JP2008544248A - Affinity analysis method and system - Google Patents

Abstract

A method for determining the affinity of a molecular binding interaction from data measured for binding in a steady state,
a) preparing a plurality of experimentally obtained data sets for binding for the interaction of two chemical species binding, wherein each data set includes a plurality of respective association phases and dissociation phases of the interaction; Contains data on binding measured at the time,
b) from a plurality of experimentally obtained binding data sets, selecting a plurality of binding data measured at a predetermined time point at or near the end of the association phase as representing steady state binding data;
c) perform quality control for each data set, including the process of estimating the reliability of the steady state combined data by evaluating other combined data of the same data set;
d) In step c), exclude each data set estimated to contain unreliable steady state combined data;
e) A method comprising determining affinity from steady state binding data in the remaining data sets.
[Selection] Figure 4

Description

  The present invention relates to a method for analyzing molecular binding interactions at a detection surface, and more particularly, a method for determining affinity for an interaction based on steady state binding data, at least in part. It relates to a method that can be automated. The present invention provides an analysis system including means for such automated affinity evaluation, as well as a computer program implementing the method, a computer program product including program code means for performing the method, and a computer comprising the program It also relates to the system.

  More and more attention is being focused on analytical detection systems that can monitor the interaction of molecules, such as biomolecules, in real time. Such a system often uses an optical biosensor and is usually referred to as an interaction analysis sensor or a biospecific interaction analysis sensor. A typical example of such a biosensor is a BIACORE (registered trademark) measuring device sold by Biacore (Uppsala, Sweden), which is composed of a molecule in a sample and a molecular structure fixed on a detection surface. In detecting the interaction, surface plasmon resonance (SPR) is used. When the sample is passed through the detection surface, the way the binding proceeds directly reflects the rate at which the interaction occurs. As the buffer is run after injecting the sample, the detector response during this time reflects the dissociation rate of the complex on the surface. A typical output obtained with the BIACORE® system is a graph or curve in which the progress of the interaction between molecules is plotted against time, which is part of the association phase and part of the dissociation phase. Including. This binding curve is usually displayed on a computer screen and is often referred to as a “sensorgram”.

Thus, with the BIACORE® system (and similar sensor systems), a sample of a specific molecule (analyte) can be obtained without labeling and often without purification of the relevant substance. It is possible to measure interaction variables such as kinetic rate constants for binding (association) and dissociation in molecular interactions, as well as the affinity of interactions at the surface, as well as the presence and concentration in the medium Become. The association rate constant (k _a ) and dissociation rate constant (k _d ) are obtained from an interaction model that mathematically describes kinetic data obtained for various concentrations of the sample analyte in the form of differential equations. It can be obtained by fitting. Affinity (Affinity represented as an affinity constant K _A or dissociation constant K _D) can be calculated from association and dissociation rate constants. However, since it is often difficult to obtain definitive kinetic data, it is usually assumed that affinity has been reached at or near the end of the association phase of the binding interaction for a range of analyte concentrations. It is more reliable to measure by equilibrium binding analysis including the process of measuring the equilibrium state, ie the level of binding in the steady state. To confirm that the association phase of the binding curve has indeed reached steady state, the time of sample injection required for the bound analyte to reach equilibrium under the conditions to be used for affinity measurements. It is customary to determine in advance the length of the sample (ie, the contact time of the sample with the sensor chip surface). Since both the time required to reach the equilibrium state and the time required for the analyte to dissociate are primarily determined by the dissociation rate constant, the approximate injection time can also be estimated from the dissociation constant. can do.

  Prior to evaluating sensorgrams for kinetic enemy properties and / or affinity, operators typically perform quality control to account for equipment-related obstacles, such as air spikes caused by air bubbles entering the fluid stream. Discard sensorgrams that are ineligible for.

  Evaluation of reaction kinetics and affinity data obtained with BIACORE® and similar instruments usually uses dedicated software, but measurement related obstacles, especially particles in the sample In the process of curve fitting, which identifies and excludes joint curves that do not fit well, for example, the operator usually had to intervene.

  The demand for throughput and information density when performing analysis continues to increase, and the burden on operators continues to increase. In response to these trends, automated procedures relating to the control of binding curve quality, such as those disclosed in WO 03/081245, and automated procedures relating to the evaluation of kinetic data, such as The procedure disclosed in 2005/029077 has been developed.

However, similar automated procedures are still needed for steady state affinity data, especially for cases where large data sets for interactions, such as sensorgrams, are generated. An important problem that must be solved in these cases is to make sure that the resulting binding curve is at least substantially in equilibrium before reading the binding level.
International Publication No. 03/081245 Pamphlet International Publication No. 2005/029077 Pamphlet

  An object of the present invention is to improve the process of determining affinity for the interaction of binding molecules. The purpose of this is to find out that data other than the binding level data for the steady state in the data set (usually the binding curve) can be used to estimate or determine whether the steady state binding data set is reliable. Realized by the method based on. In other words, data from one domain (eg, dissociation phase) of a binding curve can be used to determine whether it makes sense to evaluate data from another domain (eg, steady state) of the same binding curve. That is. This method can be automated, at least in part, to automatically exclude unreliable data about steady state binding levels.

Therefore, in one aspect of the present invention, a method for determining the affinity of a molecular binding interaction from measured data for binding in a steady state, comprising:
a) For the interaction of two chemical (including biochemical) species binding, a plurality of data sets experimentally obtained for binding are prepared, where each data set includes an association phase and a dissociation phase of the interaction. Including data on binding measured at each of a plurality of time points,
b) from a plurality of experimentally obtained binding data sets, selecting a plurality of binding data measured at a predetermined time point at or near the end of the association phase as representing steady state binding data;
c) perform quality control for each data set, including the process of estimating the reliability of the steady state combined data by evaluating other combined data of the same data set;
d) In step c), exclude each data set estimated to contain unreliable steady state combined data;
e) providing a method comprising determining affinity from binding data at steady state in the remaining data set;

In one form of this aspect, steps c) -e)
Select multiple non-steady-state binding data from experimental data sets for multiple data sets,
Determining which non-steady-state binding data is within a predetermined range, the predetermined range representing the quality of the data in the experimental binding data set;
Exclude poor quality experimental binding datasets from the final remaining experimental binding datasets, where poor quality experimental binding datasets are non-steady state binding data And determining the affinity from the steady state binding data of the plurality of experimental binding data sets that ultimately remain.

  In another aspect of the present invention, there is provided an analytical system for intermolecular interaction studies comprising data processing means for performing the above method.

  In yet another aspect of the present invention, a computer program comprising program code means for performing the above method is provided.

  In yet another aspect of the present invention, there is provided program code means for performing the above method, wherein the program code means is stored on a computer readable medium or carried in an electrical or optical signal. A computer program product is provided.

  In still another aspect of the present invention, there is provided a computer system equipped with a computer program comprising program code means for carrying out the above method.

  For a more complete understanding of the present invention and for further understanding of the features and advantages of the present invention, reference should be made to the following detailed description and drawings.

  Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the areas relevant to the present invention. In addition, even when written in the singular using definite articles and indefinite articles, unless otherwise specified, includes plural references.

  All references, including publications, patent applications, and patents, are hereby incorporated by reference in their entirety.

  As described above, the present invention evaluates the binding data for the steady state of the molecular binding interaction among a plurality of data sets regarding the interaction between the molecules, and determines the affinity of the interaction. In this process, the interaction data of the dataset other than the binding data for the steady state is used to estimate the reliability of the binding data for the steady state. The binding data obtained in the experiment is usually obtained by a technique that mainly uses a sensor that investigates the interaction between molecules. In this technique, the interaction between molecules is examined, and the result is expressed as the progress of the interaction. Present in real time at the same time. However, before describing the present invention in further detail, the general context in which the present invention will be used will be described.

Chemical sensors or biosensors are usually based on methods that detect changes in sensor surface properties, such as mass, refractive index, fixed layer thickness, and do not use labels. However, some sensors rely on signs. Typical sensor detection methods include mass detection methods such as optics, thermo-optics, and piezoelectric, or acoustic wave methods (eg, surface acoustic wave (SAW) and quartz crystal microbalance (QCM) methods) and electrochemistry. Examples of such methods include, but are not limited to, potentiometric measurement, conductivity measurement, current measurement, and capacitance / impedance method. With respect to optical detection methods, typical methods utilized mass surface concentration detection methods such as reflection, including both external and internal reflection methods that resolve in terms of angle, wavelength, deflection, or phase. Examples of such methods include evanescent wave ellipsometry and evanescent wave spectroscopy (EWS, or internal reflection spectroscopy) (both of which are evanescent fields via surface plasmon resonance (SPR)). Enhancement), Brewster angle refractometry, critical angle refractometry, attenuated total reflection (FTR), scattered total internal reflection (STIR) (can include scattering enhancement labels), optical waveguide sensor , External reflection photography, photography based on evanescent wave, for example, critical angle resolution photography, Brewster angle resolution photography, SPR angle resolution photography Etc. can be mentioned. Also, for example, surface-enhanced Raman spectroscopy (SERS), surface-enhanced resonance Raman spectroscopy (SERRS), evanescent wave fluorescence spectroscopy (TIRF), and phosphorescence-based photometry and imaging / microscopy itself, or The combination of these and the reflection method can be mentioned, and also includes a waveguide interferometer, a waveguide leakage mode spectroscopy, a reflection interference spectroscopy, a transmission interference method, a holography spectroscopy, and an atomic force microscope (AFR). be able to.
As a commercially available biosensor, the above-described Biacore, which can monitor in real time the binding interaction between the ligand bound to the surface and the target analyte, is a device utilizing surface plasmon resonance (SPR). Mention of BIACORE system equipment manufactured and sold by the company (Uppsala, Sweden). In this context, “ligand” refers to a known or unknown affinity molecule for a given analyte and can include, for example, any capture substance immobilized on a surface. On the other hand, examples of the “analytical substance” include any specific binding partner substance for this ligand.

  In the detailed description of the invention and the examples that follow, the invention is illustrated in the context of SPR spectroscopy, and more particularly with respect to the BIACORE® system, but the invention is limited to this detection method. Please understand that it is not. If the detection method uses affinity that binds the analyte to the ligand immobilized on the detection surface, it can measure changes in the detection surface that quantitatively suggest the binding of the analyte to the ligand immobilized on the detection surface. Any method can be used as long as it is a method.

  The surface plasmon resonance (SPR) phenomenon is well known, in which light is reflected at an interface between two media having different refractive indices under specific conditions, and this interface is coated with a metal film, usually silver or gold. It is sufficient to state that it occurs when In the Biacore instrument, these media are the sample and the glass of the sensor chip, which is in contact with the sample by a microfluidic flow system. The metal film is a thin gold layer provided on the surface of the chip. At a specific reflection angle, the intensity of the reflected light is reduced by SPR. The angle at which the intensity of the reflected light is minimized varies depending on the refractive index on the side opposite to the reflected light, that is, near the surface on the sample side in the BIACORE® system.

  A schematic diagram of the BIACORE® system is shown in FIG. The sensor chip 1 is provided with a gold film 2, and this gold film 2 carries a capturing molecule (ligand) 3, for example, an antibody. The capturing molecule (ligand) 3 is in contact with a flow of a sample containing a detection target substance 4, for example, an antigen, flowing through the flow channel 5. A p-polarized monochromatic light 6 emitted from a light source 7 (LED) is coupled by a prism 8 to a glass / metal interface 9 where the light is totally reflected. The intensity of the reflected light beam 10 is detected by an optical detection unit 11 (photodetector array).

  The technical aspects of BIACORE® equipment and the SPR phenomenon are described in detail in US Pat. No. 5,313,264. The matrix film provided on the detection surface of the biosensor is further described in detail in, for example, US Pat. Nos. 5,242,828 and 5,436,161. Further, the technical aspects of the biosensor chip used with the BIACORE (registered trademark) device are described in detail in US Pat. No. 5,492,840.

When molecules in the sample bind to the capture molecules on the sensor chip surface, the surface concentration, and thus the refractive index, changes and the SPR response is detected. Plotting this response against the time of interaction gives a quantitative measure of the progress of the interaction. Such plots, ie kinetic curves or binding curves (binding isotherms), are usually referred to as sensorgrams and are sometimes referred to in the art as “affinity trajectories” or “affinograms”. In the BIACORE (registered trademark) system, the response value of SPR is expressed in resonance units (RU). 1RU represents a change of 0.0001 ° of the angle at which the intensity of reflected light is minimized, and this change is about 1 pg / mm ² of sensor surface concentration for various biomolecules including most proteins. Corresponding to When a sample containing an analyte is brought into contact with the sensor surface, a capture molecule (ligand) bound to the sensor surface interacts with the analyte in a process called “association”. This process is indicated in the sensorgram by an increase in RU as the sample first contacts the sensor surface. Conversely, “dissociation” usually occurs when the sample stream is replaced by, for example, a buffer stream. This process is indicated in the sensorgram by the decrease in RU over time as the analyte is dissociated from the ligand bound to the surface.

  A typical sensorgram (binding curve) for the reversible interaction at the sensor chip surface is shown in FIG. On the detection surface, a capturing molecule, that is, a ligand, such as an antibody, is immobilized, and interacts with a binding partner molecule in the sample, that is, an analyte. Binding curves generated by biosensor systems based on other detection principles as described above should also have similar shapes. The vertical axis (Y axis) represents the response (here, the resonance unit RU), and the horizontal axis (X axis) represents the time (here, seconds). First, a buffer is passed over the detection surface to obtain a baseline response A of the sensorgram. During sample injection, an increase in signal due to analyte binding is observed. This B portion of the binding curve is usually referred to as the “association phase”. As a result, a steady state condition is reached at or near the end of the association phase, and the resonance signal plateau shown in C is reached (but this state is not always realized). It should be noted that the term “steady state” is used synonymously with “equilibrium” in the present invention (in other contexts “equilibrium” must be reserved to describe an ideal model of interaction). This is because, in actual situations, the binding may remain constant over time even though the system is not in equilibrium.) When the sample injection is finished, If the buffer is flowed continuously instead of the sample, the signal is reduced, and this reduction reflects the dissociation, ie, release, of the analyte from the surface. That is. This D portion of the binding curve is usually referred to as the “dissociation phase”. The analysis ends with a regeneration process in which the bound analyte can be removed from the surface and (ideally) a solution that retains the activity of the ligand is injected into the sensor surface. This process is shown in the E part of the sensorgram. When the buffer is injected, the baseline A returns and the surface returns to a state where a new analysis can be performed.

  The association phase B and dissociation phase D profiles provide information on the kinetic nature of the binding and dissociation, and the strength of the resonance signal at C is determined by the affinity (reciprocity associated with the change in surface mass concentration). Response as a result of action). This will be described in more detail below.

The reversible reaction between the surface binding rate reaction target substance A and the capture molecule (fixed) bound to the surface, that is, B, which is neither diffusion nor mass transfer is expressed by the following pseudo first order reaction kinetics: Is assumed to follow and follow this.

A + B⇔AB
In this interaction model (usually called Langmuir model), it is assumed that the analyte (A) is monovalent and homogeneous, the ligand (B) is homogeneous, and all binding events occur independently. In fact, this is true for most cases.

  The change rate of the surface concentration of the analysis target substance A during the injection of the analysis target substance A (= change rate of the concentration of the formed complex AB) is the sum of the inflow rate and the outflow rate of the analysis target substance A.

([A] is in the formula, the concentration of the analyte A, [B] the concentration of ligand B, [AB] is the concentration of the reaction complex AB, _{k a} is the association rate constant, _{k d} is the dissociation rate constant is there.)
After the elapse of time t, the concentration of unbound ligand B on the surface becomes [B _T ] − [AB] (where [B _T ] is the total or maximum concentration of ligand B). Substituting this into equation (1) gives the following equation:

Considering the response unit of the detector (AB is detected), this equation can be expressed as:

(Where R is the response at time t (resonance units (RU)), C is the initial or bulk concentration of the free analyte (A) in solution, and R _max Is the response (resonance unit (RU)) obtained when the analyte (A) binds to all the ligands (B) on the surface. By rearranging the equation (3), the following equation is obtained. It is done.

(Where R is the response (resonance unit (RU))). When integrated, the following equation is obtained.

Here K _a of the _{k d} calculated, according to equation (4), the dR / dt, is plotted against the concentration R of bound analyte was, the gradient _{- (k a C + k d} ) , and the ordinate intercept Becomes k _a R _max C. Bulk concentration C is known, (e.g., the surface, large excess by saturated with analyte in) if R _max is Zumi determined, determine the association rate constant k _a and dissociation rate constant k _d in the calculation be able to. However, a simpler method is a method in which the fitting of the integral function (5) or the numerical calculation and fitting of the differential equation (4) are preferably performed using a computer program.

k _d can also be determined by the following method. That is, the dissociation rate can be expressed by the following equation:

Integrating this gives the following equation:

Here, R ₀ in the formula is a response at the beginning of the surface washing with the buffer at the beginning of the dissociation phase.

  Linearizing equation (6) gives the following equation:

By plotting ln [R / R ₀ ] against t, a straight line with a slope of −k _d is obtained. However, as a simpler method, the association rate constant k _d can be determined by fitting the exponential rate equation (7).

  In order to obtain a reliable reaction rate constant, the above analysis is usually repeated for several different analyte concentrations, and the above analysis is repeated for at least one different sensor surface ligand concentration. Is appropriate.

Affinity calculations affinity, association constant _K A _{= k} a / k _d or, (also referred to as equilibrium _constant) the dissociation constant is represented by _{K D = k d / k a} .

The association constant K _A can also be determined from the equation (3). When dR / dt = 0 in the equilibrium state, this equation becomes the following equation.

Where R _eq in the equation is the response of the detector at equilibrium. Since k _a / k _d = K _A , substituting this into equation (9) and rearranging gives the following equation.

When carrying out the binding reaction in a plurality of concentrations, K _A, the data can be a, obtained by non-linear curve fitting. Also, for example, if the kinetic data is unreliable or the association and dissociation proceed too quickly for accurate measurements, R _eq / C is plotted against R _eq And then the slope is -K _A (Scatchard plot).

  By rearranging the equation (10), the following equation is obtained.

Upon insertion of the _K A = 1 / _{K D} in equation (11), it is the following Shikigae.

Expression (12) is usually transformed into the following expression.

("Offset" in the equation is a compensation factor for the balanced movement of the baseline due to the refractive index error of the entire system.)
Equations (11) and (12) can be modified by introducing a steric interference factor n describing how many binding sites are blocked on average by one molecule of the analyte.

Software assisted analysis Software for analyzing kinetics and affinity data is commercially available. Thus, for example, evaluation of kinetics and affinity data generated by a BIACORE® instrument is typically performed with dedicated BIAevaluation software (Biacore, Uppsala, Sweden), with the remaining Calculate the differential rate equation and nonlinear regression using numerical integration to find the value for the variable that gives the closest fit that reduces and minimizes the sum of the squares of the differences, and the kinetics And fitting affinity variables. The “residual” is the difference between the calculated curve at each point and the experimental curve, and the square of the residual weights the deviation equally above and below the experimental curve. Used for. The sum of the squares of the residuals is expressed by equation (16).

(S in the equation is a sum of squares of residuals, r _f is a value obtained by fitting at a predetermined point, and r _x is a value obtained by an experiment at the same point.)
For example, for the molecular interactions described above, such software-assisted data analysis is performed after subtracting background noise, and the simple 1: 1 described above expressed by equations (5) and (7) above. The Langmuir coupling model is to be fitted to the measurement data.

Typically, the binding model is fitted simultaneously to multiple binding curves obtained at different analyte concentrations C (and / or different levels of direction induction R _max ). This method is called a "global fitting", based on the data of the sensorgram, the global fitting, single global k _a or k _d to determine whether to show a good fit for all data. The completed fitting result is graphically presented to the operator, and the fitted curve is superimposed on the original sensorgram curve. The proximity of the fitting is also indicated by a chi-square (χ ² ) value (hereinafter referred to as “chi-square”), and this chi-square (χ ² ) is a standard statistical value. For good fit, the value of the chi-square is the noise about the same RU ^2. If necessary, a “residual plot” can be generated that graphically shows how the experimental data deviates from the fitted curve. The operator then determines whether the fitting is good enough. If it is not good enough, the sensorgram or the sensorgrams showing the worst fitting are excluded and the fitting process is performed again with the set of sensorgrams after exclusion. This process is repeated until a good fitting is obtained.

The process of determining the affinity constant from the measured steady state binding level using BIAevaluation software is as follows:
(I) Obtain the steady state binding level (Req) from the sensorgram reporting point located in the steady state region of the curve;
(Ii) plot Req against C, and
(Iii) fitting this plot to a general “steady state affinity” fitting model (eg, equation (13 or 14)) to obtain K _A / K _D and R _max .

The present invention as described above, the present invention is the combined data of a plurality of steady-state measured for binding, relates to a method determining the affinity (K _A or K _D) of molecular binding interactions, the method, It can be automated to collect experimentally obtained data sets for multiple different interactions and determine the affinity of each of the different analytes with minimal operator effort. The term “data set” as used herein refers to data representing a binding curve, eg, a sensorgram. The term “batch” of data sets, as used herein, includes two or more data sets.

  What is required in the process of such automated affinity assessment is the automatic exclusion of unreliable steady state data, i.e., binding interactions are substantially reduced when binding level data is read, for example. It excludes data that has not reached equilibrium or has some other problem in the steady state part of the binding curve.

  Steady state binding data is usually read at the reporting point at or near the end of the meeting phase. Such a “report point” is actually a very short interval of about 5 seconds, and the response values detected for this interval are averaged. At this stage of the binding curve, it should be sufficient to examine the slope of this binding curve region (which should be flat if in equilibrium) to determine if an equilibrium has been reached. However, since the reporting point is a short interval, even a combined curve that has reached an equilibrium state due to detection noise usually causes a gradient at the reporting point. Increasing the reporting point section does not improve the situation significantly.

  In the present invention, the reliability of the steady state data of the binding curve is estimated based on the binding data obtained from other parts of the binding curve, that is, the domain. From such combined data, various “reliability indicators” can be estimated.

An example of such a reliability indicator can first be the dissociation rate, which is used in determining whether the interaction has reached equilibrium in the association phase under the experimental conditions used. be able to. (To consider that equilibrium has been substantially reached, the curve slope of the equilibrium detection interval is usually less than about 1% / second, preferably about 1 ‰ / second (for an R _max of about 30 RU). Less than about 0.03 RU / sec). That is, if the dissociation rate is too low, the contact time of the sample with the detection surface may be insufficient to reach an equilibrium state. More specifically, for BIACORE® and similar systems where the sample passes through the sensing surface as a fluid flow, the length of each sample injection time (ie, the contact time of the sample with the sensing surface) is dissociated. There is a rate limit (this limit is substantially independent of the association rate) below which the affinity (K _A or K _D ) is calculated reliably with steady state analysis. I found something I couldn't do. The limit of the dissociation rate is preferably determined empirically by, for example, an actual test or computer simulation described later.

FIG. 3 shows 12 different diagrams (plots) in which the curve fitting error for affinity is plotted on a log (dissociation rate) vs. log (association rate) map at different injection lengths. These figures, MATLAB (TM) 6.5 by (MathWorks Inc. (Natick, Massachusetts)), (1) various at _{R max} of about _k a ⁼ 10 4 _~10 ⁸ and _k d = 10 Simulate sensorgrams at a series of concentrations for ^-1 to ^10-4 (using equation 5 and random noise above), (2) extract reporting points from each sensorgram, and (3) report the value of the point fitting to an affinity model steady state (12 and non-linear regression described above), and (4) those obtained by plotting an error of curve fitting for affinity in k _a -k _d map It is. This procedure was repeated for 12 different injection lengths.

In each plot, the area of the right dark diet, indicates that the error is relatively small in the fitted K _D, pale areas on the left, indicates that the error is relatively large in the fitted K _D, these These two regions are separated by a relatively thin gray border region. “Injlen 20s” or the like of each plot information indicates the injection length.

As can be seen from the plot, for each injection length, made of whether the limits can be calculated K _D dissociation rate exists. As can also be seen from the plot, this is not necessarily true when the association rate is high. However, it should be noted that the simulation was performed with a 1: 1 model without mass transport limitations. It should be noted that the time required to reach equilibrium is longer at higher association rates in the presence of mass transport limitations.

Based on empirical values for each plot, the various K _D, (specifically polynomial) equation can be calculated minimum dissociation rate used.

  The binding data for determining the dissociation rate is usually taken from the dissociation phase, but may be taken from the association phase or from both domains of the binding curve. To automatically exclude (by software) sensorgrams that are unlikely to reach equilibrium, set a limit on the dissociation rate determined by the injection length, and the estimated dissociation rate is the set value for the used injection length. Lower sensorgrams may be automatically excluded from the data set used for affinity calculations.

  A second reliability indicator that can basically be determined from a specific domain of the binding curve that is different from the reporting point interval is the last part of the sensorgram, for example the last 30 seconds of the meeting phase. Can be mentioned. If the slope of this curve (including the reporting point section) is downward (negative), something is wrong and the sensorgram should be excluded from the sensorgram used to determine affinity. It becomes an indicator of being.

If necessary, other indicators of reliability can also be used in combination with the two above-mentioned indicators. These other confidence indicators may relate to multiple curves (eg, like a sensorgram at a series of concentrations), even if they relate to a single curve (eg, like the reliability factor described above). For example, one or more of the following may be included.
・ There are too few curves with signal (reporting point) higher than noise level ・ Graph signal (reporting point) vs. concentration is not a constant tone ・ Insufficient difference between highest signal and lowest signal ・The concentration range is too narrow. • The number of curves that can be used for analysis is too small.

  After excluding unreliable batches of the data set, the “affinity model” is fitted to all qualified batches as is known in the art.

  Preferably, the fitting quality, the reliability of the affinity calculation, and optionally other quality measurements are calculated before presenting the results of the affinity evaluation to the operator.

An example of a flowchart for evaluation using software is shown in FIG. Although this flowchart is considered to be generally applicable, the following describes a case where the flowchart is used with a BIACORE® system (eg, a BIACORE® T100 system or a BIACORE A100 system). It is assumed that the user has generated a fairly large number of sensorgrams (eg, about 300-2000 per 24 hours) for multiple (eg, about 30-200) analyte-ligand interactions (see above). Thus, as used herein, “ligand” refers to an interactant specific for the analyte to be immobilized on the sensor surface). For each analyte and ligand pair, measurements are performed using different analyte concentrations, and multiple sensorgrams obtained with these series of concentrations are referred to below as a batch of sensorgrams. " Performance of affinity assays that the optimum is at least six as the concentration of analyte, for example using a series of concentrations that include the 10 kinds, equilibrium is reached in all these levels, the highest concentration, equilibrated with R _max This is the case when the state is reached and the minimum concentration Req is lower than 0.2 × Rmax. (Alternatively, instead of using a single ligand density and several different analyte concentrations, several surface areas with different fixed ligand densities (surface concentrations) and a single analyte concentration It can also be used.)
Prior to assessing for affinity, the binding data has already been subtracted with respect to the criteria, corrected for bulk effects, and curve quality control, e.g. as described in WO 03/081245 above. It is assumed that nicked sensorgrams are automatically removed. It is also preferable to use manual quality control for manual tests, and to present the operator with a plot of the reporting points so that the operator can make control judgments, carryover, summarization to a reference, etc. .

  In a first step 401, the software presents to the user the analyte to be analyzed and the immobilized ligand (“spot”), and the user then sets the data set to be analyzed, ie the sensorgram. Select a batch of

  In the second step 402, the software extracts data for affinity analysis for each analyte / ligand combination, and for the baseline, association phase, and dissociation phase, the flow rate and “injection stop”. Prepare data for analysis by defining by event). The reporting point “binding late” is used to determine affinity, and sensorgram data is used for quality control.

In the third step 403, the software prepares for each batch by applying a first set of “rules” predefined for each sensorgram as well as the entire batch based on various reliability indicators. Perform a first quality control on the data. The following two rules are particularly useful in evaluating steady state combined data.
(1a) Exclusion of all curves whose association phase has a negative slope.
(1b) For each curve, compare the dissociation rate estimate to a preset limit for this estimated value for the injection length used (this limit can be defined as described above). Do. As a result, a single curve, or (usually) the entire batch is excluded, or a “penalty” is added to the batch, as described below.

  The determination of the dissociation rate is usually (i) cutting out the dissociation phase, (ii) fitting the dissociation rate, and if the fitting is successful, (iii) whether the obtained dissociation rate is within a preset allowable range This is done by determining whether or not. If it is not acceptable, a penalty is added to the batch containing the curve.

The first rule set may further include the following.
(1c) A batch with too few curves with a signal above the noise level (binding late at the reporting point) gives a penalty.
(1d) Penalizes the batch if the signal in the graph (the combined delay at the reporting point) is not a constant tone.
(1e) If the difference between the highest signal and the lowest signal is insufficient, a penalty is given to the batch.
(1f) If the concentration range is too narrow, a penalty is given to the batch.
(1 g) If the number of curves that can be used for analysis is too small, a penalty is given to the batch.

For example, the following four batch penalty levels can be used.
3-Request removal of batch 2-Request user inspection 1-Suspected 0-No problem detected.

  Batches without penalty or batches with a penalty of 1 can be used for analysis as they are. A batch is automatically excluded if the batch has a penalty of 3 for any point or the sum of all the penalty of the batch is 5 or more. If the total penalty is 2, 3, or 4, mark the user for inspection.

  In a fourth step 404, the software fits an affinity model (eg, the “1: 1 + Offset” model described above) by non-linear regression as described above for all qualified batches.

In the fifth step 405, a cross check, preferably “leave-one out cross check” (see, eg, Wold S., Technometrics, 20 (1978) 397-406) is performed (i ) Consider whether out-of-line values are included, and (ii) obtain an estimate for the variability of the calculated parameter (χ ² , standard deviation)
If χ2 is significantly improved by removing one curve in the batch (approximately 99% confidence level), the curve is removed in step 405a. Only one curve need be removed from each batch. If there is more than one out-of-line value, a penalty is given to the batch. If the curve is removed in step 405a, the first set of rules in the third step 403 is re-evaluated in step 405b, the affinity model is re-fitted to the data in step 405c, and cross-validation is again performed in step 405d. Calculate, but do not remove curves.

In the sixth step 406, the second rule set is applied to the data obtained in step 405 or 405c. This rule can include the following:
(2a) If the standard deviation of the calculated _{K D} in cross-validation process 405 or 405c is too large, a penalty to the batch.
The standard deviation of the calculated _{K D} in cross-validation of (2b) process 405 or 405c is too large, if χ2 is too large, a penalty to the batch.

In step 406, a third rule set can be applied to check if the resulting variable is a valid value. This rule can include the following:
(3a) _Penalize the batch if R _max obtained in step 404 is significantly higher than the highest signal in the batch.
(3b) If the offset term obtained in step 404 is larger than the noise level and the average signal level, a penalty is given to the batch.

  In the seventh step, all penalties for each batch are collected, and the batch is scored for exclusion, inspection, or qualification based on the same penalty criteria as described above. For all batches, all batches that have been graded to be excluded are clearly marked out, and batches that have been graded to be inspected are clearly marked. A qualified batch performs a qualified operation.

  In an eighth step 408, a batch of curves is presented to the user. In step 408a, the user can also modify the batch by returning one or more previously excluded batches as needed. Also, the user can change the initial value of the fitting process by taking a curve into or removing it from the batch. If such a batch has been modified, the affinity model is re-fitted to the modified batch in step 408b.

  In a ninth step 409, the final result is displayed and the user can “end and lock” the affinity evaluation.

  The invention also relates to a computer program, in particular a computer program on or in the carrier, suitable for carrying out the method of the invention. A carrier can be any entity or device capable of carrying a program. For example, the carrier can be a storage medium, such as a ROM, CD ROM, DVD, or semiconductor ROM, or a magnetic recording medium, such as a floppy disk or hard disk. The carrier can also be a movable carrier, for example an electrical or optical signal that can be transported via an electrical or optical cable, or by radio or other means. The carrier can also be an integrated circuit in which a program is incorporated. Any known or newly developed medium capable of storing information suitable for use with a computer system can be used.

  It should be understood that the invention is not limited by the specific forms of the invention described above, and the scope of the invention is defined by the appended claims.

This figure is a schematic diagram of a biosensor system based on surface plasmon resonance (SPR). This figure is a representative sensorgram, where the binding curve has distinct association and dissociation phases. This figure shows 12 different plots of curve fitting errors for affinity plotted in a log (dissociation rate constant) vs. log (association rate constant) map for each case with varying sample injection times. . This figure is a flowchart showing one form of affinity determination of the present invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Sensor chip 2 ... Gold film 3 ... Capture molecule (ligand) 3
4 ... Detection target material 5 ... Flow channel 6 ... Monochromatic light 7 ... Light source 8 ... Prism 9 ... Glass / metal interface 10 ... Reflected light beam 11 ... Optical detection unit

Claims (31)

A method for determining the affinity of a molecular binding interaction from data measured for binding in a steady state,
a) preparing a plurality of experimentally obtained data sets for binding for the interaction of two chemical species binding, wherein each data set includes a plurality of respective association phases and dissociation phases of the interaction; Contains data on binding measured at the time,
b) from a plurality of experimentally obtained binding data sets, selecting a plurality of binding data measured at a predetermined time point at or near the end of the association phase as representing steady state binding data;
c) perform quality control for each data set, including the process of estimating the reliability of the steady state combined data by evaluating other combined data of the same data set;
d) In step c), exclude each data set estimated to contain unreliable steady state combined data;
e) A method comprising determining affinity from steady state binding data in the remaining data sets. Processes c) -e)
Select multiple non-steady-state binding data from experimental data sets for multiple data sets,
Determining which non-steady-state binding data is within a predetermined range, the predetermined range representing the quality of the data in the experimental binding data set;
Exclude poor quality experimental binding datasets from the final remaining experimental binding datasets, where poor quality experimental binding datasets are non-steady state binding data And determining the affinity from the steady state binding data of the plurality of experimental binding data sets that ultimately remain. 3. A method according to claim 1 or claim 2, wherein one of the interacting chemical species is immobilized on a solid support and the other chemical species is in solution. 4. A method according to any one of claims 1 to 3, wherein step c) according to claim 1 comprises the step of estimating at least one reliability indicator. 5. The method of claim 4, wherein the at least one reliability indicator is a dissociation rate, and a dissociation rate lower than a predetermined value indicates that the steady state binding data is unreliable. 6. The method of claim 5, wherein the predetermined value of the dissociation rate is determined from a pre-established relationship between the length of time of the association phase for the interaction to reach a substantially steady state and the minimum dissociation rate. 7. The method of claim 6, wherein the relationship between the length of time of the association phase for the interaction to reach a substantially steady state and the minimum dissociation rate is based on empirical data about the interaction. 8. The method of claim 7, wherein the empirical data for the interaction is computer simulated data for the interaction. The method according to claim 5, wherein the dissociation rate is estimated from the binding data of the dissociation phase. 5. The at least one reliability indicator is a slope of the binding curve represented by association phase binding data, and the presence of a negative slope indicates that the steady state binding data is less reliable. A method according to any one of claims 9 to 9. 11. A method according to any one of claims 1 to 10, wherein the affinity determination comprises fitting a steady state affinity model. 8. A method according to any one of claims 3 to 7, wherein a plurality of data sets comprises binding data for several, preferably 6 or more, different concentrations in the solution of the chemical species. The step c) of claim 1 includes the step of applying a first rule set of predefined quality to the data set to grade the data set or batch of data sets for data reliability. The method according to claim 1. 14. The method of claim 13, wherein the predefined quality rules include at least one rule that can exclude at least one data set. 15. A method according to claim 13 or claim 14, wherein the quality rules comprise at least one rule that penalizes a batch of data sets. 16. The method of claim 15, wherein a batch of datasets is excluded when the total penalty exceeds a predetermined value. 17. A method according to any one of claims 11 to 16, wherein the quality of fitting to a steady state affinity model is determined. 18. Method according to claim 17, wherein the determination of the quality of the fitting comprises a process of cross validation, preferably a single point exclusion cross validation. Based on the determined fitting quality, modify the batch of datasets by excluding specific datasets in the batch, and then reapply the first rule set for the predefined quality to the modified batch. 19. A method according to claim 17 or claim 18, wherein the steady state affinity model is refit into the batch. 20. A method according to any one of claims 13 to 19, wherein the second rule set for quality is matched to the affinity determined in step e) of claim 1. 21. The method of claim 20, wherein the second rule set for quality includes at least one rule that penalizes a batch of data sets. 24. The method of claim 21, wherein when the penalty sum exceeds a predetermined value, the batch is excluded and the steady state affinity model is reapplied to the batch. 23. The method according to any one of claims 1 to 22, wherein data obtained experimentally for binding is determined with a biosensor. 24. The method according to claim 23, wherein the biosensor is based on evanescent wave detection, in particular surface plasmon resonance (SPR). 25. A method according to any one of claims 1 to 24, wherein the method is implemented on a computer. (I) detection data including a plurality of binding curves each representing at least one detection surface, a detection means for detecting a molecular binding interaction on the at least one detection surface, and a time course of the binding interaction; A sensor device comprising means for generating;
(Ii) an analytical system for detecting a molecular binding interaction comprising data processing means for carrying out steps a) to e) according to claim 1 defined in any one of claims 1 to 24 . 25. A computer program comprising program code means for performing affinity determination according to any one of claims 1 to 24 when the program is run on a computer. 25. Program code means for carrying out affinity determination according to any one of claims 1 to 24 when the program is run on a computer, stored in a computer readable medium, electrical or optical A computer program product comprising program code means carried in a dynamic signal. 26. A computer system including a program for executing affinity determination according to any one of claims 1 to 25 when the program is run on a computer. Preparing a data set including data on binding interactions between two species, including data representing binding curves for the interactions;
To determine the affinity of the interaction, select the first domain of the binding curve;
Selecting data from the second domain of the binding curve to estimate the reliability of the data obtained from the first domain;
An affinity analysis method comprising a step of using this data in determining the affinity of a binding interaction when it is estimated that the data obtained from the first domain is reliable. In determining the affinity for the molecular binding interaction between the first species and the second species,
Prepare a solid support surface to which the second chemical species is fixed,
Contacting the fluid containing the first chemical species with the solid support surface for a predetermined contact time;
Determine the association of the first species with the surface and obtain data about the association;
Contacting the solid support surface with a fluid not containing the first chemical species, determining dissociation from the surface and obtaining data on dissociation,
Based on a predetermined relationship between the contact time to reach equilibrium and the minimum dissociation rate, estimate from the dissociation data whether the interaction has substantially reached equilibrium within the predetermined contact time And
An affinity determination method that uses association data to determine the affinity value of an interaction when an equilibrium state is substantially reached.