Method for detecting the impacts of interfering effects on experimental data

Abstract

A method for identifying the impact on data, such as experimental data, of interfering effects, such as unwanted auto-fluorescence, fluorescence quenching, and fluorescent-sample deterioration, whether or not the data fulfill certain criteria with respect to a threshold indicative of the interfering effects.

Claims

1. A system for detecting the impact of and correcting for interfering effects of auto-fluorescence, fluorescence quenching, and fluorescence-signal deterioration on fluorescence emission data resulting from fluorescence measurements, said system comprising: (i) a device for supporting one or a plurality of samples selected from the group consisting of drug-candidate samples and patient-specimen samples in an inspection station, (ii) one or a plurality of fluorescence signal detectors which are positioned relative to the inspection station so that a fluorescence signal emitted from the samples impinges on the fluorescence signal detectors, and (iii) a fluorescence signal processing unit programmed to perform the steps of receiving fluorescence emission data generated by the fluorescence signal detectors, determining values of a plurality of fluorescence identification parameters from said fluorescence emission data including at least first and second corresponding fluorescence identification parameters, storing the determined values in such a manner that all the determined values which relate to any one of the samples are linked, creating a multi-dimensional histogram or distribution of the determined values of the fluorescence identification parameters, determining thresholds for the determined values of the fluorescence identification parameters from said multi-dimensional histogram or distribution, which thresholds are indicative of interfering effects selected from the group consisting of auto-fluorescence, fluorescence quenching, and fluorescence-signal deterioration, wherein the threshold for the determined values of the first fluorescence identification parameter is a function of the corresponding second fluorescence identification parameter, analyzing the determined values of the plurality of fluorescence identification parameters whether or not the determined values fulfill one or a plurality of criteria with respect to the thresholds, supplying as output information fluorescence emission data influenced by the interfering effects, fluorescence emission data not affected by the interfering effects, or fluorescence emission data influenced by the interfering effects and fluorescence emission data not affected by the interfering effects, and using the output information to detect at least one of false positive drug-candidate test results in the fluorescence emission data, false negative drug-candidate test results in the fluorescence emission data, false positive diagnostic test results in the fluorescence emission data, and false negative diagnostic test results in the fluorescence emission data.

2. The system of claim 1 further comprising a fluorescence reader.

3. The system of claim 1 further comprising a fluorescence reader including a confocal optical set-up.

4. The system of claim 1 wherein the photosensitive detector comprises an avalanche photodiode or a charged coupled device (CCD) camera.

Description

(1) Other objects, advantages, and novel features of the invention will become apparent from the following detailed description of the invention when taken in conjunction with the accompanying drawings.

(2) FIGS. 1A-F illustrates impacts of the effect of quenching compounds on fluorescence emission data and presents different identification methods.

(3) FIGS. 2A-E illustrates impacts of the effect of auto-fluorescent compounds on fluorescence emission data using polarization and 2D-FIDA read-outs and presents different identification methods, a correction procedure, and a procedure for checking possible failures of the correction step.

(4) FIGS. 3A-B illustrates impacts of the effect of auto-fluorescent compounds on fluorescence emission data using FIDA read-outs and presents a correction procedure.

(5) FIGS. 4A-B illustrates screening for activating compounds using polarization and 2D-FIDA read-outs and the identification of auto-fluorescent and quenching compounds, a correction procedure, and a check of the correction step in the case of auto-fluorescence.

(6) FIGS. 5A-B illustrates an identification of auto-fluorescent and quenching compounds in high-throughput-screening (HTS).

(7) FIG. 6 shows a schematic diagram of a preferred system for detecting the impacts of interfering effects on experimental data resulting from fluorescence measurements.

EXAMPLES

(8) The measurements presented in the following were performed on an epi-illuminated confocal fluorescence microscope as described in [P. Kask, K. Palo, N. Fay, L. Brand, Ü. Mets, D. Ullmann, J. Jungmann, J. Pschorr, K. Gall (2000) Biophys. J, 78, 1703-1713]. A polarized continuous-wave (cw) laser either at 543 nm or 633 nm was used to excite a fluorophore (Tetramethyl-Rhodamine (TAMRA) for 543 nm excitation, MR-121 for 633 nm excitation) alone or covalently linked to a molecule of interest. Detection was performed with a single detector or two detectors (Avalanche-Photo-Diode, APD) monitoring the fluorescence light emitted with parallel or perpendicular polarization with respect to the polarization of the exciting light. While the one-detector set-up was used for the fluorescence data analysis via FIDA, the two-detector set-up was used for the determination of the polarization or anisotropy values and analysis via 2D-FIDA.

Example 1

(9) In a first measurement series, different amounts of various water soluble chemical compounds were added to an aqueous TAMRA solution (about 15 nM, resulting in 96 different samples) and the total intensity, I.sub.tot, as well as the polarization, P, were determined for the 96 different samples (two-detector set-up and measurement time of two seconds). I.sub.tot and P were calculated from the intensities with parallel, I.sub.P, and perpendicular, I.sub.S, polarization with respect to the exciting light.
I.sub.tot=I.sub.P+2I.sub.SP=(I.sub.P−I.sub.S)/(I.sub.P+I.sub.S)×1000
Subsequently, 4 different methods were applied to identify samples with deteriorated signal as presented in FIG. 1.

(10) FIG. 1A shows the one-dimensional distribution of I.sub.tot over all 96 samples (dotted line) together with the thresholds (vertical lines) for identification of deteriorated signal. The thresholds were set according to the median, med(I.sub.tot)=1119.9 kHz, and the median like standard deviation, s*(I.sub.tot)=145.8 kHz, of I.sub.tot from all 96 samples (s* has been described previously); threshold(I.sub.tot)=med(I.sub.tot)±3×s*(I.sub.tot). All samples that exhibited a value of I.sub.tot outside these thresholds were identified to be deteriorated, in this case 12 samples.

(11) FIG. 1B shows the one-dimensional distribution of the mathematically transformed total intensity, TI, together with a Gaussian fit to the distribution (gray line; G(TI)=A×exp[−(TI−TI.sub.0).sup.2/(2σ.sup.2)] with the variables A, TI.sub.0, and σ subject to fitting and resulting in A=13.9, TI.sub.0=0.44, and σ=0.3) and the thresholds (vertical lines) for identification of deteriorated signal. The mathematical transformation was performed according to the steps; (a) calculation of the median, med(I.sub.tot), of I.sub.tot from all 96 samples, (b) calculating the difference, res(I.sub.tot)=I.sub.tot−med(I.sub.tot), for each sample, (c) determination of the mean, mean(res(I.sub.tot)), and the standard deviation, s(res(I.sub.tot)), of res(I.sub.tot) from all 96 samples, (d) calculating TI=[res(I.sub.tot)−mean(res(I.sub.tot))]/s(res(I.sub.tot)). The thresholds were determined from the resulting value of σ and TI.sub.0 of the Gaussian fit; threshold(TI)=TI.sub.0±3×σ. All samples that exhibited a value of TI outside these thresholds were identified to be deteriorated, in this case 13 samples.

(12) FIG. 1C shows the one-dimensional distribution of the slightly different mathematically transformed total intensity, TI*, together with a Gaussian fit to the distribution (gray line; G(TI*)=A×exp[−(TI*−TI.sub.0*).sup.2/(2σ*.sup.2)] with the variables A, TI.sub.0*, and σ* subject to fitting and resulting in A=3.788, TI.sub.0*=0.266, and σ=1.11) and the thresholds (vertical lines) for identification of deteriorated signal. In this case, the mathematical transformation was performed according to the steps; (a) calculation of the median, med(I.sub.tot), of I.sub.tot from all 96 samples, (b) calculating the difference, res(I.sub.tot)=I.sub.tot−med(I.sub.tot), for each sample, (c) determination of the median, median(res(I.sub.tot)), and the median like standard deviation, s*(res(I.sub.tot)), of res(I.sub.tot) from all 96 samples, (d) calculating TI*=[res(I.sub.tot)−median(res(I.sub.tot))]/s*(res(I.sub.tot)). The thresholds were determined from the resulting value of σ* and TI.sub.0* of the Gaussian fit; threshold(TI*)=TI.sub.0*±3×σ*. All samples that exhibited a value of TI outside these thresholds were identified to be deteriorated, in this case 12 samples.

(13) FIG. 1D represents the two-dimensional distribution of joint total intensity-polarization pairs, (I.sub.tot, P), from all 96 samples (black dots) together with the thresholds (black lines) for identification of deteriorated signal. The thresholds were set according to the median, med(I.sub.tot)=1119.9 kHz and med(P)=30.72, and the median like standard deviation, s*(I.sub.tot)=145.8 kHz and s*(P)=5.15, of I.sub.tot and P respectively from all 96 samples; threshold(I.sub.tot)=med(I.sub.tot)±3×s*(I.sub.tot) and threshold(P)=med(P)±3×s*(P). All samples that exhibited a value of I.sub.tot or P outside these thresholds were identified to be deteriorated, in this case 13 samples.

(14) By one or the other method, the same conspicuous samples were identified by a decreased intensity and an increased polarization. To find the reason behind this deterioration, one conspicuous compound was added at rising concentrations to the dye solution. The measured total intensity, I.sub.tot, and polarization, P, are shown in FIGS. 1E and F, respectively. One clearly observes, that the deterioration was caused by a quenching interference of the compound to the fluorescence emission of the dye, which was a decreasing intensity accompanied by an increase in polarization.

Example 2

(15) In a second measurement series, the binding of a small MR-121-labeled peptide to the SH2-domain of the Grb2-protein was monitored by a change in the fluorescence polarization, P, of the MR-121 fluorescence emission (two-detector set-up, measurement time of ten seconds). In the different samples, the binding was increasingly inhibited by the titration of unlabeled peptide. Thereby, nine different concentrations of unlabeled peptide were measured five times each, i.e. 45 samples were observed. While one set of 45 samples only contained the assay components (labeled and unlabeled peptide and protein), auto-fluorescent compounds (1 μM Rhodamine 800) had been added to another set of 45 samples. 2D-FIDA with a one-component fit was applied to the signal of all samples. This analysis yielded values of concentration, c, brightness, q.sub.1 and q.sub.2, of each detection channel monitoring the light emission with parallel and perpendicular polarization with respect to the exciting light, respectively, and of chi.sup.2, which is the quality parameter of the fit (as presented previously). The total signal intensity was once again calculated from the intensities with parallel, I.sub.P, and perpendicular, I.sub.S, polarization with respect to the exciting light, while the polarization, P, was calculated from q.sub.1 and q.sub.2.
I.sub.tot=I.sub.P+2I.sub.SP=(q.sub.1−q.sub.2)/(q.sub.1+q.sub.2)×1000
In addition, two control samples were measured ten times each, resulting as well in values of c, q.sub.1 and q.sub.2, chi.sup.2, I.sub.tot, and P. The ten high control samples, which contained only labeled peptide and protein (resulting in mainly bound labeled peptide), resulted in values of c(high), q.sub.1(high) and q.sub.2(high), chi.sup.2(high), I.sub.tot(high), and P(high). The low control, which contained labeled peptide, excess of unlabeled peptide, and protein (resulting in totally inhibited binding, thus mainly unbound labeled peptide), resulted in values of c(low), q.sub.1(low) and q.sub.2(low), chi.sup.2(low), I.sub.tot(low), and P(low). This enabled the calculation of the normalized total intensity, NI, and the inhibition, Inh, for each measurement X.
NI(X)=[I.sub.tot(X)−I.sub.tot(low)]/[I.sub.tot(high)−I.sub.tot(low)]×100
Inh(X)=[P(high)−P(X)]/[P(high)−P(low)]×100

(16) FIGS. 2A and B present two different methods how samples with auto-fluorescence can be identified.

(17) FIG. 2A plots the two-dimensional distribution of joint normalized total intensity-inhibition pairs, (NI, Inh), from both sets of 45 samples (black dots), the high samples (gray cross), and the low samples (gray circles) together with the threshold functions (black lines) for said identification. The thresholds were set by the mean, m(NI,high)=0, m(Inh,high)=0, m(NI,low)=100, and m(Inh,low)=100, and the standard deviation, s(NI,high)=19.7, s(Inh,high)=5.8, s(NI,low)=17.9, and s(Inh,low)=2.7, of NI and Inh from all ten high and low samples, respectively; t_1(Inh)=m(Inh,high)−3×s(Inh,high), t_2(Inh)=m(Inh,low)+3×s(Inh,low), t_3(NI)=m(NI,low)−3×s(NI,low), t_4(NI)=m(NI,low)+3×s(NI,low), t_5(NI)=m(NI,high)−3×s(NI,high), t_6(NI)=m(NI,high)+3×s(NI,high),

(18) An auto-fluorescent sample was identified if its read-out, NI or Inh, obeyed one of the following conditions; Inh<t_1(Inh), Inh>t_2(Inh), NI<a.sub.1+b.sub.1×Inh, or NI>a.sub.2+b.sub.2×Inh, with a.sub.1=t_3(NI)−b.sub.1×m(Inh,low), a.sub.2=t_4(NI)−b.sub.2×m(Inh,low), b.sub.1=[t_3(NI)−t_5(NI)]/[m(Inh,low)−m(Inh,high)], and b.sub.2=[t_4(NI)−t_6(NI)]/[m(Inh,low)−m(Inh,high)].
In this way, all 45 samples with added auto-fluorescence were identified.

(19) FIG. 2B plots the two-dimensional distribution of chi.sup.2-inhibition pairs, (chi.sup.2, Inh), from both sets of 45 samples (black dots), the high samples (gray cross), and the low samples (gray circles) together with the threshold functions (black lines) for said identification. The thresholds were set by the mean, m(chi.sup.2,high)=1.37, m(Inh,high)=0, m(chi.sup.2,low)=0.71, and m(Inh,low)=100, and the standard deviation, s(chi.sup.2,high)=0.24, s(Inh,high)=5.8, s(chi.sup.2,low)=0.06, and s(Inh,low)=2.7, of chi.sup.2 and Inh from all ten high and low samples, respectively; t_1(Inh)=m(Inh,high)−3×s(Inh,high), t_2(Inh)=m(Inh,low)+3×s(Inh,low), t_3(chi.sup.2)=m(chi.sup.2,low)−5×s(chi.sup.2,low), t_4(chi.sup.2)=m(chi.sup.2,low)+5×s(chi.sup.2,low), t_5(chi.sup.2)=m(chi.sup.2,high)−5×s(chi.sup.2,high), t_6(chi.sup.2)=m(chi.sup.2,high)+5×s(chi.sup.2,high),
An auto-fluorescent sample was identified if its read-out, chi.sup.2 or Inh, obeyed one of the following conditions; Inh<t_1(Inh), Inh>t_2(Inh), chi.sup.2<a.sub.1+b.sub.1×Inh, or chi.sup.2>a.sub.2+b.sub.2×Inh, with a.sub.1=t_3(chi.sup.2)−b.sub.1×m(Inh,low), a.sub.2=t_4(chi.sup.2)−b.sub.2×m(Inh,low), b.sub.1=[t_3(chi.sup.2)−t_5(chi.sup.2)]/[m(Inh,low)−m(Inh,high)], and b.sub.2=[t_4(chi.sup.2)−t_6(chi.sup.2)]/[m(Inh,low)−m(Inh,high)].
Once again, all 45 samples with added auto-fluorescence were identified.

(20) FIG. 2C shows the titration curves for the pure samples (black dots) and the samples with added auto-fluorescence (transparent dots), i.e. the curve shows the change of the polarization, P, with increasingly added unlabeled peptide (the error bars were obtained from the results of the five samples observed for each titration point). The effect of the auto-fluorescence on the detected fluorescence becomes evident by a decreased polarization value. Fitting the auto-fluorescent samples with an additional pair of floating FIDA-background values resulted in a correction of the read-out. This is demonstrated in FIG. 2D, where the corrected read-out of the auto-fluorescent samples coincides with the read-out of the pure samples.

(21) FIG. 2E demonstrates the test procedure of the correction step. Similar to FIG. 2B, it plots the two-dimensional distribution of corrected chi.sup.2-inhibition pairs, (chi.sup.2, Inh), from both sets of 45 samples (black dots), the low samples (gray cross), and the high samples (gray circles) together with the threshold functions (black lines) for said identification. The thresholds were identically set as in FIG. 2B.

(22) The failure of the correction step was identified if the according read-out, chi.sup.2 or Inh, obeyed one of the following conditions; Inh>t_2(Inh), chi.sup.2<a.sub.1+b.sub.1×Inh, or chi.sup.2>a.sub.2+b.sub.2×Inh, with a.sub.1=t_3(chi.sup.2)−b.sub.1×m(Inh,low), a.sub.2=t_4(chi.sup.2)−b.sub.2×m(Inh,low), b.sub.1=[t_3(chi.sup.2)−t_5(chi.sup.2)]/[m(Inh,low)−m(Inh,high)], and b.sub.2=[t_4(chi.sup.2)−t_6(chi.sup.2)]/[m(Inh,low)−m(Inh,high)].
In this way, only one failure of the correction step was identified.

Example 3

(23) In a third measurement series, the binding of a TAMRA-labeled ligand to membrane vesicles with the appropriate G-protein coupled receptors was monitored using FIDA (one-detector set-up, measurement time of two seconds). The ligand bound to the vesicles can be distinguished from the free ligand by an increase in the fluorescence brightness, q, since the vesicles can bind several ligands. In FIDA, these two components were distinguished in a two-component fit and their brightness, q(ligand) and q(vesicle), and concentration values, c(ligand) and c(vesicle), were determined. For every sample the binding degree was determined according to the equation,
binding degree=c(vesicle)×q(vesicle)/[c(vesicle)×q(vesicle)+c(ligand)×q(ligand)].

(24) 48 high control and 48 low control sample were measured. The high control contained both, labeled ligand and vesicles, while the low control solely contained labeled ligand. In a first set of measurements, the pure 96 samples were observed. In additional sets of measurement, the 96 samples were observed after adding different amounts of auto-fluorescent substances (0.05 μM, 0.5 μM, and 1 μM of the dye C682). The binding degree resulting from the two-component FIDA fit is shown in FIG. 3A. The apparently decreased binding degree shows the effect of the increasingly added auto-fluorescent substances.

(25) For the correction, a three-component FIDA analysis was performed on the same fluorescence data sets. Thereby, an additional component with floating concentration, c(auto-fluorescence), and fixed brightness value, q(auto-fluorescence)=1 kHz, was added to the two-component fit of FIG. 3A. The fixed brightness value was rather low compared to the mean brightness values obtained for the free ligand, q(ligand)=9 kHz, and the ligand bound to the vesicle, q(vesicle)=1350 kHz. The resulting values of the binding degree coincides with that of the pure samples, which indicates the success of the correction procedure. However, a decreased accuracy of the determination of the binding degree becomes evident by the increased error bars, which is due to the presence of the interfering auto-fluorescence as well as the correction procedure. Therefore, it is recommendable to apply this correction step solely to those samples which are identified to emit auto-fluorescence.

Example 4

(26) In a fourth measurement series, 96 different compounds were tested for the activation of a DNA-binding protein. Upon activation, the protein was able to bind the single DNA strand. Since the DNA strand was labeled with TAMRA, the activation was accompanied by an increase in the polarization, P. A positive compound, which activated the protein, should therefore result in an increase of polarization, P. To check the reactivity of the compounds, the polarization read-out was compared to that of positive and negative controls. While the negative control just like a non-activating compound comprised the unbound DNA strand (low polarization), the positive control just like an activating compound comprised the DNA-peptide complex (high polarization). As in the previous example 2, the measurements were performed with two detectors monitoring the different polarization directions of the light emission and analyzed using 2D-FIDA regarding only one fluorescent component. This resulted in values of the intensity, I.sub.P and I.sub.S, as well as of the brightness, q.sub.1 and q.sub.2, of the fluorescence with parallel and perpendicular polarization with respect to the polarization of the exciting light, respectively, and of the mean concentration, c, of the fluorescent component. This enabled the calculation of the total intensity, I.sub.tot, the total brightness, q.sub.tot, the activation, Act, as well as the normalized total signal, NI.
I.sub.tot=I.sub.P+2I.sub.Sq.sub.tot=q.sub.1+q.sub.2P=(q.sub.1−q.sub.2)/(q.sub.1+q.sub.2)×1000
NI(X)=[I.sub.tot(X)−I.sub.tot(pos)]/[I.sub.tot(neg)−I.sub.tot(pos)]×100
Act(X)=[P(X)−P(neg)]/[P(pos)−P(neg)]×100

(27) The whole experiment included the measurement (two second duration) of 96 different compounds added to the assay (labeled DNA and protein) as well as nine positive controls and 96 negative controls.

(28) For the identification of possible auto-fluorescent or quenching compounds, FIG. 4A plots the two-dimensional distribution of joint normalized total intensity-activation pairs, (NI, Act), from the 96 compound samples (black dots), the 96 negative control samples (gray cross), and the six positive control samples (gray circles) together with the threshold functions (black lines) for said identification. In the same way as in FIG. 2A, the thresholds were set by the mean, m(NI,pos)=0, m(Act,pos)=100, m(NI,neg)=100, and m(Act,neg)=0, and the standard deviation, s(NI,pos)=4.1, s(Act,pos)=4.7, s(NI,neg)=9.3, and s(Act,neg)=6.2, of NI and Act from all six positive and 96 negative control samples, respectively; t_1(Act)=m(Act,neg)−3×s(Act,neg), t_2(Act)=m(Act,pos)+3×s(Act,pos), t_3(NI)=m(NI,neg)−3×s(NI,neg), t_4(NI)=m(NI,neg)+3×s(NI,neg), t_5(NI)=m(NI,pos)−3×s(NI,pos), t_6(NI)=m(NI,pos)+3×s(NI,pos).

(29) An auto-fluorescent compound sample was identified if its read-out, NI or Act, was above the upper threshold line, i.e. obeyed the following condition;
NI>a.sub.2+b.sub.2×Act, with a.sub.2=t_4(NI)−b.sub.2×m(Act,neg), and b.sub.2=[t_4(NI)−t_6(NI)]/[m(Act,neg)−m(Act,pos)].
In this way, 67 compound samples were identified to be auto-fluorescent.

(30) A quenching compound sample was identified if its read-out, NI or Act, was below the lower threshold line or elsewhere to the left or right of the two vertical lines, i.e. obeyed one of the following conditions and was not auto-fluorescent;
Act<t_1(Act),Act>t_2(Act), or NI<a.sub.1+b.sub.1×Act, with a.sub.1=t_3(NI)−b.sub.1×m(Act,neg), and, b.sub.1=[t_3(NI)−t_5(NI)]/[m(Act,neg)−m(Act,pos)].
In this way, two compound samples were identified to be quenching and taken away from further analysis (bad data points).

(31) In a second step, the correction procedure was applied to the fluorescence data from the compound samples identified as being auto-fluorescent (while the results of the analysis were maintained for the valid compound samples). The correction procedure comprised a 2D-FIDA fit regarding one component as before and in addition two floating FIDA-background values as already applied in example 2. The success of the correction procedure was checked (see FIG. 4B). Similar to FIG. 2B, it plots the two-dimensional distribution of the corrected total brightness-activation pairs, (q.sub.tot, Act), from the 94 left samples (black dots), the 96 negative samples (gray cross), and the six positive samples (gray circles) together with the threshold functions (black lines) for the identification of failures of the correction algorithm or the analysis in principle. The thresholds were set by the mean, m(q.sub.tot,pos)=68.0, m(Act,pos)=100, m(q.sub.tot,neg)=58.4, and m(Act,neg)=0, and the standard deviation, s(q.sub.tot,pos)=3.7, s(Act,pos)=4.7, s(q.sub.tot,neg)=3.5, and s(Act,neg)=6.2, of q.sub.tot and Act from all six positive and 96 negative control samples, respectively; t_1(Act)=m(Act,neg)−3×s(Act,neg), t_2(Act)=m(Act,pos)+3×s(Act,pos), t_3(q.sub.tot)=m(q.sub.tot,neg)−5×s(q.sub.tot,neg), t_4(q.sub.tot)=m(q.sub.tot,neg)+5×s(q.sub.tot,neg), t_5(q.sub.tot)=m(q.sub.tot,pos)−5×s(q.sub.tot,pos), t_6(q.sub.tot)=m(q.sub.tot,pos)+5×s(q.sub.tot,pos).
The said failure was identified if the according read-out, q.sub.tot or Act, obeyed one of the following conditions;
Act>t_2(Act),q.sub.tot<a.sub.1+b.sub.1×Act, or q.sub.tot>a.sub.2+b.sub.2×Act, with a.sub.1=t_3(q.sub.tot)−b.sub.1×m(Act,low), a.sub.2=t_4(q.sub.tot)−b.sub.2×m(Act,low), b.sub.1=[t_3(q.sub.tot)−t_5(chi.sup.2)]/[m(Act,low)−m(Act,high)], and b.sub.2=[t_4(q.sub.tot)−t_6(q.sub.tot)]/[m(Act,low)−m(Act,high)].
In this way, eight failures of the whole analysis process were identified.

(32) Using the identification step and correction procedure, together with the failure check, one can not only exclude possible false positive compounds (i.e., apparently activating in this case) due to auto-fluorescence or quenching, but also correct the read-out for auto-fluorescent compounds. In a drug discovery process, this does not only save precious money and time, but also helps to find possible positive and auto-fluorescent drug candidates which would otherwise be lost.

Example 5

(33) In a further measurement series, the identification step was applied to a high-throughput-screening (HTS) run. In this HTS run different compounds were tested for the inhibition of the dephosphorylation of a phosphotyrosine-containing peptide by an appropriate protein tyrosine phosphatase. An antibody was used in this experiment which only binds to the phosphorylated peptide. Since the peptide was fluorescently labeled, binding of the antibody to the phosphorylated peptide increased the polarization, P, of this complex. Therefore, dephosphorylation resulted in a loss of antibody binding and concomitant decrease of polarization. A positive compound, which inhibited the dephosphorylation, should therefore result in an increase of polarization, P. To check the reactivity of the compounds, the polarization read-out was compared to that of positive and negative controls. While the negative control just like a non-inhibiting compound comprised the dephosphorylated peptide (low polarization), the positive control just like an inhibiting compound comprised the antibody-peptide complex (high polarization). As in the previous examples 2 and 4, the measurements were performed with two detectors monitoring the different polarization directions of the light emission and analyzed using 2D-FIDA with a one-component fit. As outlined, this enabled the calculation of the inhibition, Inh, as well as the normalized total signal, NI.

(34) 6144 different compounds were added to the assay (labeled peptide, antibody, and phosphatase) and distributed on four different nanotiter-plates with 2080 wells each. Furthermore, each plate contained 24 positive and 24 negative control samples. The HTS run was performed by measuring each sample once for one second. The identification step for auto-fluorescent or quenching compounds is outlined in FIG. 5A. In the same way as outlined in detail in FIG. 2A and FIG. 4A, the threshold conditions for the identification were set individually for each plate according to the mean values and standard deviations of Inh and NI of the positive and negative controls (mean±3×standard deviation).

(35) This is shown in FIG. 5A for one of the four plates, where the threshold lines (black lines) are drawn such as in FIGS. 2A and 4A. The compound samples exhibiting a read-out pair of (Inh,NI) above the upper line were classified as auto-fluorescent compounds (gray cross), while compound samples exhibiting a read-out pair of (Inh,NI) below the lower line were classified as quenching compounds (gray circles). Valid compound samples as well as positive and negative controls (black circles) lie in between the threshold lines. In this way, 1313 compounds were classified to be valid, 166 (10.8%) to be quenching, and 57 (3.7%) to be auto-fluorescent.

(36) FIG. 5B plots the pairs (Inh, NI) from all four plates. The identification step was performed for each plate independently. In this way, 4966 valid (black circles), 819 quenching (13.3%, gray circles), and 365 auto-fluorescent compounds (5.9%, gray cross) were identified in this HTS run.

(37) Since the inhibition values obtained from the samples with auto-fluorescent and quenching compounds in a lot of cases pretend a positive inhibiting property of the according compound (compare FIG. 5), this identification step avoids the detection of false positives and helps to save precious money and time in the drug discovery process when using HTS.

Example 6

(38) FIG. 6 shows a schematic diagram of a preferred system for detecting the impacts of interfering effects on experimental data resulting from fluorescence measurements. Preferably, the fluorescence measurements are performed with a confocal epi-illuminated microscope.

(39) Means in an inspection station (2) support one or a plurality of samples (e.g. a moveable microscope table with a 4×6-, 96-, 384-, 1536-, or 2080-well glass bottom well plate, the wells being filled with the samples). Preferably, the samples comprise dye-labeled molecules at a rather low concentration below 20 nM. Furthermore, the inspection station can preferably be moved with respect to the rest of the system.

(40) One or a plurality of light sources (3) serve for the excitation of fluorescence emission within the sample. Preferably, the light sources are linearly polarized lasers at wavelengths between 350 and 700 nm, which are either continuous wave or pulsed in the case of fluorescence lifetime measurements. In order to guide the excitation light onto the sample, it is reflected by a mirror (4) and focused into the sample by a lens (5). Preferably, the mirror is dichroitic, i.e. it reflects the excitation light and transmits the fluorescence light. Preferably, the lens is an objective lens, which focuses the light to an almost diffraction limited spot of about 1 μm diameter, thereby causing fluorescence emission within the sample.

(41) For the detection of the fluorescence emission, the system comprises an optical set-up which directs the fluorescence on at least one of the detectors (9, 10). The fluorescence of the sample is collected by the same lens (5), transmits the mirror (4), and is focused onto a pinhole (6). The pinhole, which preferably has a diameter of 10 to 200 μm, blocks out-of-focus light and transmits only fluorescence light, which is emitted within the focal part of the excitation light, i.e. a volume of about fL-size. After the pinhole, the fluorescence is guided to one or more detectors (9, 10). It can be split into several components by one or more mirrors (7), which preferably split the fluorescence into its components of different polarization and/or color. Before impinging onto the detectors, the fluorescence radiation can pass optical filters (8), which preferably transmit the fluorescence and block unwanted radiation such as scattering from the solvent. Preferably, the detectors (9, 10) are avalanche photodiodes, which convert the impinged fluorescence radiation into an electrical signal with a very high efficiency.

(42) A signal processing unit (11) converts the electrical signal of the one or the plurality of detectors into experimental data, which is preferably a stream of fluorescence photon counts. In further processing steps, the unit (11) determines the values of one or a plurality of identification parameters from the experimental data, which is e.g. the amount of detected fluorescence—e.g. the fluorescence intensity, the number of counts and/or the count-rate—, a ratio of fluorescence intensities at selected wavelengths, a ratio of fluorescence intensities at different polarization directions, a fluorescence anisotropy, a fluorescence polarization, a fluorescence lifetime, a rotational correlation time, a diffusion constant, a concentration of fluorophores, a specific fluorescence brightness, and/or a function of these. For this determination, the signal processing unit uses preferably analysis techniques such as FCS, 1D- and/or 2D-FIDA, FILDA, fluorescence lifetime and/or time-resolves anisotropy analysis, and/or FIMDA. Furthermore, the signal processing unit (11) might coordinate the movement of the sample support within the inspection station. The signal processing unit preferably contains a storage unit, which stores the determined values of identification parameters in relation to the respective position of the sample support. The signal processing unit (11) as well creates an histogram or distribution of the values of the identification parameters and determines thresholds for the values of the identification parameters, which thresholds are indicative for the impact of interfering effects. It analyzes the values of the identification parameters for the different positions of the sample support within the inspection station and determines whether or not these values fulfill criteria with respect to the thresholds. It also supplies as output information those data which are influenced and/or not influenced by the interfering effects. Furthermore, the unit (11) includes means for correcting the data for the impact of the interfering effect and means for the conductance of a control step to check the success of the correction.