IMPROVEMENTS IN OR RELATING TO AN APPARATUS FOR CHARACTERISING A COMPONENT

Abstract

An apparatus for characterising a biomolecule is provided. The apparatus comprising a sample inlet channel configured to introduce a sample fluid including the biomolecule to the apparatus; an auxiliary inlet channel configured to introduce an auxiliary fluid to the apparatus; a distribution channel in fluid communication with the sample inlet channel and the auxiliary inlet channel; wherein the distribution channel is adapted to generate a distribution of biomolecules; a measurement module configured to detect a signature profile of the biomolecule to obtain a measured dataset of the detected biomolecule; a storage location configured to store and maintain a stored dataset comprising a plurality of parameters that are associated with the measured dataset obtained from the measurement module; and an analysis module configured to receive the stored dataset from the storage location and correlate the stored dataset with the measured dataset from the measurement module to provide a correlation value, wherein the analysis module is further configured to use the correlation value to determine at least two characteristics of the biomolecule simultaneously using Bayesian analysis. A method for characterising a biomolecule is also provided.

Claims

1. An apparatus for characterising a biomolecule, the apparatus comprising: a sample inlet channel configured to introduce a sample fluid including the biomolecule to the apparatus; an auxiliary inlet channel configured to introduce an auxiliary fluid to the apparatus; a distribution channel in fluid communication with the sample inlet channel and the auxiliary inlet channel; wherein the distribution channel is adapted to generate a distribution of biomolecules; a measurement module configured to detect a signature profile of the biomolecule to obtain a measured dataset of the detected biomolecule; a storage location configured to store and maintain a stored dataset comprising a plurality of parameters that are associated with the measured dataset obtained from the measurement module; and an analysis module configured to receive the stored dataset from the storage location and correlate the stored dataset with the measured dataset from the measurement module to provide a correlation value, wherein the analysis module is further configured to use the correlation value to determine at least two characteristics of the biomolecule simultaneously using Bayesian analysis.

2. The apparatus according claim 1, wherein the apparatus further comprises a controller configured to receive the stored dataset from the storage location and to receive the measured dataset from the analysis module to further tune one or more parameters associated with the measured dataset; and an output module associated with the controller configured to provide a notification to an operator indicating a further cycle of measurements such that further measurements obtained provide a pre-determined level of confidence in the determined characteristics of the biomolecule.

3. The apparatus according to claim 1, wherein the distribution channel is adapted to generate a lateral distribution of biomolecule.

4. The apparatus according to claim 1, wherein the apparatus further comprises at least two outlet channels to divide the fluid in the distribution channel into two or more flows.

5. The apparatus according to claim 1, wherein the analysis module is configured to detect and determine the concentration and the affinity of each biomolecule.

6. The apparatus according to claim 1, wherein the analysis module is configured to detect and determine the concentration and the avidity of each biomolecule.

7. The apparatus according to claim 1, wherein the analysis module is further configured to detect and determine the stoichiometry of each biomolecule.

8. The apparatus according to claim 1, wherein the distribution channel is a T-sensor.

9. The apparatus according to claim 1, wherein the biomolecule is labelled with a fluorophore, a quantum dot or a nanoparticle.

10. The apparatus according to claim 1, wherein the fluorophore selected for labelling the biomolecule is in the far-red spectral region.

11. The apparatus according to claim 1, wherein an electrode is provided at the upstream of the distribution channel and an electrode provided at the downstream of the distribution channel, configured to apply an electric field across the distribution channel.

12. The apparatus according to claim 1, wherein the biomolecule is an antibody, a polypeptide, a polynucleotide or a polysaccharide.

13. The apparatus according to claim 1, wherein the biomolecule is an antibody or an antibody fragment thereof.

14. The apparatus according to claim 1, wherein the antibody is an allo-antibody.

15. The apparatus according to claim 1, wherein the biomolecule is a multi-biomolecule mixture.

16. The apparatus according to claim 1, wherein the multi-biomolecule mixture comprises an antibody and an antigen.

17. The apparatus according to claim 16, wherein the antigen is labelled.

18. The apparatus according to claim 1, wherein the sample fluid is a human serum comprising the biomolecule.

19.-27. (canceled)

Description

[0083] FIG. 1 shows an apparatus for characterising a biomolecule according to the present invention;

[0084] FIGS. 2A to 2C provide plots showing binding experiments of HLA;

[0085] FIGS. 3A to 3D show binding curves for HLA against an antibody in human serum;

[0086] FIGS. 4A to 4D show data on testing of human patient serum against different HLA variants;

[0087] FIGS. 5 and 5B provide a labelling strategy of the sample according to FIG. 1;

[0088] FIG. 6A and FIG. 6B provide a further labelling strategy of the sample according to FIG. 1;

[0089] FIG. 7 shows a binding curve for interaction;

[0090] FIGS. 8A to 8E show data of covalent labelling according to the present invention;

[0091] FIGS. 9A to 9F show the determination of the absorbance of varying concentration of a label in human serum;

[0092] FIG. 10 shows the structure of bilirubin;

[0093] FIGS. 11A and 11B show plots of negative control experiments;

[0094] FIG. 12 shows a further plot of a negative control experiment;

[0095] FIGS. 13A and 13B show global fit for the binding curves for interactions between HLA and an antibody;

[0096] FIGS. 14A and 14B shows the binding curve of human serum against HLA; and

[0097] FIG. 15 showing an affinity plot of serum antibody reactivity against the Receptor Binding Domain (RBD) of SARS-CoV-2 Spike protein.

[0098] According to the present invention, there is provided an apparatus for characterising a biomolecule and in some cases the apparatus can be used to characterise biomolecule interactions such as antibody-antigen interactions. The apparatus comprises a sample inlet channel configured to introduce a sample fluid including the biomolecule to the apparatus, an auxiliary inlet channel configured to introduce an auxiliary fluid to the apparatus and a distribution channel which is in fluid communication with the sample inlet channel and the auxiliary inlet channel. The distribution channel is adapted to generate a distribution of biomolecules through diffusion such as lateral diffusion and/or through electrophoresis upon the application of an electric field or through thermophoresis upon the application of heat. Additionally or alternatively, the distribution of biomolecules within the distribution channel can be generated via microfluidic diffusional sizing (MDS).

[0099] In utilising the MDS assay for absolute quantification and characterisation of molecules in solution, it is possible to determine the hydrodynamic radius, Rh, in human serum and hence, it is possible to characterise the interactions of biomolecules in measured in human serum. This is because MDS has been show to allow the determination of binding parameters by utilising the measured hydrodynamic radius, Rh, of a fluorescently labelled protein by tracking its spatial and temporal evolution in a microfluidic channel under laminar flow.

[0100] The apparatus further comprises a measurement module configured to detect a signature profile of the biomolecule to obtain a measured dataset of the detected biomolecule; a storage location configured to store and maintain a stored dataset comprising a plurality of parameters that are associated with the measured dataset obtained from the measurement module; and an analysis module configured to receive the stored dataset from the storage location and correlate the stored dataset with the measured dataset from the measurement module to provide a correlation value. The analysis module is further configured to use the correlation value to determine at least two characteristics of the biomolecule simultaneously using Bayesian analysis.

[0101] Referring to FIG. 1, there is provided a schematic showing the apparatus and method of the present invention. Human serum sample is taken from a patient such as a human patient that is then incubated with labelled HLA. The sample is then introduced into a device 100. The device 100 comprises a sample inlet channel 102 where the human serum sample is loaded onto the device and it also comprises an auxiliary inlet channel 104 for introducing a blank fluid flow. The sample fluid flow and the auxiliary fluid flows are introduced into a diffusional channel 110 to generate a laminar flow within the distribution channel 110. The component or components, which may be biomolecules, are distributed between the sample and auxiliary flows within the distribution channel 110. The distribution of the biomolecule occurs via on or more of diffusion, electrophoresis or thermophoresis.

[0102] The device 100 further comprises at least two outlet channels 120, 122 located downstream from the distribution channel 110 where the fluid flow comprising the sample and auxiliary flows in the distribution channel 110 is divided the fluid into two outlet flows, one flow entering each outlet channel 120, 122. As shown in FIG. 1, the distribution channel 110, together with the sample inlet channel 102, auxiliary inlet channel 104 and the outlet channels 120, 122 forms an H-filter.

[0103] A measurement module comprising a detector may be positioned at the outlet channels to detect the biomolecule of interest. The fluorescence at each of the outlet channels 120, 122 can be measured. From the ratio between the fluorescence in both the outlet channels, the hydrodynamic radius, Rh, of the protein can be determined. In some cases, the measurement module is able to detect and/or identify a signature profile of the biomolecule of interest to obtain a measured dataset of the detected biomolecule. The measured dataset can then be transmitted to an analysis module.

[0104] The size of the complex is subsequently determined by microfluidic diffusional sizing of which the dissociation constant K.sub.d and the antibody concentration are evaluated using Bayesian analysis.

[0105] MDS has been shown to allow the determination of binding parameters by utilising the measured hydrodynamic radius, Rh, of a fluorescently labelled protein by tracking its spatial and temporal evolution in a microfluidic channel under laminar flow as illustrated in FIG. 1.

[0106] Referring to FIGS. 2A to C, there is shown in FIG. 2A binding experiment of HLA A*03:01 using diffusional sizing. Every bar shows the average of triplicate measurements. A significant change in hydrodynamic radius indicates a positive binding event between HLA A*03:01 and W6/32. In contrast, no binding can be observed between HLA and OUW4F11 or BSA. Referring to FIG. 2B, there is shown a correlation of hydrodynamic radii with the number of residues. The radii determined for HLA both free and bound to the antibody W6/32 agree well with the assumption of a folded protein.

[0107] Referring to FIG. 2C, there is shown a binding curve of 25 nM HLA A*03:01 with varying concentrations of antibody W6/32. The points give the size measured for the labelled HLA species in varying fractions of unbound and unbound state, averaged over the data of at least three replicates, the curve is a least-square fit to a non-linear binding equation. From the fit, the dissociation constant K.sub.d=961.3±115.7 pM and the binding site concentration B.sub.1=12.98±1.00 nM could be determined, resulting in a stoichiometry of one antibody to two HLA molecules.

[0108] Referring to FIG. 3A, there is shown a binding curve of 5 nM HLA A*02:01 against antibody SN23OG6 in human serum (red) and PBS (blue). From the fit, the K.sub.d=3.83±1.26 nM in serum and K.sub.d=3.64±0.43 nM in PBS was determined, which perfectly agree with each other. Further, a binding ratio 1 to 2 for binding of 2 antigens per antibody is obtained. The inserted graph shows the increase of the Stokes radius (Rh) for HLA A*02:01 upon addition of the specific antibody SN23OG6.

[0109] Referring to FIG. 3B, there is shown a binding curve of 1.2 nM HLA B*08:01 against antibody OUW4F11 in human serum (red) and PBS (blue). From the fit, K.sub.d=50.74±8.09 nM in serum and K.sub.d=57.38±7.19 nM in PBS and a binding ratio of n=½ was determined, showing consistency. The inserted graph shows the increase of the Stokes radius for HLA B*08:01 upon addition of the specific antibody OUW4F11.

[0110] Referring to FIG. 3C, there is illustrated a summary of the hydrodynamic radii for the unbound antigens and in FIG. 3D, there is shown the dissociation constants in human serum in comparison to their values in pure buffer, demonstrating consistent values under both buffer conditions.

[0111] In order to demonstrate the usability of the MDS assay for absolute quantification and characterisation of molecules in complex media, well characterised interactions were measured in human serum. First, varying concentration of antibodies SN23OG6 and OUW4F11 are added into blank human serum and measured their interactions with HLA A*02:01 and HLA B*08:01, respectively. As a first step, the hydrodynamic radii of both antigens are determined. The radii of pure HLA obtained human serum are consistent with the theoretical values as well as the radii obtained in buffer as shown in FIG. 3C, suggesting applicability of the MDS assay for human serum.

[0112] This also shows that human serum does not contain any other protein that can perturb the measurements by binding to the investigated antigen. For these two antigen/antibody pairs, the dissociation constants determined are independent of the buffer conditions used. More specifically, for the interaction of the antibody SN23OG6 against HLA A*02:01, the dissociation constant K.sub.d=3.83±1.26 nM in human serum is in good agreement with the K.sub.d=3.64±0.43 nM in PBS (see FIG. 3A. Similar can be observed for the antibody OUW4F1 against HLA B*08:01, with a K.sub.d=50.74±8.09 nM similar to K.sub.d=57.38±7.19 nM in PBS (see FIG. 3B). The different saturation levels under the different media conditions lie within the error range, and are most likely caused by minor conformational variations between individual antigens.

[0113] Referring to FIGS. 4A to 4D, there are shown data on testing of human patient serum against different HLA variants. By way of example, the size increase of both HLA A*02:01 and HLA A*24:02 can be an indication of protein interactions.

[0114] The hydrodynamic radius, Rh, of the measured species can be correlated with the molecular weight (Mw) of the biomolecule, which is then used to sum and examine whether the Rh values observed make sense as Rh can't be added since it is a volumetric measure and scales with molecular weight by approximately a one third power.

[0115] Referring to FIG. 4A, there is shown Bayesian analysis of the binding interaction of HLA A*02:01 with HLA-specific antibody in patient serum. From this, an antibody concentration in the nanomolar range can be determined, with a maximal probability value 6.02±2.66 nM. The K.sub.d is not well constrained, but has an upper limit K.sub.d≤30 nM.

[0116] Referring to FIG. 4B, there is shown Bayesian Analysis for the interaction of HLA A*24:02 with HLA-specific antibody in the serum of the same patient. Here, an antibody concentration of about 15.27±0.24 nM is found, for the K.sub.d only an upper limit of K.sub.d; 3 nM.

[0117] In some embodiments, the mean fluorescence intensities of a patient serum against HLA A*02:01 and A*24:01 at different serum concentration can be shown.

[0118] As shown in FIG. 4C, there is shown the Bayesian analysis of the binding interactions of antibody at ten times dilution of Patient Serum against HLA A*02:01. As shown in FIG. 4D, there is shown the Bayesian analysis of the binding interactions of antibody at ten times Dilution of Patient Serum against HLA A*24:02.

[0119] Subsequently, the applicability of this platform in patient serum is further described herein. For this purpose, the interaction of both HLA A*02:01 and HLA A*24:02 with the same patient serum is of interest. The patient can be highly sensitised for both HLA types, showing a higher Mean Fluorescence Intensity (MFI) on the Luminex platform for the HLA A*24:02. As shown in FIG. 4A, a size increase can be observed under the given conditions, showing that binding occurs.

[0120] By using Bayesian analysis, both the affinity and the antibody concentration can be quantified. For the interaction of serum allo-antibody with HLA A*02:01, the concentration of the antibody could be constrained to 6.02±2.66 nM, with a K.sub.d≤30 nM, assuming a binding ratio of one to two. Similarly, for the same patient serum, antibody binding to HLA A*24:02 can be investigated, which show a higher Luminex signal than for HLA A*02:01. The antibody concentration could be determined as 15.26±3.17 nM, and the affinity can be given with an upper limit of K.sub.d≤3 nM.

[0121] As shown in FIGS. 5A and 5B, there is shown a labelling strategy of the biomolecule of interest. FIG. 5A refers to a reaction scheme and FIG. 5B shows chromatogram on a Superdex 200 increase column from GE Healthcare with a flow rate of 0.5 mL/min and PBS as elution buffer. It is possible to deconvolute the signal into the fluorescence of interest and the background contribution.

[0122] Referring to FIG. 6A and FIG. 6B, there are shown analyses of sample autofluorescence. As shown in FIG. 6A, the fluorescence intensity increases with increasing serum concentration linearly, as expected. The absolute fluorescence intensity, thereby, differs between different patients. As shown in FIG. 6B, the serum fluorescence Intensity increases over course of 12 days, whilst the radii of labelled biomolecules in serum does not change over the same time course.

[0123] Using the apparatus and method as disclosed in the present invention, it is possible to determine the hydrodynamic radius Rh in human serum. The auto-fluorescence of human serum above 600 nm seems to be reduced, although a signal is detectable on the microfluidic platform.

[0124] Referring to FIG. 7, there is shown binding curve for the interaction between SN23OG6 against HLA A*02:01. The dissociation constants in both cases are consistent, and both have an equal stoichiometry of 1 to 2. It is possible to determine the stoichiometry and dissociation constant in antibody doped serum consistently with the buffer conditions.

[0125] FIGS. 8A to 8E shows a schematic of covalent labelling strategy as disclosed herein. FIG. 8A shows a reaction Mechanism of linking an amine on a protein to Alexafluor® 647 using an amide coupling. FIG. 8B shows a chromatogram of the purification of HLA A*02:01. FIG. 8C shows HLA A*03:01 and FIG. 8D HLA B*08:01 after labelling, showing the elution of HLA in one fraction. FIG. 8E shows a chromatogram for the purification of the commercially available streptavidin mixture.

[0126] FIG. 9A illustrates the determination of the Absorbance of varying concentration of Alexafluor® 647 in in human serum and FIG. 9B in buffer, showing no increased absorbance around 663 nm in human serum. FIG. 9C shows the fluorescence emission of Alexafluor® 647 in human serum and FIG. 9D in PBS. FIG. 9E shows a comparison of fluorescence emission of both human serum and PBS shows no difference. FIG. 9F shows the signal to noise aspect ratios in human serum and in PBS. The signal-to-noise ratio is slightly reduced in human serum compared to PBS.

[0127] FIG. 10 shows the structure of billirubin. The arrow indicates the rotation which is hindered by complexation of billirubin to HSA, a possible source of the fluorescence of the human serum.

[0128] Referring to FIGS. 11A to 11B, there is shown negative control experiments. FIG. 11A shows a comparison of hydrodynamic radii of Alexafluor® 647 labelled BSA, both pure and after incubation with different antibodies. This demonstrates that the HLA specific antibodies do not recognise the fluorophore, thus, every binding interaction determined can be assumed specific. FIG. 11B shows the hydrodynamic radii of different HLA variants determined purely or after incubation with 200 nM IgG (ab205198), showing no size increase and, thus, suggesting selective interaction between these HLA variants and specific alloantibodies.

[0129] FIG. 12 provides another negative control experiment. As shown in FIG. 12, the binding curve of 25 nM HLA A*03:01 with varying concentration of antibody W6/32. The blue points are averaged over the data measured in three replicates; the red line is the fit according to a Hills equation. From this data, the Hills coefficient h=1.01±0.15 could be determined.

[0130] Referring to FIG. 13A to 13B, there is shown global fit for binding curves. FIG. 13A shows a global fit for the binding curves for the interaction between HLA A*03:01 and antibody W6/32 at 25 nM and 500 pM HLA A*03:01. From this fit, a K.sub.d=466.0±149.4 pM could be determined. FIG. 13B shows a global fit for the interaction between HLA A*02:01 and SN23OG6 at 5 nM. 1 nM and 82 pM HLA A*02:0. This yields a K.sub.d=5.27±0.49 nM.

[0131] Referring to FIG. 14A there is shown a binding curve of blinded human serum against 1 nM HLA A*02:01. An antibody concentration [AB].sub.cal=138.7±47.9 and a K.sub.d=961.3±115.7 pM were obtained. After analysis, it was revealed that the blinded serum had 110 nM specific antibodies, showing the effectiveness in determining the [AB].sub.cal. Bayesian analysis of the blinded human serum is shown in FIG. 14B. The Bayesian plot shows that the ratio between antibody concentration and K.sub.d is well constrained. Binding curve with non-linear fit and with Bayesian analysis can be measured.

[0132] Analysis of MDS Data

[0133] The analysis of MDS data to establish a binding equilibrium between antibody and antigen are shown below. Assuming 1:1 binding, we therefore have the following equilibrium

Ab+HAbH  (1)

[0134] where AbH denotes the bound antibody-HLA complex.

[00001]$\begin{array}{cc}{K}_d=\frac{\left[\mathrm{Ab}\right]\left[H\right]}{\left[\mathrm{AbH}\right]}& \left(2\right)\end{array}$

[0135] Expressing the total concentrations of antibody and HLA as [Ab].sub.0 and [H].sub.0, respectively:

[00002]$\begin{array}{cc}\left[\mathrm{AbH}\right]=\frac{\left({\left[\mathrm{Ab}\right]}_0-\left[\mathrm{AbH}\right]\right)\left({\left[H\right]}_0-\left[\mathrm{AbH}\right]\right)}{K_d}& \left(3\right)\end{array}$$\begin{array}{cc}{\left[\mathrm{AbH}\right]}^2-\left({\left[\mathrm{Ab}\right]}_0+{\left[H\right]}_0+K_d\right)\left[\mathrm{AbH}\right]+{{\left[\mathrm{Ab}\right]}_0\left[H\right]}_0=0& \left(4\right)\end{array}$

[0136] Solving this:

[00003]$\begin{array}{cc}\left[\mathrm{AbH}\right]=\frac{{\left[\mathrm{Ab}\right]}_0+{\left[H\right]}_0+K_D-\sqrt{{\left({\left[\mathrm{Ab}\right]}_0+{\left[H\right]}_0+K_D\right)}^2}-{{4\left[\mathrm{Ab}\right]}_0\left[H\right]}_0}{2}& \left(5\right)\end{array}$

[0137] In the microfluidic diffusional sizing method, the measured hydrodynamic radius of our reaction mixture is calculated by recording the total fluorescence intensities of the ‘diffused’ and ‘undiffused’ channels, termed I.sub.d and I.sub.u, respectively.

[0138] Since the measured radius is not linearly correlated with the concentrations of free and bound labelled HLA, in order to avoid skewing the data, the fraction of labelled HLA that ends up in the ‘diffused’ channel I.sub.d/I.sub.d+I.sub.u can be observed and used. In order to relate the observed quantity to the predicted concentration bound, [AbH], pf and ρb are introduced as the fractions of free and bound HLA, respectively, that are detected in the ‘diffused’ channel. The following equations can be obtained as shown below:

I.sub.d=κ([AbH].sub.ρ.sub.b+([H].sub.0−[AbH]).sub.ρ.sub.f)  (6)

I.sub.u=κ([AbH](1-ρ.sub.b)+([H].sub.0−[AbH])(1-ρ.sub.f))  (7)

where κ is a constant that relates the concentration of HLA to the fluorescence intensity observed. Therefore the predicted fraction to end up in the diffused channel, f.sub.d is:

[00004]$\begin{array}{cc}{f}_d=\frac{\left({\left[\mathrm{AbH}\right]}_{{\rho }_b}+{\left({\left[H\right]}_0-\left[\mathrm{AbH}\right]\right)}_{{\rho }_f}\right)}{{\left[H\right]}_0}& \left(8\right)\end{array}$

with [AbH] as determined in the above equation. Therefore, the predicted fraction to end up in the diffused channel is a function of ρf.sub.d, ρb, K.sub.d and the total concentrations of antibody and HLA and this is denoted by f.sub.d(ρ.sub.f,ρ.sub.b,K.sub.d,[Ab].sub.0,[H].sub.0).

[0139] Using f.sub.d as the observable, the ρf.sub.d, ρ.sub.b, and K.sub.d can therefore be determined as the unknown parameters by Bayesian inference as disclosed in further detail below. In order to account for the actual 1:2 non-cooperative binding stoichiometry of antibody:HLA, we assume that the radii of the singly and doubly bound antibody are equal, as their sizes are within the expected error of the experimental method. Therefore, we simply use [Ab] to denote concentration of antibody binding sites, rather than antibody concentration, and K.sub.d is thus the dissociation constant with respect to binding sites. As each IgG antibody contains two binding sites, we simply multiply antibody concentration by two to obtain [Ab].

[0140] Determining Unknown Antibody Concentration

[0141] In patient serum samples, the concentration of antibody in serum is unknown. The concentration of HLA-reactive antibody binding sites in a given sample, [Ab].sub.0 is related to the concentration of binding sites in the original serum, [Ab].sub.tot, and the fraction of serum used in this particular sample, α, by

[Ab].sub.0=α[Ab].sub.tot  (9)

[0142] where 0≤α<1.

[0143] Substituting this into equation 5 as above, an updated equation is shown below:

[00005]$\begin{array}{cc}\left[\mathrm{AbH}\right]=\frac{\begin{array}{c}{\alpha \left[\mathrm{Ab}\right]}_{\mathrm{tot}}+{\left[H\right]}_0+K_D+\\ \sqrt{{\left({\alpha \left[\mathrm{Ab}\right]}_{\mathrm{tot}}+{\left[H\right]}_0+K_D\right)}^2-4{{a\left[\mathrm{Ab}\right]}_{\mathrm{tot}}\left[H\right]}_0}\end{array}}{2}& \left(10\right)\end{array}$

[0144] Serum Auto-Fluorescence

[0145] Using this expression in equation 8 we get a new expression for the predicted fraction to end up in the diffused channel, f.sub.d(ρ.sub.f,ρ.sub.b,K.sub.d,[Ab].sub.tot,α,[H].sub.0).

[0146] A further complication of performing the measurements in human serum is serum auto-fluorescence, which requires pre-processing of the data. Since the raw channel fluorescence intensities are measured, this background fluorescence can be corrected for. By performing a calibration of the serum fraction and intensity in each channel in the absence of HLA, a linear relationship between the fluorescence intensity of each channel and the serum fraction, a can be obtained:

I.sub.d,s=s.sub.dα  (11)

I.sub.u,s=s.sub.uα  (12)

where I.sub.d,s and I.sub.u,s are the intensities in the ‘diffused’ and ‘undiffused’ channels arising from the serum, respectively, and s.sub.d and s.sub.u are constants obtained through linear regression.

[0147] A datapoint, y, is thus computed during pre-processing by

[00006]$\begin{array}{cc}y=\frac{I_d-s_d\alpha }{\left(I_d+I_u\right)-\alpha \left(s_d+s_u\right)}& \left(13\right)\end{array}$

[0148] It is therefore possible to analyse the data by Bayesian inference, with ρ.sub.f, ρ.sub.b, K.sub.d, and [Ab].sub.tot as unknown parameters, thereby obtaining values for both K.sub.d, and [Ab].sub.tot through binding measurements by microfluidic diffusional sizing.

[0149] Bayesian Inference

[0150] The Bayesian inference analysis method utilises Bayes' theorem, and allows the determination of the probability distribution of unknown parameters, given the observed data, by the following equation

P(parameters|data)∝P(parameters)P(data|parameters)  (14)

where P(parameters|data) is known as the posterior, P(parameters) as the prior, and P(data|parameters) as the likelihood. The prior probability distribution is an expression of our information about the system before any measurement data can be acquired. For ρ.sub.f and ρ.sub.b, prior is assumed to be flat in linear space, whereas for the K.sub.d and total concentration of antibody, a prior that is flat in logarithmic space is more appropriate, to reflect the scale invariance of the problem.

[0151] Experimental measurement data as described herein are normally to be distributed about the true value, and the likelihood function is therefore a Gaussian, centred on the theoretical measurement value.

[00007]$\begin{array}{cc}P\left(\mathrm{data}.Math.\mathrm{parameters}\right)\propto \exp \left[-\frac{1}{2{\sigma }^2}{.Math.}_{i=1}^N{\left(y_i-f_d\left({\rho }_f,{\rho }_b,K_d,{\left[\mathrm{Ab}\right]}_{\mathrm{tot}},{\alpha }_i,{\left[H\right]}_i\right)\right)}^2\right]& \left(15\right)\end{array}$

where f.sub.d is defined in equation 8 with [AbH] defined in equation 10, α.sub.i and [H].sub.i are the concentrations of serum and antibody, respectively, in the i.sup.th measurement and yi is the pre-processed data point obtained in the i.sup.th measurement. In order to define an appropriate standard deviation, σ, for each dataset, the standard deviations of repeats of each measurement are calculated, and the maximum of these values are used as a global standard deviation for that dataset.

[0152] In order to maximise the information gained in each experiment, the entropy of the posterior distribution of the quantities of interest, e.g. the antibody concentration and affinity is used. The entropy of a distribution can simply be interpreted as a measure for how certain we are that a parameter had a specific value, the lower the entropy the more certain. Thus, the more a measurement decreases the entropy, the more information it contains. It can be predicted or estimated the entropy change if an additional measurement at a specific concentration of serum and antigen can be recorded. This is referred to as the expected entropy.

[0153] By calculating the expected entropy for all possible measurements i.e. all combinations of concentrations of serum and antigen, within the limits imposed by the experiment, the measurement point associated with the biggest expected decrease in entropy can be found. This can be performed in an iterative manner: after taking the first measurement, the expected entropy is calculated and the best next measurement is proposed. Once this measurement has been recorded, the expected entropies are updated and a new best next measurement is proposed. This process is repeated until the desired level of confidence and accuracy is obtained.

[0154] Referring to FIG. 15, there is provided a plot that shows a microfluidic antibody affinity profiling (MAAP) of serum antibody reactivity against the Receptor Binding Domain (RBD) of SARS-CoV-2 Spike protein in patients with COVID-19. Sera from patients with COVID-19 are examined at 28 days and at 90 days after onset of symptoms. Patients with the full spectrum of COVID-19 severity, from asymptomatic to patients admitted in intensive care unit, were examined. MAAP analysis showed that affinities (K.sub.D) of anti-RBD antibodies spanned over two orders of magnitude at both time points revealing a spectrum of humoral immunity against SARS-CoV-2.

[0155] Moreover, MAAP analysis showed strengthening of the antibody response against SARS-CoV-2 within this time frame (reduction in K.sub.D at 3 months compared to the 1 month time point) in the majority of the patients examined, providing direct evidence of affinity maturation in patients recovering from COVID-19. The data as shown in FIG. 15 demonstrates that MAAP analysis enables assessment of the magnitude and evolution of the humoral response directly in patient sera enabling further insights into the quality of the immune response after viral infection.

Examples—Experimental Details

[0156] Determination of Auto Fluorescence in Human Serum

[0157] Human serum (not containing specific anti-HLA antibodies) can be supplemented with PBS (pH 7.3, Oxoid tablets, Thermo Fisher Scientific Inc., Waltham, US; supplemented with NaN3 (0.02% (w/v), Sigma Aldrich, St Louis, US)) and the fluorescent label Alexafluor® 647 (Thermo Fisher Scientific Inc., Waltham, US),to yieldfluorophore concentrations between 10 pM and 1 pM in serum. Similar dilutions of fluorophore in buffer can be prepared for comparison. Subsequently, both absorption spectra and the emission spectra upon excitation at two wavelengths Δ.sub.ex,1=481 nm and Δ.sub.ex,2=632 nm can be recorded on a plate reader (Clariostar BMG Labtech, Ortenberg, Del.).

[0158] Labelling of HLA with Alexafluor® 647 Fluorophore

[0159] To label HLA (different variants from Emory, Atlanta, US; in NaHCO.sub.3(Sigma Aldrich, St Louis, US), 0.89 nmol, 1 equiv.), Alexa Fluor® 647 N-Hydroxysuccinimide ester (in DMSO 3 equiv.) can be added into a mixture containing HLA. The reaction mixture can be incubated for 1 hour at approx. 20° C., protected from light. The sample can be purified by size exclusion chromatography on a Superdex 200 increase 10/300 GL column (GE healthcare, Chicago, US) with a flow rate of approximately 0.5 mL/min and PBS (pH 7.3, supplemented with NaN3 (0.02% (w/v)) as eluent buffer, to yield labelled HLA (370 nM, DOL between 0.33 to 2.25, depending on variant). Conjugated HLA can then be stored at 4° C. until further use.

[0160] Microfluidic Diffusional Sizing (MDS)

[0161] All MDS experiments were performed using Fluidity One W (Fluidic Analytics, Cambridge, UK). The basic principle of MDS has been described elsewhere. In brief, labelled protein streams into the diffusion chamber from one side, auxiliary buffer from the other side. Due to the small channel size, laminar flow can be assumed, meaning that the particles can move into the buffer stream by diffusion only, whereby the rate depends on the size of the molecular complex. At the end of the diffusion channel or chamber, the stream is split again and the fluorescence at both sides of the chambers is measured. From the ratio between the fluorescence in both chambers, the hydrodynamic radius, Rh of the protein can be determined.

[0162] Verification of Binding

[0163] Labelled HLA A*03:01, A*02:01, and B*08:01, respectively, is diluted in PBS (pH 7.3, supplemented with NaN3 (0.02% (w/v)) to yield a 5 nM solution. The size of the HLA conjugates can be determined by MDS. Similarly, labelled HLA variants and respective antibodies are mixed, to yield a 5 nM concentration of HLA with 1 μM antibody. These conjugates were incubated at 4° C. for approx. 1 hour, then heated up to room temperature for 5 min and sized by MDS.

[0164] Negative Control Experiments

[0165] Labelled HLA variants were mixed with human IgG ab205198 (abcam, Cambridge, UK) or bovine serum albumin (BSA; Sigma Aldrich, St Louis, Del.) to yield a concentration of 5 nM of HLA variants and 1 μM antibody. These conjugates are incubated at 4° C. for 1 hour, then heated up to room temperature for 5 min and sized by MDS. Similarly, Alexafluor® 647 labelled BSA is diluted in PBS, to yield a solution of 5 nM of labelled BSA, as well as 250 equivalent of antibody, to demonstrate absence of unspecific interactions between the antibody and the fluorophore.

[0166] Measurements in PBS

[0167] Labelled HLA of a particular variant, together with a varying concentration of the antibody of interest, were added and diluted in PBS. The samples are incubated at 4° C. for approx. 1 hour or at room temperature for approx. 30 min. Subsequently, the size was determined by MDS. The data is fitted according to the non-linear binding equation (equation 16), which relates the measured hydrodynamic radius, R.sub.h,m, to the total concentrations of antibody, [Ab].sub.0, the binding site concentration, [B].sub.0, the dissociation constant K.sub.d, the stoichiometric ratio, n, the total increase in the hydrodynamic radius, ΔR.sub.h,tot and the hydrodynamic radius of unbound antigen, R.sub.h,0, using GraphPad Prism (Version 8.2).

[00008]$\begin{array}{cc}{R}_{h,m}=\left[\left(\frac{{\left[\mathrm{Ab}\right]}_0+{n\left[B\right]}_0+K_d}{2}-\sqrt{\frac{{\left({\left[\mathrm{Ab}\right]}_0+{n\left[B\right]}_0+K_d\right)}^2}{4}-{{n\left[B\right]}_0\left[\mathrm{Ab}\right]}_0}\right)\frac{\Delta R_{h,\mathrm{tot}}}{{n\left[B\right]}_0}\right]+R_{h,0}& \left(16\right)\end{array}$

[0168] For the fitting of the data to a Hill's equation 17 is used as shown below:

[00009]$\begin{array}{cc}{R}_{h,m}=\left[\left(\frac{{\left[\mathrm{Ab}\right]}_0^h+{\left[B\right]}_0+K_d^h}{2}-\sqrt{\frac{{\left({\left[\mathrm{Ab}\right]}_0^h+{\left[B\right]}_0+K_d^h\right)}^2}{4}-{{\left[B\right]}_0\left[\mathrm{Ab}\right]}_0^h}\right)\frac{\Delta R_{h,\mathrm{tot}}}{{\left[B\right]}_0}\right]+R_{h,0}& \left(17\right)\end{array}$

[0169] As, experimentally, a non-cooperativity of HLA to antibody binding has been validated, a binding ration of n=½ for 2 to 1 binding has been used further on in equation 16.

[0170] Binding Measurements in Human Serum

[0171] A range of different HLA types can be added to human serum, supplemented with varying concentrations of specific antibodies. The samples are incubated at room temperature (r.t.) for 30 min and subsequently, the size is determined by MDS. The data is fitted according to equation 18, using GraphPad Prism (Version 8.2).

[00010]$\begin{array}{cc}{R}_{h,m}=R_{h,\mathrm{HLA}}.Math.\left(1-\frac{\mathrm{qf}+p}{\mathrm{mf}+n}\right)+\left(\frac{\mathrm{qf}+p}{\mathrm{mf}+n}\right).Math.R_{h,\mathrm{bg}}& \left(18\right)\end{array}$

[0172] Bayesian Analysis

[0173] The priors used for the radius of the free species, rf, and the radius of the bound species, rb, are flat in linear spice, while a flat log-space prior is used for K.sub.d and binding site concentration, meaning that the probability of the K.sub.d being between 1 nM and 10 nM equals the probability of lying between 10 nM and 100 nM. A flat log-space allows facilitates the constraint the order of magnitude.

[0174] Various further aspects and embodiments of the present invention will be apparent to those skilled in the art in view of the present disclosure.

[0175] “and/or” where used herein is to be taken as specific disclosure of each of the two specified features or components with or without the other. For example “A and/or B” is to be taken as specific disclosure of each of (i) A, (ii) B and (iii) A and B, just as if each is set out individually herein.

[0176] Unless context dictates otherwise, the descriptions and definitions of the features set out above are not limited to any particular aspect or embodiment of the invention and apply equally to all aspects and embodiments which are described.

[0177] It will further be appreciated by those skilled in the art that although the invention has been described by way of example with reference to several embodiments. It is not limited to the disclosed embodiments and that alternative embodiments could be constructed without departing from the scope of the invention as defined in the appended claims.