Method for classifying monitoring results from an analytical sensor system arranged to monitor molecular interactions - Patents.com

Abstract

A method is disclosed for classifying monitoring results from an analytical sensor system 20 arranged to monitor molecular interactions on a sensing surface, wherein detection curves representing the progression of the molecular interactions over time are generated. The method includes steps of obtaining 100 a set of detection curves, fitting 101 a first mathematical model to the set of detection curves, calculating 102 a set of features from the set of detection curves and the fitted mathematical model, classifying 103 each detection curve into a quality classification group based on the calculated set of features, and determining a detection curve for use in kinetic analysis of the monitored molecular interactions based on the classification.

Description

[Technical field]

[0001] The present disclosure relates to a method for classifying monitoring results from an analytical sensor system arranged to monitor molecular interactions, as well as a computer program and a computer program product comprising program code means for carrying out said method. This document also relates to an analytical system for detecting molecular binding interactions and classifying the monitoring results. [Background technology]

[0002] Analytical sensor systems arranged to monitor interactions between molecules, such as biomolecules, in real time are often based on label-free biosensors, such as optical biosensors. A representative such biosensor system is the BIACORE® instrumentation, which uses surface plasmon resonance (SPR) to detect interactions between molecules in a sample and molecular structures immobilized on a sensing surface. The sample is passed over the sensor surface, and the progress of binding directly reflects the rate at which the interaction occurs. After the sample is injected, a buffer is allowed to flow, during which the detector response reflects the rate of dissociation of the complex on the surface.

[0003] A typical output from the BIACORE® system and similar biosensor systems is a response graph or curve that describes the progression of a molecular interaction over time, including association and dissociation phase portions. This response graph or detection curve, usually displayed on a computer screen, is often called a binding curve or "sensorgram."

[0004] Using the BIACORE® system (and similar sensor systems), it is possible to determine several interaction parameters for molecules used as ligands and analytes, including the kinetic rate constants for binding (association) and dissociation of the intermolecular interaction, as well as the interaction affinity.

[0005] In the evaluation of the kinetics of a ligand-analyte system, classifying the detection curves as good or bad can be cumbersome and difficult even for an experienced user of the sensor system, especially when the evaluation includes a large number of detection curves. Curves of low quality should be excluded from further analysis of the interaction kinetics, as they may negatively affect the evaluation. Different users may classify detection curves differently (i.e., as good or bad quality), so the evaluation of the ligand-analyte system may vary and be user-dependent. Curves of low quality may include, for example, unstable baselines, air spikes, responses below baseline, etc., and may be caused, for example, by a contaminated flow system or running buffer, the ligand capture approach used, or too low concentration detergent in the running buffer.

[0006] WO2003081425 describes a method and an analysis system for data processing of a large set of detection curves representing molecular interactions on a sensor surface, and assists the user in classifying the curves with respect to quality. The detection curves are subjected to a quality assessment that includes the steps of selecting a quality parameter for each detection curve based on at least one quality-related parameter (e.g., baseline slope, air spikes, oscillations), each parameter being defined by at least one quality descriptor, calculating for each detection curve value in terms of the descriptor, and calculating for each detection curve a quality classification indicating the quality of the detection curve with respect to all detection curves of the set of curves. Detection curves classified as outliers are selected and a validation procedure is performed to determine whether the deviating detection curves should be included in the subsequent kinetic analysis.

[0007] Use of the method of WO2003081425 requires an understanding of the algorithms used in data processing, and the method also takes a significant amount of time to fine-tune (train). Summary of the Invention [Problem to be solved by the invention]

[0008] It is an object of the present disclosure to provide a simpler and faster, or at least an alternative, method to known methods for classifying monitoring results from an analytical sensor system arranged to monitor molecular interactions. A further object is to provide a computer program and a computer program product comprising product code means for carrying out the method. Furthermore, an object is to provide an analytical system for monitoring molecular binding interactions and classifying the monitoring results.

[0009] The invention is defined by the accompanying independent patent claims. Non-limiting embodiments arise from the dependent patent claims, the accompanying drawings and the following description. [Means for solving the problem]

[0010] According to a first aspect, there is provided a method for classifying monitoring results from an analytical sensor system arranged to monitor an intermolecular interaction, wherein a detection curve is generated that represents the progression of the intermolecular interaction over time. The method includes the steps of: obtaining a set of detection curves, the set of detection curves including one or more detection curves that represent the intermolecular interaction at respective molecular concentrations; fitting a mathematical model to the set of detection curves; and calculating a set of features from the set of detection curves and the fitted mathematical model, the calculated set of features including: association rate constant ka divided by standard error of ka, dissociation rate constant kd divided by standard error of kd, maximum binding capacity Rmax divided by standard error of Rmax, mass transfer limitation tc divided by standard error of tc, delayed binding response B divided by Rmax, and mean squared error (MSE) between the detection curve and the fitted mathematical model. 2 Based on the set of calculated features, each detection curve is classified into a quality classification group indicating a quality of the detection curve.

[0011] The molecular interaction to be monitored may occur at a sensing surface, for example a substance detection biosensor system such as an optical biosensor, where one molecule, the ligand, may be attached / immobilized on the sensor surface and the other molecule, the analyte, may pass over the sensor surface to monitor the progress of analyte-ligand binding over time. Analytical sensor systems based on other detection principles, such as electrochemical systems, are also possible.

[0012] A set of detection curves can include at least one detection curve. When the set of detection curves includes more than one detection curve, e.g., 2 to 10 detection curves, each detection curve in the set of detection curves represents a molecular interaction (between the same ligand and analyte) at a different molecular concentration, i.e., a different analyte concentration.

[0013] The mathematical model that is fitted to the set of detection curves is a model that describes the molecular interactions over time. The model may be, for example, a 1:1 binding model. Alternatively, a heterogeneous ligand binding model, a heterogeneous analyte binding model, or a bivalent analyte binding model can be used.

[0014] The parameter values ​​of ka, kd, Rmax and tc are divided by their respective standard errors. The standard error of a parameter is a measure of how important the parameter is to the closeness of the fitted mathematical model. Rmax is the calculated Rmax from the fit. The delayed binding response B is the maximum binding response in the detection curve relative to the baseline in the detection curve. The delayed binding response feature B is normalized by dividing it by Rmax, and the mean squared error feature MSE is the squared delayed binding response B 2 The features are normalized to be applicable to all types of input data.

[0015] The steps of fitting a mathematical model to the set of detection curves, calculating a set of features from the set of detection curves and the fitted mathematical model, and classifying each detection curve into a quality classification group may be performed by an experienced user of the analytical sensor system. Alternatively, the quality classification step may be performed by an ANN or expert system trained on multiple training sets of detection curves with the respective fitted mathematical models, calculated sets of features and quality classifications. The parameters used in the training are determined by the experienced user.

[0016] Detection curves are classified into quality classification groups that indicate the quality of the detection curve using the method. Curves of low quality should be excluded from further analysis of the kinetics of the monitored interaction, as they may negatively affect the evaluation. Low quality curves may include, for example, unstable baselines, air spikes, responses below baseline, etc., and may be caused, for example, by a contaminated flow system or running buffer, the ligand capture approach used, or too low concentration detergent in the running buffer. What is a sufficiently good quality of a detection curve and whether to include it in the kinetic analysis of the monitored interaction depends on the purpose of the analysis / experiment.

[0017] The number of different quality classification groups can be two in one example, the first group includes detection curves with sufficiently good quality, and the second group includes detection curves with low quality.The group with detection curves with sufficiently good quality can then be selected for dynamic analysis of monitored molecular interactions.

[0018] In another example, the number of quality classification groups can be 100, with group 1 containing detection curves of low quality and group 100 containing detection curves of very good quality. The user can adjust the stringency of the method and place a cutoff at, for example, >50, thereby including only detection curves classified in classification groups 51-100 in the dynamic analysis of the monitored interactions. Again, whether the quality of the detection curves is good enough and where to place the cutoff depends on the goal of the analysis / experiment.

[0019] The set of features calculated may be three or more. Using multiple features in this way can improve the classification of the detection curve if the features complement each other and highlight something that distinguishes what an experienced user considers to be good enough quality from not so good quality. Too many features that do not contribute to the classification may reduce the quality of the neural network training and classification. It is also possible to calculate additional features not mentioned above and include them in the step of classifying the detection curve.

[0020] The mathematical model can be selected from a 1:1 binding model, a heterogeneous ligand binding model, a heterogeneous analyte binding model, and a bivalent analyte binding model.

[0021] Based on the classification, a detection curve can be determined for use in the kinetic analysis of the monitored molecular interactions.

[0022] As mentioned above, the decision of which of the classified detection curves should be included in the kinetic analysis of the monitored interactions and where the cut-off should be placed depends on the goal of the analysis / experiment.

[0023] Through such kinetic analysis, the association constant, ka, the dissociation constant, kd, and the interaction affinity for the interaction can be obtained.

[0024] The method may further include determining a second mathematical model for use in the dynamic analysis.

[0025] The second mathematical model used in the dynamic analysis of the detection curve does not necessarily have to be the same mathematical model as the mathematical model used in the quality assessment procedure described above.

[0026] The second mathematical model can be selected from a 1:1 binding model, a heterogeneous ligand binding model, a heterogeneous analyte binding model, and a bivalent analyte binding model.

[0027] Classifying the detection curves into quality classification groups can be done by an artificial neural network or an expert system.

[0028] The classification can be performed by one ANN / expert system or two or more such systems working in conjunction with each other.

[0029] The artificial neural network can be trained with a plurality of sets of detection curves representing the progression of different molecular interactions over time, the artificial neural network being provided with, for each set of detection curves, a) a set of features calculated from the set of detection curves and a mathematical model fitted to the set of detection curves, the calculated set of features being: the association rate constant ka divided by the standard error of ka; the dissociation rate constant kd divided by the standard error of kd; the maximum binding capacity Rmax divided by the standard error of Rmax; the mass transfer limiting value tc divided by the standard error of tc; the delayed binding response B divided by Rmax; and the mean squared error MSE between the detection curves and the fitted mathematical model. 2 and b) a classification of each detection curve into a quality classification group.

[0030] The classification of the detection curves into quality classification groups is obtained by visual inspection of the detection curves by an experienced user and is fed into an artificial neural network (ANN) along with the calculated features. The mathematical model used to fit the data is fixed for training the ANN and can be a 1:1 binding model, a heterogeneous ligand model, a heterogeneous analyte model, or a bivalent analyte binding model.

[0031] When using such a trained ANN to perform the classification step of the above-mentioned method, the ANN is preferably trained on the same set of calculated features that are fed into the trained ANN for performing the classification step. However, in some embodiments, it is possible to feed a set of calculated features in the classification step into a trained ANN that includes a smaller number of calculated features than those used to train the ANN.

[0032] Determining the second mathematical model used in the dynamic analysis may be performed by an artificial neural network or an expert system.

[0033] The artificial neural network can be trained using a number of detection curves representing the progression of molecular interactions over time, and the artificial neural network is provided with a classification of the detection curves in terms of a mathematical model that fits the detection curves.

[0034] A classification of mathematical models for use in the dynamic analysis of different detection curves can be evaluated by an experienced user and the results are fed into the ANN.

[0035] According to a second aspect, there is provided an analytical system for monitoring molecular binding interactions and classifying the monitoring results, comprising: a) a sensor device comprising at least one sensing surface, detection means for detecting molecular interactions at said at least one sensing surface and means for generating a detection curve representative of the progression of the interaction over time, and b) data processing means for classifying each detection curve in terms of a first or second group, said data processing means performing said steps b) to d).

[0036] According to a third aspect, there is provided a computer program comprising program code means for performing the method as defined above, when the program is run on a computer.

[0037] According to a fourth aspect there is provided a computer program product comprising program code means stored on a computer readable medium for performing the method as set out above when the program is run on a computer. [Brief description of the drawings]

[0038] [Figure 1] FIG. 1 shows a schematic diagram of an analytical system for detecting molecular binding interactions and classifying the monitoring results. [Diagram 2] 1 shows a detection curve over time from the interaction of a sample with a target molecule. [Diagram 3] Two acceptable detection curves (left) and two unacceptable detection curves (right) are shown. [Figure 4] 1 is a flow chart showing steps in a method for quality assessment of a set of detection curves representative of monitored intermolecular interactions. [Diagram 5] 1 is a flow chart showing inputs to an artificial neural network (ANN) for training the ANN to classify detection curves into quality classification groups. [Figure 6] Delayed binding response B is the detection curve shown. [Figure 7]Dynamics Screen Shows an example of how a machine learning application processes training and predicting data. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0039] The present disclosure relates to analytical sensor methods, in particular biosensor-based methods, in which molecular interactions are studied and the results are presented in the form of detection curves, often called sensorgrams, in real time as the interaction progresses.

[0040] Biosensors can be based on various detection methods. Typically, such methods include, but are not limited to, substance detection methods such as piezoelectric, optical, thermo-optical and surface acoustic wave (SAW) device methods, as well as electrochemical methods such as potentiometric, conductive, amperometric and capacitive methods. With regard to optical detection methods, representative methods include methods that detect substance surface concentrations, such as reflectance-optical methods (including both internal and external reflection methods), angle, wavelength or phase resolution, such as ellipsometry and evanescent wave spectroscopy (EWS), the latter including surface plasmon resonance (SPR) spectroscopy, Brewster angle refractometry, critical angle refractometry, frustrated total reflection (FTR), evanescent wave ellipsometry, scattered total internal reflection (STIR), optical waveguide sensors, evanescent wave-based imaging (such as critical angle resolved imaging, Brewster angle resolved imaging, SPR angle resolved imaging, etc.). Additionally, photometric methods based on, for example, evanescent fluorescence (TIRF) and phosphorescence, as well as waveguide interferometry, can also be used.

[0041] One commonly used detection principle is surface plasmon resonance (SPR) spectroscopy. An exemplary type of SPR-based biosensor is sold under the trade name BIACORE® (hereinafter referred to as "BIACORE device"). These biosensors utilize SPR-based substance sensing technology to provide "real-time" label-free binding interaction analysis between surface-bound ligands and analytes of interest.

[0042] The BIACORE device includes a light-emitting diode (LED), a sensor chip containing a glass plate covered with a thin gold film, an integrated fluidic cartridge that provides a liquid flow over the sensor chip, and a photodetector array. Incident light from the LED is totally internally reflected at the glass / gold interface and detected by the photodetector array. At a certain angle of incidence (the "SPR angle"), surface plasmon waves are generated in the gold layer and detected as an intensity loss "or dip" in the reflected light. The phenomenon of SPR relevant to the BIACORE device relies on resonant coupling of monochromatic p-polarized light incident on a thin metal film through a prism and the glass plate with vibrations of conducting electrons, called plasmons, in the metal film on the other side of the glass plate. These vibrations give rise to an evanescent field that extends from the surface to the liquid flow for distances on the order of one wavelength (~1 μm). When resonance occurs, light energy is lost to the metal film by collective excitation of electrons, and the intensity of the reflected light drops off at a sharply defined angle of incidence, the SPR angle. The SPR angle depends on the refractive index within the reach of the evanescent field near the metal surface.

[0043] As mentioned above, the SPR angle depends on the refractive index of the medium close to the gold layer. In the BIACORE device, dextran is typically bound to the gold surface and an analyte-binding ligand is bound to the surface of the dextran layer. The analyte of interest is injected in solution form onto the sensor surface via a fluidic cartridge. Since the refractive index in the vicinity of the gold film depends on (i) the refractive index of the solution (which is constant) and (ii) the amount of material bound to the surface, the binding interaction between the bound ligand and the analyte can be monitored as a function of the change in the SPR angle.

[0044] In Fig. 1, a schematic diagram of an analytical system 20 for detecting molecular binding interactions and classifying the monitoring results is shown. The analytical system includes a BIACORE™ device, which is a sensor device, a chip 1 with a sensing surface, a gold film 2, supporting capture molecules 3 (e.g. antibodies), exposed to a sample flow with an analyte 4 (e.g. antigen) via a flow channel 5. Monochromatic p-polarized light 6 from a light source 7 (LED) is coupled by a prism 8 to a glass / metal interface 9 where the light is totally internally reflected. The intensity of the reflected light beam 10 is detected by a detection means 11, such as an optical photodetector array.

[0045] A typical output from a BIACORE instrument is a "sensorgram," which is a plot of response (measured in "resonance units" or "RU") as a function of time. An increase of 1,000 RU corresponds to an increase in material on the sensor surface of approximately 1 ng / mm2. When a sample containing an analyte is contacted with the sensor surface, a ligand bound to the sensor surface interacts with the analyte in a step called "association." This step is indicated on the sensorgram by an increase in RU when the sample is first contacted with the sensor surface. Conversely, "dissociation" usually occurs when the sample flow is replaced by, for example, a buffer flow. This step is indicated on the sensorgram by a decrease in RU over time as the analyte dissociates from the surface-bound ligand.

[0046] A representative sensorgram of the BIACORE device is shown in FIG. 2. FIG. 2 shows a sensing surface with immobilized ligands (e.g., antibodies) that interact with an analyte in a sample. The y-axis shows the response (here in resonance units (RU)) and the x-axis shows time (here in seconds). Initially, a buffer solution is passed over the sensing surface, giving a "baseline response" in the sensorgram. During sample injection, an increase in signal is observed due to binding (i.e., association) of the analyte to a steady-state condition where the resonance signal reaches a plateau. At the end of sample injection, the sample is replaced with a continuous flow of buffer and a decrease in signal reflects dissociation or release of the analyte from the surface. The slope of the association / dissociation curve provides valuable information about the kinetics of the interaction, while the height of the resonance signal represents the surface concentration (i.e., the response resulting from the interaction is related to the change in the concentration of the substance on the surface). The analysis system 20 shown in FIG. 1 comprises a means 12 for generating a detection curve that represents the progress of the interaction over time.

[0047] Detection curves or sensorgrams generated by biosensor systems based on other detection principles have a similar appearance.

[0048] Sometimes, the generated sensorgrams are of unacceptable quality for various reasons and therefore must be discarded. Figure 3 shows examples of two acceptable and two unacceptable sensorgrams. The two curves on the left are both acceptable. On the other hand, the curve on the top right is too unstable and the curve on the bottom right is distorted by air peaks (air bubbles in the fluid flow). The control of the quality of the sensorgrams is usually done by the user by creating an overlay plot of the curves to be analyzed and visually searching for curve anomalies. Unacceptable curves can be due to, for example, an unstable baseline, responses below the baseline, etc., and can be caused, for example, by a contaminated flow system or running buffer, the ligand capture approach used, or a too low detergent concentration in the running buffer.

[0049] The current trend in biosensor systems is the development towards high throughput systems that can generate large sets of sensorgrams in a relatively short time. With already modest increases in throughput, it is easy to see that it is not feasible for a user, even an experienced user, to inspect all sensorgrams at once to assess their quality. Furthermore, since different users may classify detection curves differently (i.e., as being of sufficiently good or poor quality), the assessment of a ligand-analyte system may vary and be user-dependent.

[0050] A method for classifying monitoring results from an analytical sensor system generating detection curves representing the progression of intermolecular interactions over time is shown in Fig. 4, and data processing means 13 for performing such a task is shown in Fig. 1. A set of detection curves comprising n (e.g., 1-8) different detection curves is obtained (100) from the analytical sensor system. The n detection curves represent the monitored molecular interactions at n different molecular concentrations, i.e. different analyte concentrations.

[0051] A mathematical model is fitted to the set of n detection curves (101). The mathematical model may be a 1:1 binding model. This is the simplest model for kinetic evaluation. The model describes a 1:1 interaction at the surface. The kinetic parameters include ka - association rate constant for the formation of the ligand-analyte complex; kd - dissociation rate constant for the ligand-analyte complex; Rmax - maximum binding capacity; and tc - mass transfer limit. Usually, these parameters are fitted broadly, but they can also be local or constant. The model is sometimes called the Langmuir model, since it corresponds to the Langmuir isotherm for the adsorption of a substance to a solid surface. The transfer of analyte from the bulk solution to the surface is directly proportional to the bulk analyte concentration, and the proportionality constant is the mass transfer coefficient, which is a function of the flow rate, the flow cell dimensions, and the diffusion properties of the analyte. This constant has units of RU·M-1s-1. For globular proteins with molecular weights of the order of 50,000 Daltons, a typical value for the mass transfer coefficient is of the order of 108 RU·M-1s-1. Reported values ​​that differ widely, by orders of magnitude (e.g., 1012 or 1014), may indicate that the parameter is not important to the fit (i.e., the observed binding is not limited by mass transport). A term, tc, for the rate of transfer of the analyte from the bulk solution to the surface is included in the 1:1 binding model.

[0052] Other mathematical / kinetic models that can be used and fitted to a set of detection curves are the heterogeneous ligand binding model, the heterogeneous analyte binding model, and the bivalent analyte binding model, which are less commonly used than the 1:1 binding model.

[0053] The heterogeneous ligand model describes the presence of two ligand species that bind to the analyte independently of each other. The species can be different molecules or different binding sites on the same ligand molecule. Here, the kinetic parameters are: association rate constant for the formation of ka1-ligand 1-analyte complex; association rate constant for the formation of ka2-ligand 2-analyte complex; dissociation rate constant for kd1-ligand 1-analyte complex; dissociation rate constant for kd2-ligand 2-analyte complex. The complexity of the mathematical model limits the number of ligand species to two.

[0054] The heterogeneous analyte model is primarily intended for situations where two analytes of different sizes are intentionally mixed. The model describes this competitive condition and returns two sets of rate constants, one set for each reaction. The kinetic parameters are the association rate constant for the formation of the ka1-analyte 1-ligand complex; the association rate constant for the formation of the ka2-analyte 2-ligand complex; the dissociation rate constant for the kd1-complex analyte 1-ligand; and the dissociation rate constant for the kd2-complex analyte 2-ligand. The heterogeneous analyte model is useful to indirectly determine the kinetics of a small analyte by competition with a large analyte. Although response contributions from both analytes were considered, the high molecular weight analyte was responsible for the major component of the observed sensorgrams. Concentrations and molecular weights are required for both analytes. If the absolute molecular weights are unknown, relative values ​​can be entered without affecting the outcome of the fit. The model cannot evaluate interactions when the proportions and relative sizes of the analytes are unknown.

[0055] The bivalent analyte binding model describes the binding of bivalent analytes to immobilized ligands, where one analyte molecule can bind one or two ligand molecules. The two analyte sites are assumed to be equivalent. The model may be particularly relevant for studies with signaling molecules that bind to immobilized cell surface receptors (receptor dimerization is common) and with intact antibodies that bind to immobilized antigens. As a result of the binding of one analyte molecule to two ligand sites, the overall binding is enhanced compared to 1:1 binding. This effect is often called avidity. The kinetic parameters are ka1 - association rate constant for the formation of the analyte-ligand site 1 complex; ka2 - association rate constant for the formation of the analyte-ligand site 2-ligand site 2 complex; kd1 - dissociation rate constant for the complex analyte-ligand site 1 complex; and kd2 - dissociation rate constant for the complex analyte-ligand site 1-ligand site 2 complex. Binding at the second site does not change the surface material, so no response occurs. For this reason, association rate constants for the second interaction are reported in units of RU-1s-1 and are only available in M-1s-1 when conversion factors between RU and M are available. Similarly, values ​​of the overall affinity or avidity constants are not reported. In general, it is preferred to immobilize bivalent reactants, thereby avoiding complications caused by combined affinities resulting from multivalent binding (avidity). In some cases, avidity effects can be reduced by using very low ligand levels and high analyte concentrations. Low ligand levels give sparsely distributed ligands where it is unlikely that two ligand molecules are within reach of a single analyte. High analyte concentrations compete for second site binding and thus favor the formation of 1:1 complexes.

[0056] After fitting the mathematical model to the set of n detection curves, a set of features from the detection curves and the fitted mathematical model is calculated 102. The set of features may include three or more of the following: - the association rate constant ka divided by the standard error of ka, - the dissociation rate constant kd divided by the standard error of kd, - the maximum binding capacity Rmax divided by the standard error of Rmax, - mass transfer limit tc divided by the standard error of tc, - delayed binding response B divided by Rmax, and - The mean squared error (MSE) between the detection curve and the fitted mathematical model is taken as the squared delayed combined response (B 2 ) divided by

[0057] The delayed binding response B is the maximum binding response in the detection curve relative to the baseline in the detection curve, see legend to FIG.

[0058] The features and number of features calculated and used in the feature set may depend on the mathematical model used.

[0059] Based on the set of calculated features, each detection curve is classified into a quality classification group indicating the quality of the detection curve, here shown as four different classification groups. The number of different quality classification groups may vary. As mentioned above, curves of low quality should be excluded from the dynamic analysis of the monitored interactions, as they may negatively affect the evaluation. What is a sufficiently good quality of a detection curve and whether to include it in the dynamic analysis of the monitored interactions depends on the purpose of the analysis / experiment.

[0060] The classification group / groups containing detection curves of sufficiently good quality can then be selected for the kinetic analysis 104 of the monitored molecular interactions. In the example of FIG. 4, the detection curves classified into the first and second classification groups are considered to have a sufficiently good quality for the kinetic analysis of the monitored molecular interactions, while the detection curves classified into the third and second groups are of lower quality and are not included in the kinetic analysis. The user can adjust the stringency of the method and set a cutoff, for example, so that only the detection curves of the first group or the detection curves of groups 1-3 are used for the kinetic analysis.

[0061] A mathematical model can be fitted to the detection curves classified as being of sufficiently good quality for dynamic analysis of the interaction to determine parameters such as ka and kd for the interaction. This mathematical model can be the same as the mathematical model used to classify the curve, or it can be a different model. The model can be selected from any of the binding models described above.

[0062] The step 103 of classifying the detection curves into different quality classification groups can be performed by an artificial neural network (ANN) or an expert system. To do this, the ANN needs to be trained on data from multiple sets 200 of detection curves representing the progression of different molecular interactions over time. The data fed into the ANN for training includes, for each set 202 of detection curves, features calculated from the set of detection curves and a mathematical model 201 fitted to the set of detection curves. The mathematical model used to fit the data can be fixed for the training of the ANN. The set of features can include three or more of the following: - the association rate constant ka divided by the standard error of ka, - the dissociation rate constant kd divided by the standard error of kd, - the maximum binding capacity Rmax divided by the standard error of Rmax, - mass transfer limit tc divided by the standard error of tc, - delayed binding response B divided by Rmax, and - The mean squared error (MSE) between the detection curve and the fitted mathematical model is taken as the squared delayed combined response (B 2 ) divided by

[0063] The ANN is also provided with a classification of each detection curve into a quality classification group indicative of the quality of the detection curve 202. Such classification of the detection curves may be determined by visual inspection of the detection curves by an experienced user and entered into the ANN model. The training of the ANN model is shown in FIG.

[0064] Such a trained ANN model then returns a quality classification of the unknown detection curve into a quality classification group.

[0065] The ANN or expert system may also be trained to assist a user of an analytical sensor system in determining the mathematical model to be used in the dynamic analysis of the detection curves. The artificial neural network can be trained using a plurality of detection curves representing molecular interactions at the sensor surface, and the artificial neural network is provided with a classification of the detection curves determined by an experienced user as to which mathematical model fits the detection curve appropriately. The trained ANN can predict with high accuracy which mathematical model to use for a set of data.

[0066] The trained ANN uses both the detection curve data features and the associated fitting parameters to mimic how an experienced user classifies the detection curves into quality classification groups. The training may be a standard ANN training that involves splitting the entire data set into a training set and a validation set, where 80% of the data is used for training and 20% of the data is used for validation. The ANN should be trained on the training data and not on the validation set. With one hidden layer and a few hidden nodes, we can obtain an ANN that is a generalized detection curve classifier. If two hidden layers and too many hidden nodes are allowed, the ANN will be very good at tuning to the training data but poor at classifying / predicting the validation set or other non-training data.

[0067] In one example, the trained ANN model included one hidden layer, 15 hidden nodes, a cross-entropy loss function, and a learning rate of 0.1, and was built using Microsoft Azure Machine Learning Studio in the cloud. For training the ANN model, for each detection curve, feature data were calculated, normalized if necessary with Biacore dynamic application, classification was added, and sent to the cloud. Data can be added to the entire training dataset, and older data of the same origin can be removed from the training set. Different algorithms may be used to handle duplicates, etc. There may be several ANN models in the cloud, one per research group or company, or there may be a shared ANN shared among several users in a community. The output from the training is an ANN model that can be used for quality classification of the detection curves.

[0068] Figure 7 shows an example of how a dynamics-based machine learning application processes training and predicting data.

[0069] In this example, we use three data sets with six detection curves each. The circled curves in each set are the detection curves, some with different concentrations. The solid black curves are the curves fitted using a mathematical model, in this case the 1:1 binding Langmuir model. There is one fitted curve for each detection curve. The application expert can label the data sets as follows: R = rejected series, and A = accepted series based on the quality of the data.

[0070] For each set, features are calculated using both the detection curve, the mathematical model curve and the results from the fitted mathematical model.

[0071] [Table 1]

[0072] Classifying each detection curve into a quality classification group that indicates the quality of the detection curve is our method of training a machine learning model, in this case an artificial neural network (ANN), using the quality label data and all (or at least three) of the feature datasets. The ANN uses one hidden layer and 15 hidden nodes, with a cross-entropy loss function and a learning rate of 0.1. Each column was normalized using a Gaussian normalization function, and the ANN was trained and evaluated using Microsoft Azure Machine Learning Studio in the cloud. The goal of the ANN is then to accurately classify the quality labels using only the features. Once trained, when a new set of data without quality labels is analyzed, the ANN can predict whether such a dataset should be automatically accepted or rejected based on the features alone, which may save the user's time and also allow operation of the system by non-expert users.

[0073] With the continuous training and development of these models, experienced users can easily correct the classification results and retrain the models in the cloud. Furthermore, by using such trained ANNs to classify detection curves, the classification process can be speeded up, especially when there are a large number of detection curves in the analysis, and the evaluation results are less user-dependent, allowing less experienced users to be assisted in the evaluation of the monitored molecular interactions.

[0074] Although the above description includes multiple specificities, these should not be construed as limiting the scope of the concepts described herein, but merely providing a description of some exemplary embodiments of the concepts described. It will be understood that the scope of the concepts described herein fully encompasses other embodiments that may become apparent to those skilled in the art, and thus the scope of the concepts described herein is not limited. Reference to a singular element is not intended to mean "one and only one" unless expressly so stated, but rather "one or more." All structural and functional equivalents to the elements of the above-described embodiments known to those skilled in the art are expressly incorporated herein and are intended to be encompassed herein.

Claims

1. 1. A method for classifying monitoring results from an analytical sensor system arranged to monitor a molecular interaction, wherein a detection curve is generated representing the progression of said molecular interaction over time, said method comprising: a) obtaining a set of detection curves (100), the set of detection curves including one or more detection curves representing molecular interactions at respective molecular concentrations; b) fitting (101) a mathematical model to said set of detection curves; c) calculating (102) a set of features from said set of detection curves and the fitted mathematical model, said calculated set of features comprising: The association rate constant ka divided by the standard error of ka, Dissociation rate constant kd divided by the standard error of kd, Maximum binding capacity Rmax divided by the standard error of Rmax, Mass transfer limit tc divided by the standard error of tc, the delayed binding response B divided by Rmax, and The mean squared error (MSE) between the detection curve and the fitted mathematical model is expressed as the squared delay combination response (B 2 ) divided by and d) classifying (103) each detection curve into a quality classification group indicative of a quality of said detection curve based on said calculated set of features; Including, the analytical sensor system is a ligand-analyte system; the monitored result is a function of a change in surface plasmon resonance (SPR) angle; the mathematical model is a model describing molecular interactions over time; the ka, kd, Rmax, and tc are determined based on the fitted mathematical model; B is the maximum binding response in the detection curve relative to the baseline in the detection curve; A method, wherein the step of classifying (103) the detection curves into quality classification groups is performed by an artificial neural network or an expert system.

2. 2. The method of claim 1, wherein the mathematical model is selected from a 1:1 binding model, a heterogeneous ligand binding model, a heterogeneous analyte binding model, and a bivalent analyte binding model.

3. The method according to claim 1 or 2, wherein the molecular interactions are monitored at a sensing surface.

4. 4. The method according to claim 1, further comprising the step of determining (104) a detection curve for use in a kinetic analysis of the monitored molecular interactions based on the classification.

5. determining a second mathematical model for use in the dynamic analysis; The method of claim 4, wherein the second mathematical model is a model describing intermolecular interactions over time.

6. 6. The method of claim 5, wherein the second mathematical model is selected from a 1:1 binding model, a heterogeneous ligand binding model, a heterogeneous analyte binding model, and a bivalent analyte binding model.

7. The artificial neural network is trained using a plurality of sets of detection curves (200) representing the progression of different molecular interactions over time, and the artificial neural network performs, for each set of detection curves: a) a set of features calculated from said set of detection curves and a mathematical model fitted to said set of detection curves (202), said calculated set of features comprising: - the association rate constant ka divided by the standard error of ka, - the dissociation rate constant kd divided by the standard error of kd, - the maximum binding capacity Rmax divided by the standard error of Rmax, - mass transfer limit tc divided by the standard error of tc, - delayed binding response B divided by Rmax, and - the mean squared error (MSE) between the detection curve and the fitted mathematical model is calculated as the squared delay combined response (B 2 ) divided by and a set including three or more of the following: b) Classifying each detection curve into a quality classification group (202); is provided, the ka, kd, Rmax, and tc are determined based on the fitted mathematical model; B is the maximum binding response in the detection curve relative to the baseline in the detection curve; 7. The method according to claim 1, wherein the classification (202) of each detection curve into a quality classification group is performed by an artificial neural network or an expert system.

8. 7. The method according to claim 5 or 6, wherein the step of determining the second mathematical model to be used in the dynamic analysis is performed by an artificial neural network or an expert system.

9. 9. The method of claim 8, wherein the artificial neural network is trained with a plurality of sets of detection curves representing the progression of molecular interactions over time, and the artificial neural network is provided with a classification of the detection curves as to which mathematical model fits the detection curves.

10. 10. An analytical system (20) for detecting molecular binding interactions and classifying monitoring results, comprising: a) a sensor device (1) comprising at least one sensing surface (2), detection means (11) for detecting molecular interactions on said at least one sensing surface (2) and means (12) for generating detection curves representative of the progression of said interactions over time; and b) data processing means (13) for classifying each detection curve into a quality classification group, said data processing means performing steps b) to d) according to any one of claims 1 to 9.

11. A computer program comprising program code means for performing the method according to any one of claims 1 to 9, when the program is run on a computer.

12. A computer program product comprising program code means stored on a computer readable medium for performing the method according to any one of claims 1 to 9, when the program is run on a computer.

Images