunction with a device that (1) generates an AC signal to perturb the sample and (2) reads the resulting impedance values. These signal readers can also contain electronics for data output to smart devices and/or data processing. Some groups used existing devices such as the PalmSens4 and Metrohm Autolab PGSTAT204, while others used traditional benchtop electrochemical workstations ^{[14][15][16][19][21][27][28]}[51,52,53,56,58,64,65]. Additionally, yet other groups developed their own electrical detectors, often with a focus on portability and use at the point of care ^{[12][14][18][22][23][24][32][34][35]}[49,51,55,59,60,61,69,70,71]. Most electrodes tended to fall within the range of a few centimeters, while devices tended to be slightly larger (a few centimeters more) ^{[12][13][14][17][18][22][23][24][25][32][34][35]}[49,50,51,54,55,59,60,61,62,69,70,71].

To perturb the system, each of the devices reviewed measured a range of frequencies to generate a semicircular Nyquist plot which could be fitted to a Randles equivalent circuit. The range of frequencies varied, but they typically began scanning around 10–100 kHz and completed their scan in the 0.01–0.001 Hz range ^{[10][11][12][13][14][15][16][17][18][19][20][21][22][23][24][25][26][27][28][29][30][34][35]}[9,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,70,71]. These measurements were typically performed in a solution containing ferricyanide and ferrocyanide as redox markers. Most groups applied a current with amplitudes between 5 mV and 10 mV (although some applied amplitudes in the range of hundreds of millivolts) and applied a DC bias of 0.1–0.2 V.

It is worth noting that more/fewer measurement points can be taken within the same range of frequencies. Salahandish et al. took both 7-point and 50-point measurements from 1–1400 Hz and found that the 7-point measurement was sufficient to generate a Nyquist plot to create an accurate calibration curve ^[35][71]. Fewer measurement points can simplify analysis and make measurements significantly faster. Some groups, such as Xue et al. and Rashed et al., use single-point impedance to generate signals from their sensors ^[31][32][68,69]. The disadvantage of measuring fewer frequency points is the potential to miss important information which could be revealed with a more detailed frequency sweep. Additionally, higher resolution frequency sweeps provide more data, which can be conducive to feeding ML algorithms to further increase diagnostic speed and accuracy.

There are two main classes of COVID-19 EIS sensor: (1) viral tests and (2) antibody tests. The choice of antigen is significantly impacted by which type of sensor is desired. Of the 24 reviewed papers, 19 were viral tests and only five sought to detect antibodies. Tests were created to detect three different portions of the SARS-CoV-2 viral particle: (i) the S protein (sometimes specifically the S1 subunit), (ii) the RBD of the S1 subunit, and (iii) the N protein.

The S protein was detected by six different groups with a range of LODs from the picogram to the femtogram-per-milliliter ranges ^{[10][22][23][29][32]}[9,59,60,66,69]. Popular sources of anti-S antibody were Sinobiological and Abcam.

Perdomo et al., Brazaca et al., and Soares et al. used anti-S protein antibodies bound to the electrode via EDC/NHS coupling ^[10][22][29][9,59,66]. Perdomo et al. and Brazaca et al. used three-carbon electrode systems to detect the S protein in PBS redox buffer ^[10][22][9,59]. Perdomo et al. achieved a limit of detection (LOD) of 1.065 fg/mL after finding linearity at concentrations between 1–20 fg/mL ^[22][59]. Brazaca et al. detected an LOD of 83.7 pg/mL when calibrated between 10 pM and 0.1 μM ^[10][9].

Soares et al. used a two-Au-electrode sensor to detect S protein samples in diluted viral solution without redox marker at up to and LOD of 0.179 fg/mL (testing a range of 10 zg/mL–10 ng/mL) ^[29][66]. It is notable they use capacitance as a detection parameter instead of the more common Rct ^[29][66].

Xue et al. used an anti-S protein antibody immobilized via biotin-streptavidin onto PEDOT:PSS nanowires to detect the S protein in a mask-integrated sensor. They obtained an LOD of 7 pfu/mL (0.35 pfu/L of air) in an aerosolized solution (mimicking breathed-out saliva) ^[32][69]. Pola et al.’s attempt at S1 protein detection resulted in an aerosol jet-printed carbon electrode functionalized with polyclonal anti-S protein antibodies which had a 110.38 ± 9.00 pg/mL LOD when exposed to a range of 1–1000 ng/mL S1 protein ^[23][60]. r Although Brazaca et al. and Pola et al. report an LOD four orders of magnitude higher than Perdomo et al. and Soares et al., this does not necessarily indicate an inferior process. This is because LODs are determined by a calculation based on either (i) the perceived intercept between the linear response and nonlinear response or (ii) the standard deviation and slope of a regression line. Both methods have objectivity issues. A bi-linear response (i.e., Ramanthan et al.,) could be misinterpreted as nonlinearity, therefore preventing full characterization of the electrode response. Variability in a manufacturing technique could, similarly, have a negative effect on LOD calculations; Soares et al.’s 10-fold lower LOD compared to Perdomo et al.’s LOD is likely due to the use of precision photolithography for fabrication and treatment, over the latter’s use of basic screen-printing technology ^[22][29][59,66].

Torres et al. and Tepeli et al. used biologically inspired design to detect the S protein using ACE2 and CD147 transmembrane glycoproteins as capture elements, since they are two receptors in the body that bind the SARS-CoV-2 virus ^[25][34][62,70]. Torres et al. were able to use ACE2 receptors immobilized with glutaraldehyde to detect S protein at a concentration of 1.39 pg/mL after testing a range of 100 fg/mL–100 ng/mL in human saliva ^[34][70]. Tepeli et al. immobilized both ACE2 and CD147 onto sensors and found LODs of 299.30 and 38.99 ng/mL (respectively) after testing in ranges of 700–7000 and 500–5000 ng/mL ^[25][62]. It is notable that Torres et al.’s carbon screen-printed system demonstrated a 100-fold superior LOD compared to Torres et al.’s commercial Au electrode system, although this, again, may be due to the latter’s testing range being significantly higher. Lasserre et al. used anti-S1 protein aptamers to detect the S1 protein, but did not demonstrate data, only correlating the results with a negative/positive result ^[18][55].

These biologically inspired sensors demonstrate performances comparable to the more traditional antibody-based sensors. The major advantage here is that existing biological structures such as ACE2 and CD147 are typically more familiar (protein structure, manufacturing, chemical stock, etc.) early on in an endemic compared to antibodies against the novel infectious agent (i.e., anti-S protein antibodies took time to produce at scale for research). Therefore, biologically inspired molecules can enable a more rapid response to a new disease than waiting for reliable antibodies to be produced. However, specific-antibody-based sensors are still important, since biologically inspired capture elements are likely to include nonspecific interactions native to healthy individuals.

Five groups detected the RBD protein and achieved LODs at the pico- and femtogram levels ^{[11][15][23][24][28]}[48,52,60,61,65]. Of the five groups, two groups used peptide sequences and DNA aptamers as capture probes ^[11][24][48,61]. The three remaining groups used anti-S1 antibodies to capture the RBD protein, since the RBD is the primary binding site of the S protein to cell receptors such as ACE2 ^{[15][23][24][28]}[52,60,61,65]. A variety of different suppliers sourced the different capture elements, including Genscript.

Soto and Orozco used a thiol-immobilized 23-amino acid peptide sequence which mimics the ACE2 receptor to capture the RBD protein with an LOD of 0.01 copies/mL in redox buffer with potassium nitrite ^[24][61]. They tested the modified commercial Au electrode in a linear range of 100–1000 copies/mL ^[24][61]. Abrego-Martinez et al. also used a thiolated aptamer to detect up to 1.30 pM (66 pg/mL) RBD with a linear range of 10 pM–25 nM in PBS redox solution ^[11][48].

Zaccariotto et al. detected RBD with an anti-S1 antibody immobilized by EDC/NHS onto a glassy carbon disk electrode ^[28][65]. They tested in the range of 0.16–40 μ

g/mL with an LOD of 150 ng/mL in a PBS redox buffer ^[28][65]. Ehsan et al. and Pola et al. printed (by hand and by CNC) graphene electrodes and modified them with anti-S1 antibodies which demonstrated 0.25 fg/mL and 22.91 ± 4.72 pg/mL LODs for RBD when tested in ranges of 0.25 fg/mL–1 ng/mL and 1–1000 ng/mL ^[15][23][52,60]. Pola et al. tested in PBS redox buffer, while Ehsan et al. tested in redox solution (without PBS) ^[15][23][52,60].

The performance of the RBD detectors is comparable to the performance of the S protein detectors. There was no significant difference in terms of selectivity between the different sensors (many groups did not report a selectivity difference, and if they did, the parameters used varied).

The N protein is not involved with the SARS-CoV-2 virus entering cells, but it is critical for packaging the RNA genome inside the viral capsid ^[43][79]. The N protein is enclosed in the viral capsid and is not accessible from outside an intact viral particle. However, the N protein is abundantly expressed during infections, and is easily detected in infected cells. Four different EIS sensors measured N-protein levels spiked into artificial solution ^{[14][27][30][35]}[51,64,67,71]. Wu et al. went further, validating their tests in artificial saliva, and Avelino et al. validated their test against qRT-PCR with real nasophrayngeal/oropharyngeal swab samples ^[14][27][51,64]. Similar to the RBD and S protein sensors, these sensors reached LODs in the range of femtograms.

Salahandish et al. and Wu et al. used anti-N-protein antibodies from Genscript and Vazyme to detect N proteins in PBS redox buffer ^[27][35][64,71]. Salahandish et al. detected up to 116 fg/mL when calibrating from 1–10,000 pg/mL with their carbon + graphene@PEDOT:PSS electrodes ^[35][71]. Wu et al. achieved an LOD of 6 pg/mL and a range of 0.1–100 ng/mL with an AuNP-modified commercial carbon electrode from Zensor ^[27][64].

Instead of antibodies, Ramanathan et al. used an aptamer immobilized onto an Au interdigitated electrode via a silanization reaction to obtain an LOD of 0.389 fM (approximately 0.443 pg/mL) with a linear range of 1 fM–100 pM ^[30][67]. Avelino et al. took a different route, using an amino-modified primer to detect the nucleocapsid gene instead of the protein, achieving an LOD of 258.01 copies/μL with a linear range of 800–4000 copies/μL ^[14][51].

As with the RBD sensors, the N-protein aptamer sensors show comparable performance to the traditional antibody-based N-protein sensors, and the difference in LOD is difficult to compare, since the tested ranges and variability are so different.

Three groups opted to detect entire viral particles instead of only proteins expressed on the surface of those particles. Ashur et al. used Traut’s reagent to thiolate an anti-S-protein antibody for immobilization on a three-electrode polytetrafluoroethylene-based Au electrode ^[13][50]. They did not detect the SARS-CoV-2 virus, but developed a pseudovirus which expressed the S protein on its surface ^[13][50]. They were able to detect a range of 104 to 109 viral particles/mL, and when detecting S protein in solution, achieved an LOD of 15 ng/mL (500 pM) ^[13][50].

Hussein et al. did not use a traditional capture element for their detector; they instead casted a layer of 3-aminophenol monomer mixed with a human sample of the SARS-CoV-2 virus, then washed out the virus to leave a viral imprint ^[17][54]. They were able to achieve an LOD of 57 pg/mL; while no range was reported, they tested the sensor at up to 320 pg/mL in redox buffer ^[17][54]. El Sharif et al. did the same with a N-hydroxymethyl acrylamide monomer with an LOD of 0.69 pfu/mL when tested in 0.477–0.845 pfu/mL; they were one of the few groups who tested in real samples with saliva biofluid ^[16][53].

This molecular imprinting technique has been used to develop NPs selective for the COVID-19 virus ^[44][80]. Thomaz et al. investigated the performance of recombinant antibodies compared to molecularly imprinted NPs for impedimetric detection of the SARS-CoV-2 RBD protein and found that they can have a binding performance comparable to high-affinity anti-RBD recombinant antibodies ^[44][80]. Such molecularly imprinted NPs have the advantage in stability and production capacity, two issues which have not yet been resolved with conventional or recombinant antibodies.

Sensors developed with a whole-virus detection mindset show comparable LODs to protein-detecting EIS sensors ^[13][16][17][50,53,54]. Additionally, testing for the detection of whole viral particles more closely mimics the physiological condition, which is beneficial for the practical translation of the technology.

The presence of antibodies is an indicator of immunity against an infection, likely caused by a previous or current infection. Antibody test currently requires blood draws, and therefore is a target for translation into low cost, less invasive EIS sensors. Five groups attempted to detect antibodies using the corresponding antigen, obtained from a variety of sources, including laboratory-synthesized antigens ^{[12][19][20][21][31]}[49,56,57,58,68].

Ali et al. detected anti-S-protein antibodies and anti-RBD antibodies by immobilizing S proteins and RBD onto 3D-microprinted Au structures coated with reduced graphene oxide via EDC/NHS chemistry ^[12][49]. They obtained an LOD of 2.8 fM and 16.9 fM for anti-S protein and anti-RBD antibodies in ranges of 1 fM–30 nM and 1 fM–20 nM in PBS buffer ^[12][49]. Li et al. immobilized RBD onto zinc oxide nanowires grown onto carbon ink for an LOD of 0.4 pg/mL when tested between 1 ng/mL–1 μg/mL in PBS redox buffer ^[19][56].

Nicoliche et al., similarly, used the S protein dropcasted onto a pyrolyzed graphitic paper electrode to detect antibodies against the S protein in PBS redox buffer ^[21][58]. Rashed et al. used single-point impedance for their sensor ^[31][68]. They used RBD nonspecifically attached to Au interdigitated electrodes embedded into the bottom of a well plate to detect a range of anti-S-protein antibodies in 3% milk buffer ^[31][68]. However, neither group reported an LOD or a detection range. Lorenzen et al. immobilized truncated N proteins onto AuNP-modified PEDOT:PSS, but tested saliva dilutions instead of precise concentration values ^[20][57].

The field seems to prefer rapid viral tests over rapid antibody tests. For one, there are so few tests looking for antibodies, and three of the tests do not report significant quantified data. However, as previously mentioned, there is a largely unexplored field of rapid, low-cost EIS devices for antibody tests that would prove extremely helpful for determining immunity and exposure to viruses such as SARS-CoV-2.