The ability to design, construct and run assays that are specific, sensitive and robust is crucial in all areas of biomedical research. By way of example, in basic research applications one may need to quantify cellular levels of a particular protein, determine the levels of a metabolite in serum or urine or, perhaps, compare the catalytic activity of an enzyme in normal and diseased tissue [1]. In the pharmaceutical and biotechnology industries assays are required for the identification and characterisation potential new drug molecules [2, 3]. Other studies may require the quantitative measurement of gene expression levels [4] or the levels of salivary IgA and IgG antibodies against SARS-CoV-2 spike protein [5].

Regardless of the specific application or the particular molecule to be measured, a number of fundamental factors must be considered when developing a biological assay. It is important that careful attention is paid not just to the assay itself, but also to the whole workflow that needs to be developed from sample preparation to analysis of the data quality delivered by the assay. The key aspects of assay development discussed in this article are of general applicability. They are valid regardless of the type of assay to be developed (e.g., antibody-based assays such as ELISA and electrochemiluminescence, determination of total protein/protein concentration, enzyme activity assays, cell-based assays or quantitative PCR, or specialized assay such as single-molecule activation assay [6] ). Similarly, the key aspects are valid regardless of the detection system employed, be it based on absorbance, fluorescence, radioactivity or even mass spectrometry. Each specific assay will have its own idiosyncrasies, but by paying careful attention to the points discussed below the researcher can help ensure that they develop a robust, reliable, fit-for-purpose assay. Often, guidelines and/or extensive discussions about assay development using specific technologies exist, such as those for targeted peptide measurements mass spectrometry [7]. Although [7] focuses on the particular issues of targeted MS assays, a number of the discussion points are relevant to any assay development project. In particular, the authors stress the need to take a fit-for-purpose approach to development and validation of assays [7]. Complex cell-based assays that aim to mimic the in vivo cellular environment are playing an increasing role in drug discovery and development. Whilst, the specific details of such assays are beyond the scope of this article, it is important to emphasize that the basic principles of assay development must be adhered to in order to build robust and reliable assays that generate meaningful data [8].

Commercial assay development kits are available to ease the assay establishment. For example, Stadtmauer EA et al utilized the DuoSet Ancillary Reagent Kit from R&D Systems to establish an ELISA assay detecting residual Cas9 protein in CRISPR-engineered T cells [9]. Commercial organizations have developed a wide array of assays, and should first be explored before undertaking the effort to develop a new assay. The sample assays include the MILLIPLEX MAP mouse angiogenesis magnetic bead-based assay from MilliporeSigma [10].

When embarking on an assay development project the researcher should ask a number of questions in order to define the exact requirements of the assay.

The starting point is to be absolutely clear as to what molecule and precisely what property of that molecule is to be measured. This may sound obvious, but it is a question of fundamental importance and underpins all subsequent assay development activities. For example, does the researcher wish to measure the total amount of a particular protein in a cell lysate or only the phosphorylated form, or both total and phosphorylated protein [11, 12] ? Similarly, the study may dictate that specific isoforms or splice variants of the protein of interest need to be measured [13, 14]. In the case of a protein, one needs to be clear as to whether the key parameter to measure is the amount of the protein present or its biological function, such as enzymatic activity or the effect of a cytokine on potential cellular targets. Alternatively, it may be acceptable or desirable to measure the level of gene expression at the mRNA level. It may be that the study dictates that separate assays be developed to measure gene expression, steady-state protein levels and biological function, respectively [12, 15, 16] ; each assay will generate different but complementary information about the molecule of interest.

It is important to consider the source of the molecule that is to be assayed. Is the molecule to be measured in a bodily fluid, such as serum or urine? Is the molecule to be measured in an organ obtained from an experimental animal, or a biopsy sample from a patient? Perhaps the source material will be post-mortem tissue. It may be that the source will be cells cultured in vitro, in which case an important consideration is whether these will be scarce primary cells or a readily scaleable immortalised cell line.

The source of the molecule will determine the sample quantity and availability. It will also determine the concentration of the molecule of interest and may profoundly influence its stability, as discussed in the following sections. Thus the source of the molecule of interest is likely to have a significant influence on the overall assay workflow that is ultimately developed.

It is also important to understand the stability of the molecule for which the assay is to be developed. Is it relatively stable, or is it extremely unstable requiring special precautions to be taken during collection and preparation of the sample for assay in order to obtain meaningful assay data? Even molecules that are stable in the isolated form may be unstable in the complex biological milieu of the samples to be assayed where they may be subject to oxidation, proteolysis or the loss of post-translational modifications. It may, for example, be necessary to include reducing agents or protease and/or phosphatase inhibitors in sample buffers to maintain molecules in their physiologically relevant forms both during sample preparation and over the course of the assay itself. The source of the molecule is an important consideration here; a molecule that is stable in, say, serum may prove extremely unstable in a liver homogenate. If assaying biopsy or post-mortem tissue it is important to know how the tissue has been handled and stored as this may profoundly affect both the quantity and quality of certain molecules.

In order to develop an assay that is fit-for-purpose it is important to decide at the outset whether a semi-quantitative measurement of the molecule, such as Western blot, meets the requirements of the project or whether a rigorously quantitative assay is required.

It is also important to consider how many samples will need to be assayed. If only a handful of samples will be assayed then a labour-intensive, multi-step manual assay format may be perfectly acceptable. Conversely, if thousands or tens-of-thousands of assays are to be run, perhaps as part of a compound profiling exercise, it will be important to simplify, streamline and automate the assay process as much as lab resources allow, with, for example, a format like microarray. However, arrays bring their own set of problems, such as intra-assay spatial variation [17].

The answers to all of these questions concerning the molecule to be assayed will play a key role in helping the researcher decide on the most appropriate format for their particular needs, and should be borne in mind when considering the assay fundamentals discussed in the next section.

Having considered these questions regarding the molecule to be assayed, careful attention must be paid to a number of fundamental technical/practical issues that apply regardless of the particular molecule of interest or the specific assay format adopted.

An absolutely key consideration is the specificity of the assay. Having decided exactly what molecule and parameter of that molecule is to be measured, the researcher needs to establish that the assay will measure only what they want it to measure and not anything else. If it transpires that the assay is measuring both the desired molecule and other molecules, then it is possible that steps can be taken to improve the specificity of the assay.

Two overlapping aspects of assay specificity need to be considered.

• Firstly, are the primary detection reagents suitably specific? For example, in the case of the HPLC-based assay, how confident is the researcher that a peak identified at a particular retention time is the molecule of interest and not a different molecule that just happens to co-elute with the molecule of interest. Here the source of the analyte should be considered. Analysis of urine samples may not give rise to peaks co-eluting with the molecule of interest. However, the same may not be true for serum or tissue homogenate samples analysed in the same assay. In an antibody-based assay, is the antibody employed absolutely specific for the target protein or does it also cross-react with other proteins? It is possible to obtain antibodies that are specific for a particular protein or a particular post-translational modification, such as phosphorylation at a specific site or to a neo-epitope generated upon cleavage by a proteinase [18, 19]. Quite often, an antibody with higher affinity is desired [20]. Such antibody reagents must be characterised rigorously under the actual assay conditions to ensure that they have the desired specificity, reacting only with the desired product and not with the substrate/precursor molecule. In assays that do not use antibody detection, such as a fluorogenic peptide cleavage assay, the degree of specificity will be much lower. Instead of cleavage at a specific site generating a signal, cleavage at any site within the peptide will now give rise to a signal.
• This brings us to the second aspect of assay specificity and is a particular issue for in vitro enzyme assays. Even if the detection reagents are specific for a particular event one wishes to assay, it may well be that in a crude cell lysate or tissue homogenate that multiple enzymes are capable of effecting that same event (e.g., phosphorylation, proteolytic cleavage or other post-translational modification). This problem will be even greater for the fluorogenic peptide cleavage assay example given above. The problem of multiple overlapping activities in crude lysates can often be overcome by the inclusion of an additional sample preparation step, such as a specific immuno-precipitation, in order to achieve the desired assay specificity [21]. Alternatively, depending on the nature of the study, the best solution may be to run the activity assay using highly purified (native or recombinant) enzyme.

How sensitive does the assay need to be? This will be determined both by the levels of the molecule of interest in the sample and the volume of sample available. The important point is that the assay must be sufficiently sensitive such that the level of the molecule falls well within the dynamic range of the assay (see below). This is an important consideration when deciding upon the assay format and detection system. If great sensitivity is required then a fluorescence, rather than absorbance, the readout is more likely to give the desired sensitivity. Sensitivity may also be increased by utilising enzymatic amplification of the original signal [22, 23]. Such ‘coupled’ assay systems can be extremely sensitive, but also provide increased opportunity for the assay sample to interfere with the assay (see below).

It is important to determine the dynamic range of the assay. In other words the range over which the assay readout is proportional to the amount of target molecule in the sample being analysed. In the case of an assay to measure the concentration of a molecule, it is important that the assay (regardless of the format adopted) is appropriately calibrated by the construction of a suitable calibration curve such as the hypothetical example in Figure 1. Similarly, in the case of an enzyme assay, one must ensure that the assay is operating in a range such that the initial rate of the reaction is measured and that the measured rate is proportional to the amount of enzyme added to the assay. In order to stay within the dynamic range of the assay (some) samples may need to be diluted or concentrated. This may be a particular issue if the molecule of interest is present at vastly different levels in different samples. Failure to stay within the dynamic range of the assay will result in the generation of spurious data. The dynamic ranges must also be tested in vivo, if the assay is to be used in vivo [24], for example, using cell lines that express an antibody target at different levels for an IHC assay development [25].

Figure 1. Assay calibration/dynamic range.

When designing an assay, it is important to consider factors that may interfere with the assay readout leading to erroneous results. The question of assay specificity is discussed above. Here we discuss other factors that may interfere with the assay readout.

In the case of a fluorescent readout it is important to ensure that the sample to be assayed does not quench the fluorescent readout. This may be a particular issue with assays utilising crude cell lysates or tissue homogenates. Similarly, in fluorogenic enzymatic or cellular assays used for compound screening/profiling purposes, the researcher needs to ensure, as far as it is possible to do so, that the test compounds do not quench the fluorescent readout. Such compounds would display apparent potent inhibitory activity whilst actually having no direct inhibitory effect on their intended target. Some compounds may themselves fluoresce, thus giving a false assay signal. Such compound interference can be reduced by using fluorescent reporters that excite and emit at longer wavelengths [26]. Another common problem is the interference by sample components, such as reducing agents or detergents with some of the popular protein assays.

In enzymatically-coupled assays (be they assays of enzyme activity or of metabolites) it is essential to determine that the coupling enzymes do not inadvertently measure some component of the sample other than the molecule of interest. If such assays are used for compound screening or characterisation it is important to ensure that the assay is configured such that it measures inhibition of the target enzyme and not of the coupling enzyme(s) [27]. If significant interference from components of the assay sample is suspected, careful consideration must be given to steps such as (immuno) precipitations or partial purification in order to remove such components [28]. The key point is, no matter what assay format is adopted, to be alert to potential sources of interference and ways of eliminating or minimising their impact on the assay. One must not assume that an assay developed for measuring a molecule in one particular tissue/cell type/biological fluid will perform acceptably when the assay is performed on samples from a different source; the assay will need to be re-validated for the new sample source.

In order to give reliable, usable data the assay must be robust and reproducible. The assay should be robust in that it is not unduly affected by changes in sample preparation and handling. Equally, it should give the same results regardless of the individual operating the assay. The assay should be highly reproducible (also referred to as precision) such that the degree of variation is as small as possible both on an intra and inter assay basis. On a single assay run, replicates of both a standard and a ‘real’ sample should give very similar values, respectively. Similarly, there should be little day-to-day variation in the assay values obtained.

If the assay is to measure the absolute (rather than relative) amount of a molecule then it must be calibrated (see above) against an accepted standard in order to give accurate quantitation. Sometimes the assay workflow requires significant sample workup (e.g., concentration, removal of interfering sample components etc..) in order that the molecule of interest can be measured reliably. In such cases it is highly desirable to include an internal standard to correct for analyte losses across the assay workflow. For example, in the case of MS assays, the extract to be assayed may be ‘spiked’ with a known quantity of stable isotope of the molecule of interest. For an HPLC assay the internal standard could be a compound structurally related to the molecule of interest and which is not otherwise present in the assay sample. The use of an internal standard is particularly important if very precise and accurate quantitative data are required [7].

The levels of robustness, reproducibility and accuracy that are acceptable will depend on the purpose of the assay and must be decided upon by the researcher for their own particular current use.

Assuming a Gaussian distribution, two results, the mean of which differ by > 2 standard deviations (SD) are considered to be statistically significantly different at the 95% confidence level. Two results similarly differing by >3 SD are considered statistically significantly different at the 99.7% confidence level [29]. Thus, the greater the reproducibility (precision) of the assay, and hence the lower the SD of experimental readings, the more discriminating the assay will be.

Figure 2. Measurement of assay performance.

Assay performance can be analysed in a number of different ways. The signal-to-background ratio or signal window (S/B = mean signal / mean background) and the signal-to-noise ratio (S/N = mean signal – mean background / SD of background) are often used as measures of assay performance. However, these ratios do not adequately take into consideration assay variability. A simple statistical test, the Z’ value, has been developed that provides a measure of assay quality that takes account of both the signal window and assay variability [30]. The higher the Z’ value (Figure 2), the more discriminating the assay, with a perfect assay having a Z’ of 1. For a high throughput screening assay a Z’ value of > 0.5 is generally considered acceptable. Although the Z’ statistic was originally derived for evaluating HTS assays, it is a general measure of assay quality and can be applied to any assay [29, 30].

High-throughput assays used in drug discovery frequently use fluorescent and luminescent reporters to enable large numbers of assays to be performed in an automated fashion. However, with the development of new technologies, there is increasing interest in the use of label-free cellular assays that utilize biophysical detection methods [31-33]. A number of instruments for label-free assays are now commercially available [31, 32]. Importantly, label-free assays avoid the potential for artefacts caused by the presence of the reporter system. For example, the commonly used bioluminescent reporter molecule firefly luciferin is itself a partial agonist of GPR35 [34], thus potentially complicating compound/pharmacological data interpretation. An additional benefit of label-free assays is that they allow the screening of compounds against endogenous levels of receptors in disease-relevant primary and stem cells [33].

Unexpected reporter-compound interactions may also occur in biochemical assays (in addition to the quenching and autofluorescence issues discussed above). For example, a number of compounds shown to activate the protein deacetylase activity of SIRT1 using a TAMRA-labelled peptide substrate had no activity in enzyme assays using an unlabelled version of the same peptide or a native full-length protein substrate. The activation observed with the TAMRA-labelled substrate was an artefact due to a direct physical interaction of the compounds with the TAMRA fluorophore [35].

Thus, when designing and validating assays for screening/compound profiling purposes, it is important to consider the potential for compound-reporter interactions. The potential benefits of using a label-free assay at some point in the screening cascade should be given serious consideration.