METHODS AND SYSTEMS FOR ANALYSING HYBRIDISATION

Abstract

A method is described for the analysis of hybridisation between a target in solution and a probe bound at a surface. The method comprises receiving detection intensity results for hybridisation of the target with a plurality of different probes, the probes being selected so that a range of hybridisation detection intensity results for the hybridisation between the target and the probe is covered. The method further comprises analysing the detection intensity results as function of the hybridisation free energy. According to embodiments of the present invention, the receiving and/or analysing takes into consideration a thermodynamic non-equilibrium state for the target-probe bounding state.

Claims

1. A method for determining in a sample solution the presence or concentration of a target nucleic acid out of a set of possible target nucleic acids, wherein each of said possible target nucleic acids differ from each other by one or two nucleotides, the method comprising contacting said sample solution with a plurality of different probes bound at a surface and receiving detection intensity results for hybridisation of the possible target nucleic acids and each of said plurality of different probes, wherein for each possible target nucleic acid, at least one probe comprises a strand of nucleic acid that is perfectly complementary with said possible target nucleic acid and at least one other probe comprises a strand of nucleic acid that is non-perfectly complementary with said possible target nucleic acid so that a range of hybridisation detection intensity results for the hybridisation between each possible target nucleic acid and the probe is covered, for each possible target nucleic acid performing the analysis of the detection intensity results in a processor of the computing system as a function of hybridisation free energy, wherein the analysis comprises: determining the hybridisation free energy for the hybridisation between a possible target nucleic acid in solution and a probe bound at a surface based on a three state model wherein a bound state, an unbound state and a thermodynamic non-equilibrium state for the hybridisation are taken into account; analysing the logarithm of the detection intensity results as a function of the difference between the hybridisation free energy of said hybridization with respect to the hybridisation free energy for the perfectly complementary probe of the possible target nucleic acid; and deriving the presence and/or concentration of a target nucleic acid in said sample by determining whether said relationship between the logarithm of the detection intensity results and the hybridisation free energy corresponds to a predetermined relationship between the logarithm of the detection intensity results and the hybridisation free energy between the target and said probe.

2. The method of claim 1, wherein for said receiving detection intensity results or for said analysis reaching of thermodynamic equilibrium is taken into consideration.

3. The method of claim 2, wherein said receiving detection intensity results taking into consideration reaching of thermodynamic equilibrium comprises receiving detection intensity results for hybridisation of the target nucleic acid with a plurality of different probes obtained under hybridisation conditions wherein thermodynamic equilibrium has been reached.

4. The method of claim 3, wherein said hybridisation conditions comprise one or a combination of hybridisation time, probe length or temperature.

5. The method of claim 1, wherein said analysing the logarithm of the detection intensity results as function of hybridisation free energy comprises determining whether one linear relationship or more linear relationships can be distinguished between parts of the logarithm of the detection intensity results and the hybridisation free energy and deriving that the hybridisation has not reached equilibrium when more than one linear relationship can be distinguished between parts of the logarithm of the detection intensity results and the hybridisation free energy.

6. The method of claim 1, wherein said determining the hybridisation free energy for the hybridisation between a target nucleic acid in solution and a probe bound at a surface is based on a nearest-neighbour model.

7. The method of claim 6, wherein for each possible target nucleic acid a perfect match probe, and probes with up to two nucleotide non-complementary elements are provided, each to separate microarray spots for interaction with the target nucleic acid during the hybridisation, and analysing the detection intensity results as function of the hybridisation free energy comprises determining a set of free energy parameters of a nearest neighbour model for the hybridisation free energy based on said hybridisation microarray experiment.

8. The method of claim 1, wherein analysing the detection intensity results as function of the hybridisation free energy comprises determining whether measured detection intensity results correspond with thermodynamic equilibrium for the hybridisation.

9. The method of claim 1, wherein the actual target nucleic acid is a minority target and wherein the set of known possible targets comprises, besides the minority target, also a main target differing from the minority target in one or two non-complementary elements.

10. The method of claim 1 for detecting single nucleotide polymorphisms, wherein said analysing comprises determining whether one linear relationship or more linear relationships can be distinguished between parts of the logarithm of the detection intensity results and the hybridisation free energy and identifying probes corresponding with intensity values deviating from the one or more linear relationships between parts of the logarithm of the detection intensity results and the hybridisation free energy, said intensity values deviating by at least more than 5% from the one or more linear relationships.

11. The method of claim 1, wherein hybridisation is performed in a microarray.

12. The method of claim 1, wherein the detection intensity results for hybridisation are induced by emission of a label associated with a hybrid formed by binding of the target and the probe.

13. A computing system for determining in a sample solution the presence or concentration of a target nucleic acid out of a set of possible target nucleic acids, the system comprising a receiving means of a computer system adapted for receiving detection intensity results for hybridisation of a target nucleic acid with a plurality of different probes, wherein the probes are nucleic acids selected so that at least one probe comprises a strand of nucleic acid that is perfectly complementary with a strand of nucleic acid of a possible target nucleic acid and at least one other probe comprises a strand of nucleic acid that is non-perfectly complementary with a strand of nucleic acid of a possible target nucleic acid so that a range of hybridisation detection intensity results for the hybridisation between each possible target nucleic acid and the probe is covered, and an analysing means adapted for (i) analysing for each possible target the detection intensity results as function of the hybridisation free energy, wherein the analysing means is a processor adapted for determining the hybridisation free energy for the hybridisation between a target in solution and a probe bound at a surface based on a three state model, wherein a bound state, an unbound state and a thermodynamic non-equilibrium state for the hybridisation is taken into account, (ii) for analysing the logarithm of the detection intensity results as a function of the difference between the hybridisation free energy with respect to the perfect match free energy for the possible target nucleic acid for each of the possible target nucleic acids for a range of hybridisation free energies; and (iii) for detecting a predetermined relationship between the detection intensity results and the hybridisation free energy between the probe and a possible target and for deriving the presence or concentration of a target nucleic acid in a sample solution therefrom.

14. The system of claim 13, wherein the system is implemented as a computer program product.

15. The method of claim 9, wherein the analysis of the detection intensity results in a processor of the computing system as function of hybridisation free energy is performed for the main target, with respect to the perfect match free energy for the main target; and wherein a deviation from the predetermined relation between the detection intensity results and the hybridisation free energy indicate the presence of the minority target within the detection limit.

16. The method of claim 15, further comprising the step of estimating the concentration of the minority target based on previously performed calibration experiments or by calculation based on an extended equation correlating the logarithm of the intensity with the hybridisation free energy.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0035] FIG. 1 illustrates a graph expressing the measured intensities in a hybridisation experiment using a plurality of different probes bound to a surface and a target as function of the hybridisation free energy whereby thermodynamic equilibrium has not been reached for all probe spots, as can be used in an embodiment according to the present invention.

[0036] FIG. 2 illustrates a graph as in FIG. 1, whereby thermodynamic equilibrium has been reached for all probe spots, as can be used in an embodiment according to the present invention.

[0037] FIG. 3A and FIG. 3B shows an example of a graph of detection intensity results as function of the hybridisation free energy for detection of a target from a known set of targets, whereby FIG. 3A illustrates the results when the supposed target used for the graph is the actual target that was in the sample, whereas FIG. 3B illustrates the results when the supposed target used for the graph is not the actual target that was in the sample, as can be obtained in an embodiment according to the present invention.

[0038] FIGS. 4A-4F illustrate graphs for analysis in a method for detecting a minority target differing a single nucleotide from a main target, as can be obtained in an embodiment according to the present invention.

[0039] FIG. 5 shows a flow chart for an exemplary method for performing hybridisation according to an embodiment of the present invention.

[0040] FIGS. 6A and 6B indicate correlation plots for the total intensities (a) and the median intensities (b) in replicated hybridisation experiments of a first particular example illustrating features of embodiments according to the present invention.

[0041] FIGS. 7A and 7B illustrate a plot of intensities (a) as function of G.sub.sol for different concentrations and a plot of predicted behaviour (b) based on the Langmuir model for a first particular example, illustrating features of embodiments according to the present invention.

[0042] FIG. 8 shows ratios of intensities and perfect match intensities as function of G.sub.array for a first particular example, illustrating features of embodiments according to the present invention.

[0043] FIGS. 9A-9D show plots of I/I*.sub.pm as function of the nearest neighbour fitted G.sub.array for a first particular example, illustrating features of embodiments according to the present invention.

[0044] FIG. 10 shows a comparison of data in tables 4 and 5 between G.sub.sol and G.sub.array for a first particular example, illustrating features of embodiments according to the present invention.

[0045] FIG. 11 shows a plot of the intensity divided by concentration as function of G for particular experiments of a second particular example with a different target concentration, illustrating features of embodiments according to the present invention.

[0046] FIG. 12 illustrates the three state model for hybridisation in DNA microarrays, as can be used in embodiments according to the present invention.

[0047] FIG. 13 illustrates a numerical solution of a fraction of hybridized probes for the three state model for a second particular example of hybridisation, illustrating features of embodiments according to the present invention.

[0048] FIGS. 14A-14D show a plot of the intensity as function of G for particular experiments of a second particular example at different hybridisation times, illustrating features of embodiments according to the present invention.

[0049] FIGS. 15A-15D show a plot as shown in FIGS. 14A-14D for hybridisation with a shorter target sequence, illustrating features of embodiments according to the present invention. Table 1 illustrates the oligos used as target in the four different hybridisation experiments of a first particular example illustrating features of embodiments according to the present invention.

[0050] Table 2 illustrates the design of the probe set used in a first particular example illustrating features of embodiments according to the present invention.

[0051] Table 3 illustrates the target conditions per microarray used in experiments of the first particular example illustrating features of embodiments according to the present invention.

[0052] Table 4 illustrates free energy differences parameters obtained from fitting microarray data to an equation expressing the logarithmic ratios of the intensities with the perfect match intensities for the first particular example, illustrating features of embodiments according to the present invention.

[0053] Table 5 illustrates data as in table 4, using the nearest neighbour parameters obtained from melting experiments in solution for the first particular example, illustrating features of embodiments according to the present invention.

[0054] Table 6 illustrates targets and probe sequences used in a second particular example, illustrating features of embodiments according to the present invention.

[0055] The drawings are only schematic and are non-limiting. In the drawings, the size of some of the elements may be exaggerated and not drawn on scale for illustrative purposes. Any reference signs in the claims shall not be construed as limiting the scope. In the different drawings, the same reference signs refer to the same or analogous elements.

DETAILED DESCRIPTION OF ILLUSTRATIVE EMBODIMENTS

[0056] Where in embodiments of the present invention the term equilibrium is used, reference is made to thermodynamic equilibrium describing a situation wherein a steady state is obtained such that the number of conventional target-probe bindings does not substantially change over time. The term non-equilibrium or non-equilibrium effects is used to refer to occurrence of a target-probe binding state that may change over time. Where in embodiments of the present invention the term free energy is used, reference is made to the Gibbs free energy (G), referring to the thermodynamic potential that measures the useful energy obtainable from an isothermal isobaric thermodynamic system change. Where in embodiments analysis is performed as function of hybridisation free energy, this includes analysis as function of G being the free energy difference between a perfect matching hybridization and an hybridization where the probe sequences have one or more internal mismatches.

[0057] The term hybridisation used in embodiments according to the present invention refers to nucleic acid hybridisation. This refers to the process of establishing a non-covalent sequence-specific interaction between two or more complementary strands of nucleic acids into a single hybrid. The strands of nucleic acids that may bind to their complement can for example be oligonucleotides, DNA, RNA or PNA. Nucleotides form the basic components of the strands of nucleic acids. Hybridisation comprises binding of two perfectly complementary strands (in the Watson-Crick base-pairing senses), but also binding of non-perfect complementary strands. With a non-perfect complementary strand reference may be made to strands having a small number of non-complementary elements such as one, two or more non-complementary elements, preferably one or two non-complementary elements. In principle there is no limit to the number of non-complementary elements but the more non-complementary elements, the easier these are detectable and thus the less it is required to have dedicated methods for detecting these. In some applications the latter is un-wanted and referred to as cross-hybridisation. Non-perfect complementary strands can contain different types of non-complementary elements like e.g. mismatches or loops, or any local alteration of binding properties. Non-complementary elements thereby may have a small effect on the free energy. The current invention is not limited to a particular type, but for clarity the examples given further below deal with small number of mismatches. Such a small number of mismatches may e.g. include one, two or more nucleotides that are mismatched. Hybridisation may occur between strands that both are in solution, but in the present embodiments, the hybridisation envisaged is the interaction between a strand that initially is in a sample solution and a strand that is bound to a surface. Where in embodiments of the present invention reference is made to the term probe a substance is envisaged that allows detection or identification of another substance in a sample. The probe may for example be a strand of oligonucleotides, DNA, RNA or PNA (partially) complementary to the strand of interest in the sample solution, e.g. referred to as actual target or target. Object of interest that may be present in a sample solution and for which a check may be performed may be referred to as a possible target. In some embodiments according to the present invention, reference is made to a minority target and a main target, the minority target being a target different from the main target, e.g. being slightly different, and being present in a concentration substantially smaller than the main target. In hybridisation experiments according to the present invention, detection of hybridisation typically may be performed by a marker associated with the formed hybrid, such as for example a radio-active marker or a fluorescence marker, embodiments of the present invention not being limited thereto. Examples of typical fluorescence markers that may be used are Cy3 or Cy5 which are dyes of the cyanine dye family, the invention not being limited thereto. Typically in hybridisation experiments, intensity of the radiation or fluorescence provided by the markers is detected and representative for the number of hybrids formed.

[0058] Examples and embodiments of the present invention may be illustrated using a particular platform for microarrays, being the Agilent platform from Palo Alto. Nevertheless, it is to be noticed that the concepts and features set forth in embodiments according to the present invention are not limited to this particular platform but can be applied mutatis mutandis to other platforms, such as for example the GeneChips platform from Affymetrix or CodeLink Bioarray platform from Amersham Biosciences. Whereas the particular fitting parameters for the model used may be platform dependent, the principles and features of the methods set out in embodiments of the present invention can be applied to these and other platforms. Furthermore, embodiments of the present invention are not limited to microarrays, but may be applied for different hybridization measurements.

[0059] In a first aspect of the present invention, a method and system is described for analysing hybridisation between a target in a sample solution and a probe bound at a surface. The system and method may be especially suitable for analysing hybridisation in a microarray, although the invention is not limited thereto. Embodiments of the present invention may also be applied for or assist in purification applications, applications for selection of sequences out of a mixture, diagnostics, applications for detecting small mutations, applications for detecting viruses, etc. According to embodiments of the present invention, the method comprises receiving detection intensity results for hybridisation of the target with a plurality of probes, the probes being selected so that a range of hybridisation detection intensities for the hybridisation between the target and the probe is covered. The plurality of probes, and hybridisation may be tested or performed for each probe type in a single spot, e.g. in a microarray spot. Detection intensity results may be obtained based on label emission from a label associated with a hybride formed during hybridisation. The range of hybridisation detection intensities implies that a range of hybridisation free energies is covered. Receiving detection intensity results may comprise obtaining detection intensity results by actually measuring the intensity, thus including the detection or measurement step, or it may comprise receiving data via an input port for processing. By way of illustration, one example for obtaining detection results may encompass providing a plurality of different probes applied to different spots in a microarray. Each probe type thereby is bound to a surface, e.g. a microarray spot surface. Obtaining detection results furthermore may encompass providing a sample solution comprising the target to the surface to which the probes are linked and allowing the hybridisation process to take place. Such hybridisation may be platform dependent but may typically acquire at least 15 hours. As will be shown and discussed below, the hybridisation time used before measurement of the hybridisation, or factors related thereto such as the length of the probes or the temperature, may play an important role in obtaining accurate results. With the formation of hybrids, typically a marker may be associated. Such markers may for example allow optical detection of the hybridisation, although the invention is not limited thereto. Detection results thus may encompass emission intensities of markers associated with hybrids formed during hybridisation.

[0060] As indicated above, in thermodynamic equilibrium the hybridisation between a target initially in the sample solution and a probe bound to a surface is characterised by the hybridisation free energy (G). This hybridisation free energy G expresses the free energy difference between the bound probe-target state (double stranded) and its unbound state. A method for determining such hybridisation free energy (G) values in the case of hybridisation between a target initially in solution and a probe bound to a surface will be discussed below. The plurality of different probes being selected so that a range of hybridisation detection intensity results for the hybridisation between the target and the probe may comprise a probe being the perfect complement of the target one wants to detect and probes that are not perfectly complementary, for example having for example one or two nucleotides being mismatched to form the perfect complement of the target, showing loops, etc. Depending on the application envisaged, the plurality of different probes may comprise such probes for a single target envisaged, for a main target and a minority target, for a set of targets of which one target is to be selected, etc.

[0061] The method furthermore comprises analysing the measured or detected intensity as function of the hybridisation free energy. One exemplary way of analysing the measured or detected intensity as function of the hybridisation free energy is discussed with reference to FIG. 1, although embodiments of the present invention are not limited thereto.

[0062] Using the plurality of different probes as described above, the hybridisation experiment results in an intensity value per probe type. Evaluating the measured intensity as function of the hybridisation free energy, for example as function of the difference of hybridisation free energy with respect to the perfect match free energy for a possible target, allows for providing additional information regarding the hybridisation process itself and the hybrid, target and/or probe. As will be discussed later, such analysis may take into consideration reaching a thermodynamic equilibrium, for example, the analysis may take into consideration that the hybridisation is not in equilibrium yet. The analysis may comprise an analysis of the logarithm of the detection intensity results as function hybridisation free energy for a range of hybridisation free energies. The analysis may comprise determining whether one linear relationship or a deviation therefrom can be distinguished between parts of the logarithm of the detection intensity results and the hybridisation free energy. A deviation from a linear relationship may be a deviation of more than 5%, e.g. more than 10%, e.g. more than 25%, e.g. more than 33%. The analysis may comprise determining whether a linear relationship with slope 1/RT can be distinguished. From this analysis different conclusions may be drawn, such as for example the presence of a target, the presence of a minority target, identification of one target out of a set of targets, identification that equilibrium has been reached or not, etc.

[0063] FIG. 1 illustrates the logarithm of the measured intensities as function of the difference of hybridisation free energy with respect to the perfect match free energy for a possible target. As shown above in FIG. 1 some intensity measurements correspond with spots where the hybridisation is in thermodynamic equilibrium, while others are not. Furthermore, as illustrated in the applications, features such as equilibrium or not, presence of a target, concentration of target, identification of mutations, etc. may be derived in the analysis, as further illustrated under the different applications discussed below. As will be described later for a plurality of applications, the analysis also may comprise considering the above analysis for a plurality of targets and selecting the actual target based thereon.

[0064] According to embodiments of the present invention, receiving detection results and/or analysing the results may take into consideration reaching of thermodynamic equilibrium. In other words, receiving detection intensity results and/or analysing them may take into consideration the fact that no thermodynamic equilibrium state has been reached already. In some embodiments according to the present invention, the step of receiving detection results may take into consideration a thermodynamic non-equilibrium state of the hybridisation. The latter may for example encompass adapting the hybridisation conditions such that thermodynamic equilibrium or thermodynamic non-equilibrium can be obtained. Such hybridisation conditions may comprise for example the temperature at which hybridisation is performed, the length of the probes used, the hybridisation time used, etc. In some embodiments according to the present invention, the step of analysing takes into consideration a thermodynamic non-equilibrium state for the hybridisation. The latter may be performed by discarding a certain number of intensity results obtained, fitting predetermined correlations to part of the data obtained. In some embodiments, it is performed by fitting different predetermined correlations to the detected intensity, one being representative for the equilibrium state and one being representative for the non-equilibrium state.

[0065] Without wanting to be bound by theory, the results obtained with methods and/or system embodiments according to the present invention could be explained by the occurrence of non-equilibrium effects. More particularly, it was surprisingly found that, although typically the assumption of thermodynamic equilibrium was made in hybridisation experiments, thermodynamic non-equilibrium effects play a significant role. Experiments show that upon increase of the hybridisation time, substantially over times that are conventionally used in hybridisation experiments, more probes reach equilibrium. The experimental results for non-equilibrium effects could be explained by determination of the hybridisation free energy (G) which could be determined using a nearest neighbour model for the hybridisation. The nearest neighbour model for hybridisation between a target and a probe bound to a surface could be applied and model parameters could be determined due to a particular experimental design and dedicated custom microarrays, an example thereof being described further below. By extending the hybridisation theory, from the two-state model mentioned above to a three state model, the form of the experimentally determined non-equilibrium effects could be explained. As shown in FIG. 1, expressing the logarithm of the measured intensity I in the hybridisation experiments as function of the hybridisation free energy (G), only a partial confirmation of the linear relationship between log(I) and the hybridisation free energy (G) having a slope 1/RT could be obtained. It can be seen that there is also a deviating regime with an approximate linear behaviour between log(I) and the hybridisation free energy (G), at a different, i.e. smaller, slope.

[0066] It is to be noticed that, whereas embodiments of the invention mainly focus on systems and methods for analysing hybridisation or setting up hybridisation experiments, in one aspect embodiments of the present invention also relate to a method and system for determining hybridisation free energy for the hybridisation of a target initially in solution and a probe bound to a surface or for setting up a nearest neighbour model for the hybridisation free energy between a target initially in sample solution and a probe bound to a surface. The method thereby comprises performing a hybridisation microarray experiment using a plurality of microarray spots, wherein for a given target a perfect match probe and probes with up to two non-complementary elements such as two nucleotide mismatches are provided, e.g. by printing, on separate microarray spots for interaction with the target during the hybridisation microarray experiment. The method also comprises determining a set of parameters for a nearest neighbour model for the hybridisation free energy based on the hybridisation microarray experiment. Optionally the method may be directly applied for determination of the hybridisation free energy.

[0067] Turning back to the non-equilibrium conditions for hybridisation, it is to be noticed that for the applications, and their interpretation, it is advantageous to know in which regime one is measuring, since the regime influences the effect on intensity due to an additional non-complementary element, e.g. mismatch, in the target-probe duplex. If for example, one wants a probe that is as specific as possible to the perfect match target, in order to avoid that non-perfect matching targets hybridize the probe i.e. in order to avoid cross-hybridisation, it is important that the measurements are performed in the regime showing a relation between the logarithmic detection intensity and the hybridisation free energy with the highest slope, as this is the regime with the highest specificity.

[0068] In one embodiment a method is described for performing a hybridisation measurement especially suitable for identifying an actual target out of a set of known possible targets. The method comprises receiving detection intensity results for hybridisation of the actual but unknown target with a plurality of different probes, the probes being selected so that a range of hybridisation detection intensities for the hybridisation between each possible target of the set of known possible targets and the probe is covered. The latter directly implies that a range of free energies is covered. The method also comprises analysing the detection intensity results as function of the hybridisation free energy for each of the possible targets of the set of known possible targets and deriving the presence or concentration of the actual target by detecting a predetermined relationship between the detection intensity results and the hybridisation free energy. It is an advantage if reaching of equilibrium is taken into account, although the embodiment is not limited thereto. In a further application the actual target is a known minority target which is present together with a known main target which differs slightly from the minority target e.g. in one or two non-complementary elements such as by one or two nucleotide mismatches, and the application allows detection of presence of the minority target and a quantification of the relative concentration. In another application, the actual target is an unknown target differing slightly from a main target, e.g. in one or two non-complementary elements such as by one or two nucleotide mismatches, and the set of known possible targets comprises the main target and a set of targets differing from the main target in one or two nucleotides, and the application allows identification of the actual target. Further details and advantages are described in the exemplary applications as provided below.

[0069] Whereas embodiments according to this first aspect have mainly been described as methods, also systems are encompassed.

[0070] Systems according to the present aspects typically comprise a receiving means, e.g. an input port but alternatively also a hybridisation measurement setup, for receiving the detection intensity data as described above and a processing means or processor for analysing the results as described above. The system furthermore may comprise additional components for performing additional functionalities as described in the method embodiments described in the present application.

[0071] In one aspect, the present invention also relates to a method for performing hybridisation, wherein first a hybridisation step is performed between a target initially in the sample solution and a probe bound to a surface and thereafter a dehybridisation step is performed for removing cross-hybridized targets. The dehybridisation step is performed over a dehybridisation time substantially equal to the relaxation time of a hybridisation process between a target and a surface-bound probe having the unwanted non-complementary element, e.g. mismatch. The latter allows removal of such hybrids having the non-complementary element, while keeping the hybrids with perfect match. The relaxation time may advantageously be determined using a method as described above, applied to a target and a probe having the non-complementary element. An example of such a method for performing hybridisation is given below by way of an exemplary application. The present aspect also relates to a controller for controlling a hybridisation setup used for hybridisation experiments and thus for controlling hybridisation experiments as described above. The controller thus may be adapted for performing a hybridisation step and for performing a dehybridisation step for removing cross-hybridised targets, whereby the dehybridisation time advantageously is determined using a method for analysing hybridisation as described above. The controller may comprise a processor for performing an algorithm implementing the steps of a method for performing hybridisation as described above.

[0072] In another embodiment, the present invention relates to a method for analysing hybridisation comprising receiving detection intensity results for hybridisation between a target initially in solution and at least one probe bound to a surface, and analysing the detection intensity results as function of the hybridisation free energy. The method thereby takes into consideration the reaching of thermodynamic equilibrium which advantageously may be determined using a method for analysing hybridisation as described above.

[0073] In still another aspect of the present invention, a hybridisation kit for hybridisation measurements for identifying an actual target out of a set of known possible targets is described. The hybridisation kit comprises a microarray having a plurality of microarray spots each of them comprising a probe so that the plurality of different probes covers a range of detection intensities and corresponding therewith a range of hybridisation free energies for the hybridisation between each possible target of the set of known possible targets and the probes. The hybridisation kit may be especially suitable for performing applications as described in the present invention, although the invention is not limited thereto. The different probes may differ from each other or from one sub-set of the plurality of different probes only slightly, being by one or two non-complementary elements such as mismatches having a small influence on the hybridisation free energy.

[0074] In yet another aspect, the methods for analysing hybridisation may be used for designing a hybridisation kit, e.g. a kit as described above, although not limited thereto. The methods may in one example be used for determining whether for a given length of the probes equilibrium would be obtained and therefore may be used in the design of a hybridisation kit for determining a length of the probes used. In another example, the methods may be used for determining whether a sufficient range of detection intensity and corresponding therewith a sufficient range of hybridisation free energy is covered by a predetermined set of probes, and if required, the set of probes may be adjusted in view of this. Based on a the above methods for analysing hybridisation, a number of design guidelines may be derived. The method for analysing hybridisation as described above thus also may be used as a method for calibrating or setting up a hybridisation experiment, used as method for calibrating a nearest neighbour model as will be described below, as well as to analyse the hybridisation experiments performed.

[0075] By way of illustration, the present invention not being limited thereto, a number of examples of applications are shown below that can be performed with or make use of embodiments of the present invention.

[0076] A first exemplary application relates to a system or method for determining whether a hybridisation experiment is in equilibrium or not. The method comprises performing a hybridisation experiment using a plurality of different probes, whereby the different probes are selected to comprise a perfect matching probe for the target and probes containing non-complementary elements, e.g. mismatches, with the target in such a way that a sufficient range of intensities and corresponding therewith G is covered. The hybridisation experiment then results in an intensity value per probe. Evaluating the measured intensity as function of the hybridisation free energy, for example as function of the difference of hybridisation free energy with respect to the perfect match free energy for a possible target, allows for determining whether a spot has reached equilibrium or not. Spots that fulfil a predetermined relation between the logarithmic intensity measured and the hybridisation free energy, e.g. that fall on a line with slope 1/RT, are in equilibrium, spots deviating therefrom are not in equilibrium. As shown above in FIG. 1 some spots may be in thermodynamic equilibrium, while others are not. The intensities corresponding to spots not in equilibrium do not fulfil a predetermined relationship between the logarithmic intensity and hybridisation free energy. In FIG. 1, only the spots on the bottom line in the graph correspond with spots that are in equilibrium. FIG. 2 illustrates an experiment whereby thermodynamic equilibrium is reached for all spots. The example shown in FIG. 2 is based on an example wherein the targets are shorter than in FIG. 1 (25 nucleotides instead of 30 nucleotides), allowing for quicker dynamics and consequently to an earlier reached equilibrium state.

[0077] A second exemplary application relates to a system or method for increasing specificity of hybridization experiments. The specificity is increased by minimising cross-hybridisation. The latter is obtained by, after a hybridisation experiment has reached equilibrium, applying a dehybridisation step for removing cross-hybridized targets. The dehybridisation step thereby is characterised by a dehybridisation time set equal to the relaxation time of the hybridization process for a target-probe complex having a non-complementary element, being a mismatch in the present example, as can be determined using a method according to an embodiment of the present invention. It may be set equal to the relaxation time of the target-probe complex having a non-complementary element having the smallest effect on the hybridisation energy (with reference to a perfect match) of all unwanted non-complementary elements. The relaxation time of target-probe complexes having at least one non-complementary element differs from, typically is shorter than, the de-hybridisation time of a target-probe having no non-complementary element, as the hybridisation of the target-probe complex containing a non-complementary element has a lower free energy. To determine the optimal dehibridization time analysis is made of the graphs of log I as function of G and selection of the dehibridisation time is performed for conditions wherein the ratio of the intensity corresponding with probes having perfect match to intensity corresponding with probes having the non-complementary element is maximal. The de-hybridisation may for example be performed by washing out the sample so that the target molecules are removed from the solution. The washing may be done with pure water or with water solution containing appropriated solvents. By applying the de-hybridisation step, most of the cross-hybridised targets will be removed, while most of the perfect matching targets will remain on the probe-spot. Such a system or method may be applied for microarray applications, although the invention is not limited thereto. The technique also can be applied for purification, for selection of sequences out of a mixture for example as a preliminary step for sequencing experiments, etc.

[0078] A third exemplary application relates to a system and/or method for identifying a target out of a set of known targets, i.e. one wants to know which of a known set of targets is in the solution. Whereas this problem could be solved by sequencing, the problem also could be solved by a hybridisation experiment. The system and/or method comprises performing a hybridisation experiment using a plurality of different probes, whereby the probes are selected to comprise, for each target of the set of known targets, a perfect matching probe for the target and probes containing non complementary elements with the target in such a way that a sufficient range of detection intensity and consequently of G is covered. The hybridisation may be performed in a micro array experiment, although embodiments of the present invention are not limited thereto. The hybridisation experiment then results in an intensity value per probe. For each possible target of the known set of targets, the measured intensity for the plurality of different probes is evaluated as function of the hybridisation free energy of the possible target, e.g. as function of the difference between hybridisation free energy with respect to the perfect match free energy for the possible target. Based on this evaluation, identification of the actual target in the solution can be performed, as only in the case of the actual target, a predetermined correlation is detectable in the evaluation. By way of illustration, an example of such an evaluation is shown in FIG. 3A and FIG. 3B. The measured intensity is set out as function of the difference of hybridisation free energy with respect to the perfect match free energy for a possible target. The x-axis of the figure is dependent on the possible target selected, as the property on the axis is function of the free energy between the probe and the possible target, i.e. function of G(probe_i, possible target). FIG. 3A illustrates the obtained result for a possible target corresponding with the actual target in the solution, resulting in a single curve or stated differently collapsing of the data points into a single curve, whereas FIG. 3B illustrates the obtained result for a possible target not corresponding with the actual target in the solution. In the latter case, no single curve is obtained. The actual target can thus be identified as the possible target for which the evaluation results in a predetermined relationship between intensity measured during the hybridisation experiments and the hybridisation free energy for the possible target G(probe_i, possible target).

[0079] A fourth exemplary application relates to a system and/or method for detecting presence and/or quantifying the presence of a minority target in a solution comprising a main target being very similar to the minority target. With very similar there may be meant that there is only one or two non-complementary element, such as one or two nucleotide difference. The method and system according to embodiments of the present invention allows deriving the presence of a minority target and/or deriving an estimate of the proportion in which the minority target is present. It is an advantage of embodiments according to the present invention that the system and method for detection of minority targets in a sample solution that differ only slightly from a main target present in the solution is sensitive. It is an advantage of embodiments according to the present invention that a detection limit for minority targets can be obtained that is substantially better than 20%, as is for example the obtainable detection limit of Sanger sequencing techniques. It is an advantage of embodiments according to the present invention that detection may be performed based on hybridisation experiments. The system and/or method comprises performing a hybridisation experiment using a plurality of different probes, whereby the probes are selected to comprise, for both the main target and the minority target, a perfect matching probe for the target and probes containing non complementary elements with the target in such a way that a sufficient range of detection intensities and consequently of hybridisation free energy G is covered.

[0080] The hybridisation may be performed in a micro array experiment, although the invention is not limited thereto. The hybridisation is performed resulting in an intensity value for each probe. For the main target, the measured intensity for the plurality of different probes is then evaluated as function of the hybridisation free energy of the possible target, e.g. as function of the difference between hybridisation free energy with respect to the perfect match free energy for the possible target. Based on this evaluation, the presence of the minority target can be derived: if a single, predetermined correlation is derived, i.e. if the data collapse in a single curve, the minority target is not present within detection limit. If there is a deviation from a single curve or predetermined correlation, the minority target is present, and the proportion can be estimated. The latter may for example be performed, based on previously performed calibration measurements, or by theoretical calculation based on an extended equation correlating the logarithm of the measured intensity with the free energy G. By way of illustration, a particular test experiment is shown in FIGS. 4A-4F, illustrating the possibilities for detecting a minority target that differs a single nucleotide from a main target using a method as described above. Experiments were performed with samples containing a known concentration of minority target, varying over the different plots from 0.1% to 30%. In FIGS. 4A-4F, the spots that are most sensitive to the presence of the minority target are indicated as triangles. It can be seen that from 1% a 3% onwards, in the present example, detection of the minority target is possible. Such a detection limit is substantially better than what is conventionally obtained using Sanger sequencing. The experiments can also be further optimised by taking into account the non-equilibrium effect occurring in hybridisation experiments, as described above. The latter typically results in that only the lowest spots indicated by triangles are in equilibrium and therefore are most sensitive. Quantification therefore preferably is based on these lower spots.

[0081] A fifth exemplary application relates to detection of a minority target in a sample, similar as described in the previous application, but wherein the minority target is not known as such, but only differs from the main target by one or two non-complementary elements, in the present example being one or two nucleotides. The sequence of the minority target then can be derived by performing an experiment as set out above, whereby it is not a priori known which probes correspond with the minority target. Identification of the probes corresponding with those intensity values deviating from the predetermined correlation for intensity as function of the free energy may allow for identification of the sequence of the minority target. It is an advantage of embodiments according to the present invention that such techniques can advantageously be applied for searching single nucleotide polymorphisms. It is an advantage of embodiments according to the present invention that alternative and complementary techniques with respect to purely statistical methods are provided for tackling searching of single nucleotide polymorphisms.

[0082] The different exemplary applications described above can be combined resulting in advantageous characterisation and detection methods.

[0083] While the invention has been illustrated and described in detail in the drawings and foregoing description, such illustration and description are to be considered illustrative or exemplary and not restrictive. The invention is not limited to the disclosed embodiments.

[0084] In one aspect, embodiments of the present invention also relate to computer-implemented methods for performing at least part of the methods for analyzing hybridization experiments or methods for setting up hybridization experiments. Embodiments of the present invention also relate to corresponding computer program products. The methods may be implemented in a computing system. They may be implemented as software, as hardware or as a combination thereof. Such methods may be adapted for being performed on computer in an automated and/or automatical way. In case of implementation or partly implementation as software, such software may be adapted to run on suitable computer or computer platform, based on one or more processors. The software may be adapted for use with any suitable operating system such as for example a Windows operating system or Linux operating system. The computing means may comprise a processing means or processor for processing data. According to some embodiments, the processing means or processor may be adapted for performing analysis of one or more hybridization experiments according to any of the methods as described above. The processor therefore may be adapted for evaluating measured intensities as function of hybridization free energies and for determining therefrom an analysis result. Performing such analysis may comprise for example determining whether hybridization with certain probes was measured in equilibrium, determining the presence of a known target, determining the presence and/or quantity of a minority target, identifying a minority target being present, etc. Besides a processor, the computing system furthermore may comprise a memory system including for example ROM or RAM, an output system such as for example a CD-rom or DVD drive or means for outputting information over a network. Conventional computer components such as for example a keyboard, display, pointing device, input and output ports, etc also may be included. Data transport may be provided based on data busses. The memory of the computing system may comprise a set of instructions, which, when implemented on the computing system, result in implementation of part or all of the standard steps of the methods as set out above and optionally of the optional steps as set out above. By way of illustration, the present invention not being limited thereto, an exemplary flow scheme of a computer implemented method is shown in FIG. 5. The computer implemented method 500 for analyzing hybridization may comprise receiving 510 measured intensities for a hybridization experiment using a plurality of different probes having a different hybridization free energy G for at least one target. Such receiving may be receiving previously obtained data via an input means. Alternatively such receiving also may comprise obtaining the data directly through measurement during the hybridization experiment. The corresponding computing system therefore may comprise an input means for receiving input. The results also may be stored intermediately. The method further may comprise evaluating 520 measured intensities as function of the hybridization free energy for at least one target for a range of detection intensities and correspondingly a range of hybridization free energies. The range of detection intensities or hybridization free energies may be selected as function of the target to be characterized. Selection of such a range may for example be based on previously obtained results, or through trial and error. Evaluating may comprise determining a correlation between the measured intensities and the hybridization free energy or a property based thereon or being function thereof. Depending on the application, such an evaluation may be performed for one target, for a plurality of known targets that may be involved, etc. Such evaluation may be performed by an evaluation component of the processor. Based on the evaluation, e.g. on an established correlation between the intensities and the hybridization free energy, the computer-implemented method also may comprise deriving information regarding the at least one target and/or its hybridization. The latter may for example comprise determining the actual target present in the sample studied by hybridization, determining the presence and/or concentration of minority targets present in the sample studied by hybridization, detecting single nucleotide polymorphisms of targets in the sample studied, characterising the actual target present in the sample studied by hybridisation, determining whether the performed hybridisation experiment was in equilibrium for given probes, etc. Such information may be derived based on fitting of predetermined curves to the obtained correlation, evaluating deviation from predetermined curves, determining the slope of parts of curves, etc. The obtained results may be outputted through an output means such as for example a plotter, printer, display or as output data in electronic format.

[0085] The computing system and/or a corresponding computer implemented method may also be adapted for controlling the performance of the hybridization experiments themselves, although the invention is not limited thereby.

[0086] Further aspect of embodiments of the present invention encompass computer program products embodied in a carrier medium carrying machine readable code for execution on a computing device, the computer program products as such as well as the data carrier such as dvd or cd-rom or memory device. Aspects of embodiments furthermore encompass the transmitting of a computer program product over a network, such as for example a local network or a wide area network, as well as the transmission signals corresponding therewith.

[0087] By way of illustration, the present invention not being limited thereby, an example is given on how the free energy parameters of the nearest neighbour model could be fitted to results obtained for hybridisation in DNA microarrays. The example illustrates features and advantages of embodiments according to the present invention. Whereas for the present example theoretical considerations are taken into account, embodiments of the present invention are not limited thereby. For the present study several hybridization experiments were performed, each with a single oligonucleotide sequence (referred to as the target in this paper) in solution at different concentrations. Four different targets were used in the experiments, and their sequences are given in Table 1. The sequences contain a 30-mer hybridizing stretch followed by a 20-mer poly(A) spacer and a Cy3 label at the 3-end of the sequence. Each target oligo was bought in duplicate in order to check the quality of the target synthesis. Reference will be made to the two duplicated oligos as a and b. The sequences printed at the microarray surfaces and referred here as the probes were chosen to contain up to two mismatches, following the scheme shown in Table 2. Mismatches were inserted from nucleotides 6 to nucleotide 25 along the 30-mer sequences in order to avoid terminal regions. In the probes with two mismatches these were separated by at least 5 nt. Given the nucleotide of the target strand there are three different possible mismatching nucleotides and 20 available positions, hence in total 60 single mismatch sequences. A similar counting for double mismatches yields 945 different sequences (Table 2). The total number of probe sequences, including the perfect matching one, is 1006. For each experiment one target and one 815K custom Agilent slide was used. This slide consists of eight identical microarrays and each of these can contain up to more than 15 000 spots. The 1006 probe sequences were spotted in the custom array 15 times: in 12 replicates a 30-mer poly(A) was added on the 3-side (surface side), in order to assess the effect of a sequence spacer. Three replicates contained no poly(A) spacer. The eight microarrays of one slide have to be hybridized during the same experiment, but a different target solution can be used. In the experiments, the target concentrations ranged from 50 to 10000 pM according to the scheme given in Table 3. In Experiment 1 only target a was used, while in the Experiments 2, 3 and 4 both replicated targets (a and b) were used. Finally, in Experiments 1 and 2 a fragmentation of the target was performed before hybridization (see section on hybridization protocol for details). The four 30-mer target sequences were selected from fragments of human genes having a GC content ranging from 43% to 50%. A criterion for selecting the target sequences was the requirement that the probes constructed following the scheme in Table 2 would yield a roughly flat histogram of mismatch types, so that all mismatches are approximately equally present in the experiments.

[0088] For the experiments, the commercially available Agilent platform was used and a standard protocol with Agilent products was followed, as described subsequently. The target oligonucleotides were OliGold from Eurogentec, Seraing, Belgium. Hybridization mixtures contained one target oligonucleotide with a 3-Cy3 endlabeling diluted in nuclease-free water to the final concentration together with 5 l 10 blocking agent and 25 l 2GEx hybridization buffer HI-RPM. In Experiments 1 and 2 the addition of the hybridization buffer was proceeded by a fragmentation step, 1 l fragmentation buffer was added followed by an incubation of 30 min at 60 C. This fragmentation buffer is customarily used in Agilent hybridization platforms and produces target sequences of reduced length in order to speed up the hybridization reaction. Too long sequences, as obtained from biological extracts, e.g. from reverse transcription of mRNA samples, have a reduced hybridization efficiency due to steric hindrance. By comparing experiments with and without fragmentation, it was found that the fragmentation step has little effect on the results. The hybridization mixture was centrifuged at 13 000 r.p.m. for 1 min and each microarray of the 815K custom Agilent slides was loaded with 40 l.

[0089] The hybridization occurred in an Agilent oven at 65 C. for 17 h with rotor setting 10 and the washing was performed according to the manufacturer's instructions. The arrays were scanned on an Agilent scanner (G2565BA) at 5 mm resolution, high and low laser intensity and further processed using Agilent Feature Extraction Software (GE1 v5 95 Feb07) that performs automatic gridding, intensity measurement, background subtraction and quality checks.

[0090] In the present example, use is made of the Langmuir model for describing the dynamics of hybridization by a rate equation for , the fraction of hybridized probes from a spot as follows

[00003]$\begin{array}{cc}\frac{}{t}={\mathrm{ck}}_1\left(1-\right)-k_{-1}.Math.& \left[3\right]\end{array}$

where c is the target concentration and k.sub.1 and k.sub.1 are the attachment and detachment rates. The equilibrium value for can be obtained from the condition d.sub.eq/dt=0. Using the link between the rates and equilibrium constants, i.e. k.sub.1/k.sub.1=e.sup.G/RT, with G the hybridization free energy, R the gas constant and T the temperature one finds

[00004]$\begin{array}{cc}{}_{\mathrm{eq}}=\frac{c.Math..Math.{}^{-.Math..Math.G.Math.\text{/}.Math.\mathrm{RT}}}{1+c.Math..Math.{}^{-.Math..Math.G.Math.\text{/}.Math.\mathrm{RT}}}& \left[4\right]\end{array}$

which is the so-called Langmuir isotherm. To link this isotherm to the measured quantities one assumes that the fraction of hybridized probes is linearly related to the measured fluorescent intensity measured from a spot, which yields

[00005]$\begin{array}{cc}I=\frac{\mathrm{Ac}.Math..Math.{}^{-.Math..Math.G.Math.\text{/}.Math.\mathrm{RT}}}{1+c.Math..Math.{}^{-.Math..Math.G.Math.\text{/}.Math.\mathrm{RT}}}& \left[5\right]\end{array}$

[0091] Here I is the background-subtracted intensity, where the background subtraction, as explained above is done by Agilent Feature Extraction software. Where reference is made to the intensities, this are intensities that are background subtracted. A is a constant which is an overall scale factor.

[0092] Far from chemical saturation, i.e. when only a small fraction of surface sequences is hybridized (i.e. c e.sup.G/RT<<1) one can neglect the denominator in Equation [4] to get:

IAce.sup.G/RT[6]

[0093] In the nearest neighbour model, the hybridization free energy of perfect complementary strands is approximated as a sum of dinucleotide terms. For instance

[00006]$\begin{array}{cc}.Math..Math.G\left(\begin{array}{c}\mathrm{ATCCT}\\ \mathrm{TAGGA}\end{array}\right)=.Math..Math.G\left(\begin{array}{c}\mathrm{AT}\\ \mathrm{TA}\end{array}\right)+.Math..Math.G\left(\begin{array}{c}\mathrm{TC}\\ \mathrm{AG}\end{array}\right)+.Math..Math.G\left(\begin{array}{c}\mathrm{CC}\\ \mathrm{GG}\end{array}\right)+.Math..Math.G\left(\begin{array}{c}\mathrm{CT}\\ \mathrm{GA}\end{array}\right)+.Math..Math.G_{\mathrm{init}}& \left[7\right]\end{array}$

where G.sub.init is an initiation parameter. Since only differences of G between a perfect matching hybridization and a hybridization with one or multiple mismatches [Equation [9]] are considered, this initiation parameter will not contribute and it is omitted further in this example. For DNA/DNA hybrids, symmetries reduce the number of independent parameters to 10. The nearest neighbour model can be extended to include single internal mismatches; as an example it is considered that the free energy of a stretch with an internal mismatch of CT type

[00007]$\begin{array}{cc}.Math..Math.G\left(\begin{array}{c}\mathrm{AT}.Math.\underset{\_}{C}.Math.\mathrm{CT}\\ \mathrm{TA}.Math.\underset{\_}{T}.Math.\mathrm{GA}\end{array}\right)=.Math..Math.G\left(\begin{array}{c}\mathrm{AT}\\ \mathrm{TA}\end{array}\right)+.Math..Math.G\left(\begin{array}{c}T.Math.\underset{\_}{C}\\ A.Math.\underset{\_}{T}\end{array}\right)+.Math..Math.G\left(\begin{array}{c}G.Math.\underset{\_}{T}\\ C.Math.\underset{\_}{C}\end{array}\right)+.Math..Math.G\left(\begin{array}{c}\mathrm{CT}\\ \mathrm{GA}\end{array}\right)& \left[8\right]\end{array}$

[0094] The mismatching nucleotides are underlined and for notational reasons the mismatch is always put in the second part of the dinucleotide (which requires the use of symmetry like here in dinucleotide term three). There are 12 types of mismatches and 4 types of flanking nucleotide pairs, hence in total there are 48 mismatch parameters of dinucleotide type. There are several possible ways of extracting the 48+10 dinucleotide parameters from the experimental data.

[0095] One can either fit the full Langmuir isotherm [Equation [4]], or for experiments at sufficiently low concentrations one could consider the limiting case of Equation [6]. In addition, the parameters could be extracted either from an experiment at fixed concentration c, by comparing the intensities of different probe sequences, or from experiments at different concentrations by analyzing the intensities of identical probe sequences over a concentration range. Focus is put on the low concentration data and use of Equation [6] for the analysis at fixed c. Equation [6] contains the constant A which is an overall scale factor relating the hybridization probability to the actual measured fluorescence intensity. This quantity may fluctuate from experiment to experiment. For instance, the optical scanning influences A, as this is proportional to the laser intensity used. Also hybridizations in different slides might occur at slightly varying conditions and there can be small differences in the manufacturing of the slides. Focus will be put on relative intensities and relative free energies, i.e. for each microarray the perfect match of that microarray will be used as a point of reference. The logarithmic ratios of the intensities with the perfect match intensity are denoted as

[00008]$\begin{array}{cc}{y}_i=\ln .Math..Math.I_i-\ln .Math..Math.I_{\mathrm{PM}}=-\frac{.Math..Math.G-.Math..Math.G_{\mathrm{PM}}}{\mathrm{RT}}=-\frac{.Math..Math.G}{\mathrm{RT}}& \left[9\right]\end{array}$

for which the exact value of A is irrelevant and only the relative free energy differences G (which is for each probe a positive number) are considered. In G of a duplex, only dinucleotide parameters which are flanking a mismatch remain, the other parameters cancel out in the subtraction. For example from Equations [7] and [8] one gets

[00009]$\begin{array}{cc}.Math..Math.G\left(\begin{array}{c}\mathrm{AT}.Math.\underset{\_}{C}.Math.\mathrm{CT}\\ \mathrm{TA}.Math.\underset{\_}{T}.Math.\mathrm{GA}\end{array}\right)=-.Math..Math.G\left(\begin{array}{c}T.Math.\underset{\_}{C}\\ A.Math.\underset{\_}{T}\end{array}\right)+.Math..Math.G\left(\begin{array}{c}G.Math.\underset{\_}{T}\\ C.Math.\underset{\_}{C}\end{array}\right)-.Math..Math.G\left(\begin{array}{c}\mathrm{TA}\\ \mathrm{AT}\end{array}\right)-.Math..Math.G\left(\begin{array}{c}\mathrm{AC}\\ \mathrm{TG}\end{array}\right)& \left[10\right]\end{array}$

[0096] In this equation the lower strand refers to the target sequence in solution, which is fixed. The upper strand is that of the probe sequence attached to the solid surface. Hence, the G of a duplex with one mismatch can be written as a sum of two mismatch dinucleotide parameters minus two matching dinucleotide parameters. As it is assumed that the nearest neighbor model is valid, the same reasoning can be applied to duplexes with two mismatches which results in a sum of four mismatch dinucleotide parameters minus four matching dinucleotide parameters. The model can now be written as

[00010]$\begin{array}{cc}{y}_i=\underset{=1}{\overset{58}{.Math.}}.Math..Math.X_{i.Math..Math.}.Math.\frac{.Math..Math.G_{}}{\mathrm{RT}}& \left[11\right]\end{array}$

where a is the index running over the 58 possible dinucleotide parameters and X is a frequency matrix, whose elements X.sub.i are the number of times the dinucleotide parameter enters in G of probe sequence i. With a simple extension of matrices and vectors one can rewrite the problem as

{right arrow over (y)}=X{right arrow over ()}[12]

[0097] Where it is defined that

.sub.=G.sub./RT. Having written the problem in Equation [12] as a linear one, one can now apply the standard approach to find the optimal values of the parameters. The procedure consists in minimizing S=({right arrow over (y)}X{right arrow over ()}).sup.2 which amounts to solving the following linear equation

X.sup.T({right arrow over (y)}X{right arrow over ()})=0[13]

where X.sup.T is the transpose of X.

[0098] To obtain {right arrow over ()} from Equation [13] one has to invert the 5858 matrix X.sup.TX. In the case that X.sup.TX is not invertible one applies a singular value decomposition. In the present case the matrix is not invertible. Zero eigenvalues of the matrix X.sup.TX come from reparametrizations that leave the physically accessible parameters G invariant. The dinucleotide mismatch parameters are not uniquely determined, as these parameters are entering in the expression for the total G in pairs [Equation (8)]. For instance, a reparametrization of the type:

[00011]$\begin{array}{cc}.Math..Math.G^{}\left(\begin{array}{cc}x& \underset{\_}{C}\\ x^{}& \underset{\_}{T}\end{array}\right)=.Math..Math.G\left(\begin{array}{cc}x& \underset{\_}{C}\\ x^{}& \underset{\_}{T}\end{array}\right)+\mathrm{.Math.}.Math..Math..Math..Math.G^{\left(\begin{array}{cc}y& \underset{\_}{T}\\ y^{}& \underset{\_}{C}\end{array}\right)}=.Math..Math.G\left(\begin{array}{cc}y& \underset{\_}{T}\\ y^{}& \underset{\_}{C}\end{array}\right)-\mathrm{.Math.}& \left[14\right]\end{array}$

for every pair of complementary nucleotides x; x and y; y leaves the total G invariant, as it can be verified directly from Equation [8]. Similar reparametrizations are possible for mismatches of type AG, AC and TG. Next to these there are three invariances of G that involve a reparametrization of both mismatch and matching dinucleotide parameters. Hence one has at least seven zero eigenvalues in X.sup.TX.

[0099] As a control of the reproducibility of the result, the intensities correlation between analogous spots in replicated experiments is considered. The replicated hybridizations were carried out on two microarrays of the same slide, with two identical but separately synthesized and labeled target oligos, at the same manually prepared concentration in solution Table 3. FIGS. 6A and 6B are an example thereof. It shows correlation plots between two replicated hybridizations. Two plots are shown, one with the full 15K intensities FIG. 6A and one in which the median of the intensities of the 15 replicated spots are taken FIG. 6B. In the former some data spreading is observed, which is greatly reduced when the median over 15 replicated spots is taken. Note that the experimental data do not align perfectly on the diagonal of the graph, this may be attributed to the manual preparation of the solutions or to differences in the oligos (synthesis or labeling). Data from different microarrays are aligned on a line of slope equal to one in the log-log plots of FIGS. 6A and 6B, which implies a linear relationship between the intensities. In general, replicates show a strong correlation between median intensities, which is an indication of a good reproducibility of the results. In this median the probes with and without poly(A) spacer are included. No significant difference was found in the intensities from spots with poly(A) and without poly(A) spacer. From this point on, the median intensity of 15 replicates is always used and simply referred to as the intensity of a probe, and because of the good reproducibility only the data produced by hybridizations with oligo synthesis a (Table 3) will be discussed. Next, the relation between the intensities and the corresponding G.sub.sol for hybridizations in solution with one or two mismatches are considered. In the case of two mismatches G.sub.sol was calculated as the sum of nearest neighbour parameters for individual mismatches, assuming that the presence of two mismatches does not involve additional terms in the free energy, i.e. they do not interact. In the experiment the minimal distance between two mismatches is 5 nt, which is considered sufficient, in first approximation to support the non-interaction assumption. In the calculation of G from the tabulated values of H and S the temperature was set to the experimental value T=65 C.

[0100] FIG. 7A shows plots of the intensities versus G.sub.sol as taken from the nearest neighbour model with the existing tabulated values for hybridization in solution. G.sub.sol is obtained by subtracting from all free energies that of the PM sequence, which is taken as a reference. As a consequence, for the PM intensities G.sub.sol=0. Each plot in FIGS. 7A and 7B contains 1006 data points obtained from the median value of the 15 replicated spots on each array.

[0101] As it is well-known from several studies of melting/hybridization in aqueous solution, the hybridization free energy G.sub.sol depends on the buffer conditions, and in particular of the ionic strength of the solution. Particularly studied was the effect of salt concentration (NaCl), which is usually assumed to be independent of sequence, but to be dependent on oligonucleotide length. Melting experiments in solution are consistent with the following dependence on Na ion concentration

G.sub.sol=G.sub.sol(1M[Na.sup.+])N ln [Na.sup.+][15]

[0102] Where G.sub.sol (1M[Na.sup.+]) is measured at 1M NaCl, N is the number of phosphates in the sequence and a is a constant.

[0103] Salt has mostly an effect on interactions with the negatively charged phosphate molecules. It is hence plausible to expect the same type of correction as Equation [15] also for sequences carrying mismatches. If that is the case, the salt dependence cancels out from G.sub.sol, which is the quantity of interest. Therefore the value will be set at 1M NaCl in G.sub.sol.

[0104] FIG. 7A shows the data for Experiment 1 at three different concentrations, from bottom to top of 50, 500 and 5000 pM. When plotted as functions of G.sub.sol, the data points tend to cluster along single monotonic curves. This already suggests a fair degree of correlation between G.sub.sol and G.sub.array. The experiment at 5000 pM shows a pronounced saturation of the intensities, as expected from the Langmuir model [Equation [4]]. Sufficiently far from saturation one expects a linear relationship between the logarithm of the intensity and G, as given by Equation [6]. FIGS. 7A and 7B shows that the low concentration data at low intensities follow approximately a straight line with the slope 1/RT expected from equilibrium thermodynamics at T=65 C., which is the experimental temperature.

[0105] However, the global behavior of the three concentrations is at odds with the Langmuir model, which predicts that intensity versus free energy plots for different concentrations should saturate at a common intensity value A, as indicated in FIG. 7B. Although one may expect some variations on A from experiment to experiment, the data of FIG. 7A are hard to reconcile with the Langmuir model. It is concluded that the hybridization data deviate from the full Langmuir model of Equation [4], but they are in rather good agreement with its limiting low intensities behavior [Equation [6]]. In order to obtain estimates of the free energies G.sub.array from microarray data, then Equation [6] will be used and restriction will be made to the lower concentration data.

[0106] To fit the 58 parameters of the nearest neighbor model the lowest concentration data are used, i.e. 50 pM. Hereto the algebraic procedure explained above is applied, which fits the logarithm of the ratios I/I.sub.PM and which assumes that the data can be described by Equation [9]. For low concentrations this assumption is expected to be correct for the lower intensities but not for the highest intensities, which deviate from the Langmuir isotherm as shown in FIGS. 7A and 7B. This poses a problem for the fitting procedure since it was designed with the perfect match intensity I.sub.PM as a reference [Equation [9]]. One may think to circumvent this problem by restricting the fit to low intensities, for instance only to probes with two mismatches and rewrite Equation [9] using as reference not I.sub.PM, but for instance one of the intensities of a probe with two internal mismatches. This procedure turns out to be of little practical use for the purpose of estimating the free energy difference between perfect matching sequences and sequences with one or multiple mismatches and for which the PM reference value is necessary.

[0107] From the analysis of plots of intensity versus G.sub.sol (FIGS. 7A-7B), one finds that the PM intensity is systematically lower than that predicted by Equation [6], which is the straight line in FIG. 7A. Hence, the relative intensities I/I.sub.PM of the probes that contain mismatches are systematically higher than those predicted by Equation [6].

[0108] Consequently, a direct fit of the experimental data to Equation [9] underestimates the effect of a mismatch, which will result in free-energy penalties that are too small. The result of the fit is shown in FIG. 8. One can notice that the G range is indeed smaller than the one from hybridization in solution (FIGS. 7A and 7B).

[0109] Moreover, the underestimation of G is more severe for probes with two mismatches than for those with only one, since G is a sum of contributions per mismatch. This produces a discontinuity of the curve from double to single mismatches. The appearance of this discontinuity is another evidence of the fact that Equation [6] is not valid in the full range of intensities.

[0110] In order to solve this problem, one would need to fit the data with a more general model I(c,G) that incorporates the observed deviations from Equation [6]. Moreover, the choice of this model may considerably influence the fitted nearest neighbor parameters. A safer compromise is to start from the observation that Equation [6] is followed by the large majority of the low concentration data points in FIGS. 7A and 7B. Hence a fit to the low concentration limit of the Langmuir model seems reasonable. Unfortunately, one of the points deviating from Equation [6] is the PM intensity, which is used as reference measure. In order to calibrate the fit correctly one should reweight the reference PM intensity. The data therefore were fitted against Equation [9] using instead of the actual PM intensity as a reference, a rescaled value I*.sub.PM=I.sub.PM, which is the value the PM intensity would have if the data would agree with Equation [6] in the whole intensity range. is estimated from the crossing of the 50 pM fitting line in FIG. 7A with the G=0 axis. This estimate is =30.

[0111] The effects of a change in a on the fitting parameters will be discussed below.

[0112] FIGS. 9A-9D show the result of the fit to Equation [9], using =30. In the main frames each experiment is fitted independently. In the insets, the free-energy parameters are obtained from a simultaneous fit of all 50 pM experiments.

[0113] The latter data produce more accurate parameters, as they come from using four independent experiments (the four experiments at 50 pM, oligo synthesis a, in Table 3), hence the 58 parameters are obtained on sampling over 10064 data points. Both the free-energy range and the continuity of the curves in FIGS. 9A-9D are now as expected. The data show very little spreading in comparison with the curves in FIG. 7A. A quantification of the spreading for a monotonic curve can be assessed by the Spearman's rank correlation coefficient, which for all four experiments is very close to 1. This is an indicator of the reliability of the nearest neighbor fitted parameters. The ratio of data points over tuning parameters is as large as 4024/58, which ought to yield a reliable fit. Moreover, although the data are fitted to a linear model, all four experiments show a clear deviation for the highest intensities.

[0114] This is an indication against overfitting, which would result in a fully linear curve with erroneous fitting parameters. Therefore, it is concluded that the deviations from the Langmuir isotherm observed in all four experiments is a robust feature of the system and that the resulting free-energy parameters are physically meaningful. It is also verified that the free-energy parameters obtained from the fit are quite stable whether one fits the whole set of experimental data, or whether the fit is restricted to the lowest intensity scales (e.g. I/I*.sub.PM510.sup.3) where all data clearly follow Equation [6]. This is because the large majority of experimental points in FIGS. 9A-9D are located in the lowest intensity scales, anyhow. Hence, this additional data filtering has little effects on the parameters.

[0115] Table 4 shows the free-energy parameters G.sub.array as obtained from the above fitting procedure. Because of the degeneracies mentioned above, the dinucleotide parameters are not uniquely determined. Triplet parameters are, however, unique, and these are given in the table. The G for triplets are defined, for instance

[00012]$\begin{array}{cc}.Math..Math.G\left(\begin{array}{c}A.Math.\underset{\_}{C}.Math.G\\ T.Math.\underset{\_}{T}.Math.C\end{array}\right)=.Math..Math.G\left(\begin{array}{c}A.Math.\underset{\_}{C}.Math.G\\ T.Math.\underset{\_}{T}.Math.C\end{array}\right)-.Math..Math.G\left(\begin{array}{c}\mathrm{AAG}\\ \mathrm{TTC}\end{array}\right)& \left[16\right]\end{array}$

where the upper strand is 5-3 oriented. The lower strand is the invariant target sequence, the upper strand is the probe sequence. Hence the G parameters are measured subtracting the reference perfect match probe. Note that because of this subtraction one has

[00013]$\begin{array}{cc}.Math..Math.G\left(\begin{array}{c}A.Math.\underset{\_}{C}.Math.G\\ T.Math.\underset{\_}{T}.Math.C\end{array}\right).Math..Math.G\left(\begin{array}{c}C.Math.\underset{\_}{T}.Math.T\\ G.Math.\underset{\_}{C}.Math.A\end{array}\right)& \left[17\right]\end{array}$

as the reference PM sequence is different in the two cases. Using standard linear regression tools, the error bar was estimated on the parameters of Table 4 to be equal to 0.2. In Table 5 the G.sub.sol for triplets following the same notation as in Table 4 are presented. As mentioned before the data in solution are at T=65 C. and 1M [Na.sup.+]. FIG. 10 shows a plot of the two free energies G.sub.array versus G.sub.sol. A clear quantitative correlation between the two is observed. The Pearson correlation coefficient is 0.839. In comparing the two sets, it can be noted that the 16 mismatches of CC appear to be the most deviating in the two cases.

[0116] As discussed above, the fit was done with a re-scaled PM intensity, using a factor =30. The analysis was repeated for other values of . Varying causes a global shift of the data in Table 4 by an -dependent constant.

[0117] This shift does not affect the slope or correlation of the data in FIG. 10. By using =50 a positive shift of 0.17 was found, while setting =20 produces a shift of 0.14. These two values of are our estimate of the largest range of variability for this parameter. In general, the procedure of re-weighting the PM intensity with introduces a global error 0.2 affecting all parameters in Table 4.

[0118] One of the advantages of the experimental setup chosen in this work is that one can obtain in principle all parameters in a single experiment, as all hybridization reactions with one or two mismatches occur in parallel on a single array. However, a drawback is that in this setup one can determine only the free energy and not the contribution of enthalpy and entropy separately, which would allow to extend the parameters to other temperatures.

[0119] In the above example, focus was put on the determination of G which is the free energy difference between a perfect matching hybridization and an hybridization where the probe sequences have one or more internal mismatches. Quantifying the effect of internal mismatches is important for a better understanding of cross-hybridization effect, which is the unintended binding of non-perfectly complementary sequences to a given probe. Moreover, this understanding could have some practical consequences for optimal probe design. An advantage of the parameter G is that it is insensitive to the free-energy initiation parameter [Equation [7]] and the scaling factor A [Equation [4] and [6]] and that it is expected to be less sensitive to buffer conditions as ionic salt etc. The example, showing custom Agilent arrays shows that there is a strong correlation, also on the quantitative scale, between G.sub.sol and G.sub.array. This correlation is shown in FIG. 10 with explicit free-energy values given in Table 4 and 5. A fit of the interaction parameters from microarray data shows a much better agreement of the data with the thermodynamic models (compare FIGS. 7A and 7B with FIG. 8). However, in the absence of dedicated experiments for the determination of interaction free energies on a DNA microarray, the results of this work suggest that one could use the corresponding hybridization free energies in solution as approximations for them.

[0120] As a correlation between G.sub.sol and G.sub.array has by now been observed in several different microarray platforms, it is fair to expect that such a correlation is a general feature of microarrays.

[0121] It is interesting to remark that the deviation from the Langmuir model enhances the cross-hybridization problem because there is a smaller effect on intensity for a given free energy penalty (smaller slope in FIGS. 9A-9D). As an example, a mismatch with G=2.5 kcal/mol (a typical value from Table 4) corresponds to a I/I.sub.PM ratio of 0.02 in the regime governed by the Langmuir model, compared to 0.2 in the deviating regime. This implies that in the deviating regime a significant fraction of the amount of target binding to a PM probe binds to a probe carrying one internal mismatch.

[0122] In a second example, hybridisation in DNA microarrays is discussed as also discussed in the first example. The second example illustrates the existence of slow relaxation phenomena for hybridisation in DNA microarrays.

[0123] Experiments are described wherein hybridization takes place between the surface-bound sequences (referred to as probes) and the sequences in solution (targets) carrying a fluorophore. The amount of hybridized target is measured from the emitted fluorescence from a given location (spot) on the microarray surface. It is illustrated that, contrary to a widespread belief, in DNA microarrays relaxation times may largely exceed typical experimental times, causing a breakdown of thermal equilibrium. Experiments are performed on a commercial microarray platform under the same buffer conditions as in typical biological experiments. They are further corroborated by the analysis of an extended kinetic model. In equilibrium one expects that the intensity measured from a spot is described by the Langmuir

[00014]$\begin{array}{cc}\mathrm{model}.Math..Math.I=\frac{\mathrm{Ac}.Math..Math.{}^{\frac{.Math..Math.G}{\mathrm{RT}}}}{1+c.Math..Math.{}^{\frac{.Math..Math.G}{\mathrm{RT}}}}\mathrm{Ac}.Math..Math.{}^{\frac{.Math..Math.G}{\mathrm{RT}}}& \left[18\right]\end{array}$

where A sets the intensity scale, R is the gas constant, T is the temperature, c the target concentration and G the hybridization free energy. It is to be noticed that a different sign convention is used compared to the first example. This is merely a matter of choosing the bound or the unbound state as the reference state for free energy differences. In Eq. [18] it took ce.sup.G/RT<<1 (weak binding and small concentrations), a limit which applies to the experiments discussed here.

[0124] The experimental setup is schematically shown in Table 6 and further similar to the first example.

[0125] FIG. 11 shows a plot of I/c vs. G for four experiments at different concentrations using the setup of Table 6 (the target sequence is the 30-mer). GGG.sub.PM is the difference in hybridization free energies between a given sequence and the perfect match (PM) sequence, calculated from the nearest-neighbour parameters. The collapse of the I/c vs. G plots into a single curve shows that Ic as expected from the low concentration limit of Eq. [18].

[0126] However, the dependence on G is not in full agreement with Eq. [18]. In the regime deviating from Eq. [18] log I scales approximately linearly with G but with a slope smaller than 1/RT.

[0127] Hybridization dynamics of oligonucleotides in aqueous solution is usually described as a two state process characterized by one association and one dissociation rate. The Langmuir isotherm itself (Eq. [18]) is a two state model. Hybridization in DNA microarrays is likely to be more complex than a simple two state process. Probes are tethered to the surface by one end and can form a dense brush, which hinders and slows down hybridization. The typical distance between probes is of about 10 nanometers, and the length of a fully stretched 30-mer duplex is of 10 nm and its thickness of 2 nm. Probe sequences in the experiment have also a poly(A) 30-mer spacer (see Table 6). Therefore a single target molecule can interact with more than one probe. Taking this into account, we have extended the two state hybridization model with an additional intermediate state (FIG. 12). Indicating with .sub.1 and .sub.2, the fraction of partially and fully hybridized probes on the microarray, the kinetics of these reactions is given by

[00015]$\begin{array}{cc}\frac{{}_1}{t}={\mathrm{ck}}_1\left(1-{}_1-{}_2\right)+k_{-2}.Math.{}_{-2}-\left(k_{-1}+k_2\right).Math.{}_1& \left[19\right]\\ \frac{{}_2}{t}=k_2.Math.{}_2-k_{-2}.Math.{}_2& \left[20\right]\end{array}$

where c is the target concentration in solution and k.sub.1, k.sub.1, k.sub.2 and k.sub.2 the four rates involved (see FIG. 12). For simplicity we have assumed that at most a single target molecule can bind to a given probe. The rate constants, using a two state model description, have been measured in several microarray experiments. The hybridization of a common target sequence to a perfect match probe and to a probe containing one mismatch were considered. One is interested in their dependence on G. The following rates were used (at 45 C.):

k.sub.1.sup.(PM)=19.Math.10.sup.4 M.sup.1 s.sup.1,

k.sub.1.sup.(MM)=21.Math.10.sup.4 M.sup.1 s.sup.1,

k.sub.1.sup.(PM)=12.Math.10.sup.4 s.sup.1 and

k.sub.1.sup.(MM)=29.Math.10.sup.4 s.sup.1.

[0128] While there is more than a factor two of difference in the detachment rates, the attachment rates differ only by 10%. These results are in agreement with observation for kinetic behaviour in bulk solution. The probes in the experiment differ by at most by two nucleotides out of 30. It is assumed that both forward rates k.sub.1 and k.sub.2 are sequence independent. The reverse rates are then fixed by the thermodynamics relations

k.sub.1=k.sub.1e.sup.G/RT;

k.sub.2=k.sub.2e.sup.(GG)/RT[21]

where G and G are the free energy differences between configurations 1 and 2, and the unbound state, respectively. It is next assumed that G, the free energy of the partially hybridized state, is monotonically dependent on G. Moreover at unbinding (G=0), also G should vanish. As a simple approximation we then take

G=G(<1)[22]

in order to approximate the expected monotonic dependence of G from G. The model is then fully characterized by three parameters k.sub.1, k.sub.2 and .

[0129] FIG. 13 shows a plot of .sub.1+.sub.2 vs. G. These are obtained from the solution of Eqs. [19], [20] for different times and the following choice of parameters k.sub.1=10.sup.5 M.sup.1 s.sup.1, k.sub.2=1 s.sup.1 and = (for k.sub.1 the value was used typically measured in kinetic experiments on microarrays, while k.sub.2 and are chosen to fit experimental data). In FIG. 13 thin solid lines are isotherms at finite times, while the thick line is the equilibrium isotherm (t.fwdarw.+).

[0130] As time increases the equilibrium intensity is approached from below. To gain some more insight the limit of fast equilibration is considered for Eq. [19]. We then solve Eq. [20] using for .sub.1 its equilibrium value:

[00016]$\begin{array}{cc}{}_1^{\left(\mathrm{eq}\right)}=\frac{c.Math..Math.{}^{.Math..Math.G^{}.Math.\text{/}.Math.\mathrm{RT}}}{c\left({}^{.Math..Math.G.Math.\text{/}.Math.\mathrm{RT}}+{}^{.Math..Math.G.Math.\text{/}.Math.\mathrm{RT}}\right)+1}c.Math..Math.{}^{.Math..Math.G^{}.Math.\text{/}.Math.\mathrm{RT}}& \left[23\right]\end{array}$

where the low concentration limit ce.sup.G/RT<<1 was taken. We have then

.sub.2(t)=ce.sup.G/RT(1e.sup.t/)[24]

.sup.1=k.sub.2=k.sub.2e.sup.(1)G/RT[25]

which is the inverse relaxation time. To get this Eqs. [21] and [23] were used in the limit ce.sup.G/RT<<1.

[0131] The relaxation time depends on G: weakly bounded sequences (small G) equilibrate faster than strongly bounded ones (large G). For fast equilibrating sequences (<<t) one recovers from Eq. [24] the usual Langmuir equilibrium; for sequences with long equilibration times >>t Eq. [24] is expanded to lowest order in t/. With this approximation we find that for a given time t

[00017]$\begin{array}{cc}{}_2\left(t\right)=\{\begin{array}{cc}c.Math..Math.{}^{.Math..Math.G.Math.\text{/}.Math.\mathrm{RT}}& .Math..Math.G.Math..Math.G^{*}\\ {\mathrm{ctk}}_2.Math.{}^{.Math..Math.G.Math.\text{/}.Math.\mathrm{RT}}& .Math..Math.G.Math..Math.G^{*}\end{array}& \left[26\right]\end{array}$

where G* is a crossover free energy which depends on time and is obtained by setting =t in Eq. [25]. The measured intensity is I=A(1+2)), however for any realistic choice of free energies .sub.1<<.sub.2, hence once can approximate IA.sub.2. Equation [26] reproduces the two slopes in the log I vs. G plots as seen in the experiments. It shows that the non-equilibrium regime is characterized by a slope equal to /RT. Turning now to experimental results, routinely, hybridization experiments are performed at constant temperature and buffer conditions for about 15 h. Experiments were performed at shorter and longer hybridization times up to more than 86 h. Once the desired hybridization time has been reached the experiment is stopped, the microarray washed and scanned to measure the emitted fluorescence from every spot. Experiments at different hybridization times thus require different slides.

[0132] FIGS. 14A-14D shows a plot of I vs. G for a 30-mer target at four different times and for a concentration of 50 pM (the 17 h hybridization data are those already shown in FIG. 11). As the hybridization time increases a larger fraction of the data aligns along a line with a slope 1/RT over the full range of intensities, confirming that the observed deviations from the Langmuir model are due to the breakdown of thermodynamic equilibrium. Surprisingly, full equilibrium has not been reached here even after 85 hours. Apart from the shortest hybridization time the data are in agreement with the behaviour predicted by Eq. [26]. From the slope of the dashed line, using Eq. [26] it is estimated that =0.32. As full hybridization involves a stretch of 30 nucleotides, this suggests that in the partially hybridized state the target and probe are bound for 10 nucleotides, the turn of an helix. The slope of the non-equilibrium regime of FIG. 14A is smaller than that of the dashed lines of FIGS. 14B, 14C, and 14D. This is probably due to the protocol followed (this is the standard protocol): at time t=0 both the solution containing the target molecules and the array are at room temperature.

[0133] They are then placed into an oven for the duration of the experiment at a constant temperature (T=65 C. for all experiments described in this letter). The slope is reminiscent of an initial low temperature hybridization.

[0134] When comparing FIGS. 14B, 14C and 14D one notices an overall decrease of the intensity scale, leading to a normalization of the constant A in Eq. [18]. This is probably due to some degradation of the fluorophores (not surprising in view of the time span involved in the experiment). The solid lines in FIGS. 14B, 14C and 14D are plots of the intensity I=A(.sub.1+.sub.2), where A was adjusted to match the global intensity scale. The agreement with the model is reasonable although not perfect.

[0135] Turning now to the case of hybridization to the shorter target sequence (25-mer, see Table. 6). Data are shown in FIGS. 15A-15D. It is expected that faster equilibration for 25-mers occurs because the sequence has a lower G and because for shorter target sequences it is expected that less entanglement occurs hence an increase of the k.sub.2 rate. Both effects lead to smaller (see Eq. [25]). As can be seen in FIGS. 15A-15D the only deviation from the equilibrium isotherm are for the shortest hybridization time. The agreement with Eq. [18] is over three orders of magnitude in the intensity scale. Note an overall decrease of the intensity scale as observed in FIGS. 14A-14D. The results are consistent with the idea that interaction of target molecule with multiple probes is responsible to the observed non-equilibrium behaviour. The present example shows that hybridization in DNA microarrays under standard conditions is characterized by a relaxation time which may largely exceed the experimental time. Since typical biological experiments involve target strands of 30-50 nucleotides, it is believed that the breakdown of equilibrium shown here on Agilent arrays, may occur in many different microarrays platforms and in biological experiments. It is found that in the non equilibrium regime the intensities are distributed according to an exponential distribution e.sup.G/RT, with <1. The breakdown of equilibrium carries important consequences: it lowers the specificity of the microarrays as devices for the detection of a desired sequence from a complex mixture. This can be illustrated with an example. Consider a probe at the microarray surface and two different sequences in solution: one perfect matching with the probe (at concentration c.sub.PM) and one with a mismatch (at concentration c.sub.MM). In the equilibrium regime the two sequences hybridize to the probe with probabilities proportional to c.sub.PMe.sup.G.sup.PM.sup./RT and C.sub.MMe.sup.G.sup.MM.sup./RT, respectively. Assuming for simplicity equal target concentrations c.sub.PM=c.sub.MM one has that the ratio of the two contributions is e.sup.(G.sup.MM.sup.G.sup.PM.sup.)/RT0.05 where a typical value G.sub.MMG.sub.PM=2 kcal/mol and a temperature of T=65 C. is used. In the non-equilibrium regime, due to the presence of a factor = in the exponential the ratio is about 0.4. Therefore in the non-equilibrium regime a significant fraction of a measured signal may be due to hybridization to non-complementary targets a phenomenon known as cross-hybridization. For an optimal functioning of the microarrays it is then desirable to work under equilibrium conditions. Several parameters may influence the relaxation time as temperature, salt and buffer conditions. The experimental setup discussed provides a good test of equilibrium (single line vs. broken line in a I vs. G plot) and can be used to investigate the best working conditions for an optimal hybridization.

[0136] Other variations to the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed invention, from a study of the drawings, the disclosure and the appended claims. In the claims, the word comprising does not exclude other elements or steps, and the indefinite article a or an does not exclude a plurality. A single processor or other unit may fulfill the functions of several items recited in the claims. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage. Any reference signs in the claims should not be construed as limiting the scope.

[0137] The foregoing description details certain embodiments of the invention. It will be appreciated, however, that no matter how detailed the foregoing appears in text, the invention may be practiced in many ways, and is therefore not limited to the embodiments disclosed. It should be noted that the use of particular terminology when describing certain features or aspects of the invention should not be taken to imply that the terminology is being re-defined herein to be restricted to include any specific characteristics of the features or aspects of the invention with which that terminology is associated.