Method and Device for Analyzing Biological Material

Abstract

A method for analyzing biological material includes reading in a measurement signal, a first reference signal and a second reference signal. The method further includes determining noise in the measurement signal in order to produce noise data, applying the noise data to the first reference signal and to the second reference signal in order to generate an adjusted first reference signal and an adjusted second reference signal, and transforming the measurement signal, the adjusted first reference signal, and the adjusted second reference signal into a frequency distribution form in order to produce a measurement signal distribution, a first reference distribution and a second reference distribution. Additionally the method includes performing a cluster analysis using the measurement signal distribution, the first reference distribution, and the second reference distribution to determine, in accordance with a result of the cluster analysis, whether the biological material has the first property or the second property.

Claims

1. A method for analyzing biological material, the method comprising: reading a measurement signal representing acquired optofluidic data of the biological material, a first reference signal representing first optofluidic model data corresponding to a first property of the biological material, and a second reference signal representing second optofluidic model data corresponding to a second property of the biological material; ascertaining a noise of the measurement signal, to generate noise data; applying the noise data to the first reference signal and to the second reference signal, to generate an adapted first reference signal and an adapted second reference signal; transforming the measurement signal, the adapted first reference signal, and the adapted second reference signal into a frequency distribution form to generate a measurement signal distribution, a first reference distribution, and a second reference distribution; and performing a cluster analysis using the measurement signal distribution, the first reference distribution, and the second reference distribution to establish as a function of a result of the cluster analysis whether the biological material has the first property or the second property.

2. The method as claimed in claim 1, further comprising: acquiring the acquired optofluidic data of the biological material to provide the measurement signal.

3. The method as claimed in claim 1, wherein: the measurement signal represents optofluidic data of the biological material acquired by a quantitative and/or qualitative polymerase chain reaction, the first reference signal has a sigmoid curve, the second reference signal has a linear curve, in the reading of the first and second reference signals, the first property of the biological material results in an amplification of at least one target molecule of the biological material and the second property of the biological material results in an absence of the amplification of the at least one target molecule the transforming of the measurement signal, the adapted first reference signal, and the adapted second reference signal includes transforming the adapted first and second reference signals in such a way that the first reference distribution is a bimodal distribution and the second reference distribution is a unimodal distribution.

4. The method as claimed in claim 1, wherein the performing of the cluster analysis includes using a k-means algorithm having a predefined distance measure, wherein the first reference distribution represents a first cluster, the second reference distribution represents a second cluster, and the result of the cluster analysis specifies whether, in consideration of the predefined distance measure, the measurement signal distribution falls into the first cluster or into the second cluster.

5. The method as claimed in claim 1, wherein, in the ascertaining of the noise, the noise of the measurement signal is ascertained over a curve of the analysis and/or via a sliding window process and/or using a noise measure to generate as the noise data a functional relationship between the noise and the curve of the analysis.

6. The method as claimed in claim 1, wherein the applying of the noise data includes adding random numbers or pseudorandom numbers dependent on the noise data to the first and second optofluidic model data of the first and second reference signals.

7. The method as claimed in claim 1, further comprising: scaling the read measurement signal by projecting absolute values on a predefined value interval, wherein the ascertaining of the noise includes ascertaining the noise of the scaled measurement signal.

8. A device for analyzing biological material, the device comprising: a control unit configured to: read a measurement signal representing acquired optofluidic data of the biological material, a first reference signal representing first optofluidic model data corresponding to a first property of the biological material, and a second reference signal representing second optofluidic model data corresponding to a second property of the biological material; ascertain a noise of the measurement signal to generate noise data; apply the noise data to the first reference signal and to the second reference signal to generate an adapted first reference signal and an adapted second reference signal; transform the measurement signal, the adapted first reference signal, and the adapted second reference signal into a frequency distribution form to generate a measurement signal distribution, a first reference distribution, and a second reference distribution; and perform a cluster analysis using the measurement signal distribution, the first reference distribution, and the second reference distribution to establish as a function of a result of the cluster analysis whether the biological material has the first property or the second property.

9. A computer program configured to execute and/or activate a control unit to: read a measurement signal representing acquired optofluidic data of a biological material, a first reference signal representing first optofluidic model data corresponding to a first property of the biological material, and a second reference signal representing second optofluidic model data corresponding to a second property of the biological material; ascertain a noise of the measurement signal to generate noise data; apply the noise data to the first reference signal and to the second reference signal to generate an adapted first reference signal and an adapted second reference signal; transform the measurement signal, the adapted first reference signal, and the adapted second reference signal into a frequency distribution form to generate a measurement signal distribution, a first reference distribution, and a second reference distribution; and perform a cluster analysis using the measurement signal distribution, the first reference distribution, and the second reference distribution to establish as a function of a result of the cluster analysis whether the biological material has the first property or the second property.

10. A machine-readable storage medium, comprising: at least one memory on which the computer program as claimed in claim 9 is stored.

Description

[0026] Exemplary embodiments of the approach presented here are illustrated in the drawings and explained in greater detail in the following description. In the figures:

[0027] FIG. 1 shows a schematic illustration of a device according to one exemplary embodiment;

[0028] FIG. 2A to FIG. 2C show schematic signal diagrams according to one exemplary embodiment;

[0029] FIG. 3A to FIG. 3C show schematic signal diagrams according to one exemplary embodiment;

[0030] FIG. 4A to FIG. 4D show schematic signal diagrams according to one exemplary embodiment;

[0031] FIG. 5A and FIG. 5B show schematic signal diagrams according to one exemplary embodiment;

[0032] FIG. 6A and FIG. 6B show schematic signal diagrams according to one exemplary embodiment;

[0033] FIG. 7A to FIG. 7C show schematic signal diagrams according to one exemplary embodiment;

[0034] FIG. 8 shows a flow chart of an evaluation process according to one exemplary embodiment; and

[0035] FIG. 9 shows a flow chart of a method according to one exemplary embodiment.

[0036] In the following description of advantageous exemplary embodiments of the present invention, identical or similar reference signs are used for elements shown in the various figures and acting similarly, wherein a repeated description of these elements is omitted.

[0037] FIG. 1 shows a schematic illustration of a device 100 according to one exemplary embodiment. The device 100 is designed to execute an analysis of biological material. The device 100 can also be referred to as an analysis device 100. The biological material includes, for example, genetic material. The device 100 is embodied in particular as a chip laboratory. The device 100 is designed here, more precisely, to execute a quantitative real-time PCR (qPCR; PCR=polymerase chain reaction). The qPCR is executed here in a plurality of cycles. For example, a presence of ultrasmall quantities of a specific DNA section (DNA=deoxyribonucleic acid) is detected quantitatively and/or qualitatively.

[0038] The device 100 includes at least one microfluidic apparatus 110. The biological material is introducible, for example, in a cartridge 115 into the device 100 or microfluidic apparatus 110. The cartridge 115 is receivable in the device 100 or the microfluidic apparatus 110. The at least one microfluidic apparatus 110 is designed to acquire optofluidic data of the biological material. For this purpose, the microfluidic apparatus 110 is designed, for example, to acquire light reflected and/or emitted from the biological material.

[0039] Furthermore, the at least one microfluidic apparatus 110 is designed to provide a measurement signal 120, for example, using the light originating from the biological material. The measurement signal 120 represents, according to one exemplary embodiment, acquired optofluidic data of the biological material. To excite the biological material, the microfluidic apparatus 110 is designed according to one exemplary embodiment to irradiate the biological material with electromagnetic radiation, for example light.

[0040] The device 100 moreover includes a control unit 130, which is also referred to as a control apparatus 130 or control device 130 for analyzing the biological material. The control unit 130 is connected to the at least one microfluidic apparatus 110 in a manner capable of signal transmission. The control unit 130 includes, according to the exemplary embodiment shown here, a read apparatus 140, an ascertainment apparatus 150, an application apparatus 160, a transformation apparatus 170, and a performance apparatus 180.

[0041] The read apparatus 140 is designed to read the measurement signal 120. Furthermore, the read apparatus 140 is designed to read a first reference signal 131 and a second reference signal 132. The first reference signal 131 represents first optofluidic model data, which correspond to a first property of the biological material, and the second reference signal 132 represents second optofluidic model data, which correspond to a second property of the biological material. The read apparatus 140 is connected in a manner capable of signal transmission to the ascertainment apparatus 150, to the application apparatus 160, and to the transformation apparatus 170.

[0042] The ascertainment apparatus 150 is designed to ascertain a noise of the measurement signal 120. Furthermore, the ascertainment apparatus 150 is designed to generate noise data 155, which represent the ascertained noise of the measurement signal 120. The ascertainment apparatus 150 is connected in a manner capable of signal transmission to the application apparatus 160.

[0043] The application apparatus 160 is designed to apply the ascertained noise data 155 to the first reference signal 131 and to the second reference signal 132 to generate an adapted first reference signal 161 and an adapted second reference signal 162. The application apparatus 160 is connected in a manner capable of signal transmission to the transformation apparatus 170.

[0044] The transformation apparatus 170 is designed to transform the measurement signal 120, the adapted first reference signal 161, and the adapted second reference signal 162 into a frequency distribution form to generate a measurement signal distribution 175, a first reference distribution 171, and a second reference distribution 172. The transformation apparatus 170 is connected in a manner capable of signal transmission to the performance apparatus 180.

[0045] The performance apparatus 180 is designed to perform a cluster analysis using the measurement signal distribution 175, the first reference distribution 171, and the second reference distribution 172. More precisely, the performance apparatus 180 is designed to perform the cluster analysis to establish as a function of a result of the cluster analysis whether the biological material has the first property or the second property. The result of the cluster analysis represents an association of the measurement signal distribution 175 with the first reference distribution 171 or the second reference distribution 172.

[0046] The control unit 130 is furthermore designed to output or provide an analysis signal 190. The analysis signal 190 represents a result of the analysis. For example, the analysis signal comprises an item of information about the property of the biological material.

[0047] According to one exemplary embodiment, the control unit 130 is also designed to scale the read measurement signal 120, in particular by projecting absolute values of the measurement signal 120 on a predefined value interval. The scaling is optionally executable by means of the read apparatus 140 or the ascertainment apparatus 150. In this case, the ascertainment apparatus 150 is designed to ascertain the noise on the basis of the scaled measurement signal.

[0048] In particular, the processes executed by the control unit 130 are also further clarified with reference to the following figures.

[0049] FIG. 2A, FIG. 2B, and FIG. 2C show schematic signal diagrams according to one exemplary embodiment. FIG. 2A shows the first reference signal 131 from FIG. 1. The first reference signal 131 has a sigmoid curve or sigma curve. FIG. 2B shows the measurement signal 120 from FIG. 1. The measurement signal 120 is subject to noise. FIG. 2C shows the second reference signal 132 from FIG. 1. The second reference signal 132 has a linear curve.

[0050] In other words, FIG. 2A, FIG. 2B, and FIG. 2C show a decision problem of a qPCR curve evaluation. If a qPCR curve was recorded in the form of the measurement signal 120 or in the form of raw data, this was assessed as to whether it is an amplification or not an amplification. The first reference signal 131 represents the amplification. The second reference signal 132 represents no amplification or an absence of amplification. The first reference signal 131 can be described as a sigmoid function having five parameters, as nonlinear, and having S shape. The second reference signal 132 can be described as a linear function having two parameters. The measurement signal 120 is typically based on n measurement points, wherein n corresponds to a number of PCR cycles of the analysis. Because of the system, these measurement points include a noise η. The decision problem then exists as to whether the measurement signal 120 having the noise η corresponds or has similarity more to the first reference signal 131 or to the second reference signal 132.

[0051] FIG. 3A, FIG. 3B, and FIG. 3C show schematic signal diagrams according to one exemplary embodiment. The schematic signal diagrams are shown in the form of intensity-time diagrams or intensity-cycle diagrams. The time t or a cycle number is plotted in each case here on the abscissa axes. An intensity In or signal intensity In is plotted in each case on the ordinate axes. FIG. 3A shows the measurement signal 120 from FIG. 2B, the first reference signal 131 from FIG. 2A, and the second reference signal 132 from FIG. 2C in superimposed form. FIG. 3B shows the measurement signal 120 and the first reference signal 131 in superimposed form. FIG. 3C shows the second reference signal 132. In this case, in FIG. 3A, FIG. 3B, and FIG. 3C, the measurement signal 120, the first reference signal 131, and the second reference signal 132 are scaled with respect to the intensity In, for example to an interval between 0 and 1.

[0052] More precisely and in other words, FIG. 3A, FIG. 3B, and FIG. 3C show an approach for solving the decision problem mentioned in conjunction with FIG. 2A, FIG. 2B, and FIG. 2C. Instead of a line fitting of the measurement signal 120 against the first reference signal 131 and the second reference signal 132 and an estimation of the error (goodness of fit or Kolmogorov-Smirnov test), which in the case of a large amount of noise η can be amplified in microfluidic devices and systems by air bubbles, a cluster analysis or a cluster method (k-means clustering) using the k-means algorithm is made use of here to maintain a robustness of the analysis of the biological material. For this purpose, the measurement signal 120 is scaled, i.e., absolute values are projected onto an interval [0, 1], and compared to two generic functions, which correspond to the first reference signal 131 and the second reference signal 132. The comparison here is a so-called k-means clustering using a predefined distance measure, for example, the Euler distance, Manhattan distance, or the like, having a total of two clusters, k=2. The generic functions (also called dummy functions) each form a cluster category or a cluster and the measurement signal 120 falls due to the respective smaller distance into one of these clusters. It is possible that the measurement signal 120 forms a separate cluster and the clusters based on the reference signals 131 and 132 coincide. Based on the system and theory, an invalid measurement would exist here, since the measurement signal 120 has a shape not to be expected. This case can be used as a quality control.

[0053] FIG. 4A, FIG. 4B, FIG. 4C, and FIG. 4D show schematic signal diagrams according to one exemplary embodiment. The schematic signal diagrams in FIG. 4A and FIG. 4C are shown in the form of intensity-time diagrams or intensity-cycle diagrams. In this case, the time t or a cycle number is plotted in each case on the abscissa axes. An intensity I or signal intensity I is plotted in each case on the ordinate axes. The schematic signal diagrams in FIG. 4B and FIG. 4D are shown in the form of intensity distribution-intensity diagrams. In this case, the intensity I or signal intensity I is plotted in each case on the abscissa axes. An intensity distribution P(I) is plotted in each case on the ordinate axes. FIG. 4A shows the first reference signal 131 from one of the above-mentioned figures. FIG. 4B shows a first reference distribution 471, which is obtained by transforming the first reference signal 131 into a distribution form. FIG. 4C shows the second reference signal 132 from one of the above-mentioned figures. FIG. 4D shows a second reference distribution 472, which is obtained by transforming the second reference signal 132 into a distribution form.

[0054] In other words, FIG. 4A, FIG. 4B, FIG. 4C, and FIG. 4D show a first step for the robust cluster decision. Instead of assessing the qPCR or the measurement signal originating therefrom in the form of a cycle-intensity function, a histogram of intensity values is created. For this purpose, a kernel density estimation is performed over the measurement signal or the PCR curve. Since a qPCR curve, amplifying or non-amplifying, is monotonously rising (with noise in the trend monotonously rising), low intensities correspond to early cycles and high intensities to late values. If the first reference signal 131 is now transformed, a bimodal distribution results as the first reference distribution 471. This is explainable by the two plateaus of the sigmoid curve of the first reference signal 131. In contrast, the linear function of the second reference signal 132 has a unimodal distribution of the intensities as the second reference distribution 472. Due to such a transformation, the decision problem is reduced to establishing a unimodal or bimodal distribution in a corresponding measurement signal transformed into the distribution form.

[0055] FIG. 5A and FIG. 5B show schematic signal diagrams according to an exemplary embodiment. The schematic signal diagram in FIG. 5A is shown in the form of an intensity-time diagram or intensity-cycle diagram. In this case, the time t or a cycle number is plotted on the abscissa axis. An intensity I or signal intensity I is plotted on the ordinate axis. FIG. 5A shows the measurement signal 120, the first reference signal 131, and the second reference signal 132 from one of the above-mentioned figures in superimposed form. The schematic signal diagram in FIG. 5B is shown in the form of an intensity distribution-intensity diagram. In this case, the intensity I or signal intensity I is plotted on the abscissa axis. An intensity distribution P(I) is plotted on the ordinate axis. FIG. 5B shows the measurement signal distribution 175 from FIG. 1, the first reference distribution 471 from FIG. 4B, and the second reference distribution 472 from FIG. 4D, which are each obtained by transformation into a distribution form.

[0056] In other words, the measurement signal 120, the first reference signal 131, and the second reference signal 132 are transformed into intensity distributions. These distributions are now subjected to a cluster analysis with k=2. The bimodal and the unimodal distribution of the generic functions or reference distributions 471 and 472 robustly form two clusters, wherein the measurement signal distribution 175 or raw data distribution falls into one of the two clusters or categories, as a function of its modality.

[0057] FIG. 6A and FIG. 6B show schematic signal diagrams according to one exemplary embodiment. The schematic signal diagram in FIG. 6A is shown in the form of an intensity-time diagram or intensity-cycle diagram. In this case, the time t or a cycle number is plotted on the abscissa axis. An intensity I or signal intensity I is plotted on the ordinate axis. FIG. 6A shows the measurement signal 120 from one of the above-mentioned figures. The schematic signal diagram in FIG. 6B is shown in the form of a noise-time diagram. In this case, the time t or a cycle number is plotted on the abscissa axis. A noise η is plotted on the ordinate axis. FIG. 6B shows the noise data 155 from FIG. 1 or the ascertained noise of the measurement signal.

[0058] In other words, FIG. 6A and FIG. 6B show a further step for obtaining robust decisions. As shown in FIG. 4A, FIG. 4B, FIG. 4C, and FIG. 4D and also FIG. 5A and FIG. 5B, a measurement signal 120 or a qPCR curve in the distribution form or intensity distribution can be characterized with little effort as amplifying or non-amplifying. However, if the noise η is high, the distribution of the intensities is thus accordingly stretched and compressed. The moments of the distribution curves, for example expected values/peaks, width of the peaks/variances, and symmetry of the curve/skewness deviate from ideal curves and become less prominent, i.e., more demanding to characterize. The existing noise η is, however, a measurable variable, which may be ascertained from the raw data of the measurement signal 120. Since air bubbles in microfluidic devices or systems provide a large interference contribution, the noise η is not constant as a function of the cycle number or time t, but increases with greater cycle number or progressing time t. This is to be attributed to the number and size of air bubbles becoming greater over the reaction time of the analysis. Using a sliding window approach, the noise η is measured as a function of the cycle number or the time t using a suitable noise measure, for example, local standard deviation, local signal-to-noise ratio, or the like. The noise data 155 are obtained therefrom, which show the functional relationship of noise η and cycle number or time t for the present analysis.

[0059] FIG. 7A, FIG. 7B, and FIG. 7C show schematic signal diagrams according to one exemplary embodiment. The schematic signal diagrams in FIG. 7A and FIG. 7B are shown in the form of intensity-time diagrams or intensity-cycle diagrams. In this case, the time t or a cycle number is plotted in each case on the abscissa axes. An intensity I or signal intensity I is plotted in each case on the ordinate axes. FIG. 7A shows the first reference signal 131 and the second reference signal 132 from one of the above-described figures. FIG. 7B shows the adapted first reference signal 161 and the adapted second reference signal 162 from FIG. 1. The schematic signal diagram in FIG. 7C is shown in the form of an intensity distribution-intensity diagram. The intensity I or signal intensity I is plotted here on the abscissa axis. An intensity distribution P(I) is plotted on the ordinate axis. FIG. 7C shows the first reference distribution 171 and the second reference distribution 172 from FIG. 1, which are obtained by transforming the respective adapted reference signals 161 and 162 into a distribution form.

[0060] In other words, FIG. 7A, FIG. 7B, and FIG. 7C show how the noise graph or the noise data 155 from FIG. 6B are used to update the reference signals 131 and 132 using the same noise conditions to which the measurement signal is subjected. The reference signals 131 and 132 represent model data having ideal values, wherein a random number or pseudorandom number is added to each ideal value for each cycle of the analysis. This random number or pseudorandom number is generated from the noise interval of the signal measured in FIG. 6A or from the noise data from FIG. 6B. Two adapted reference signals 161 and 162 thus result, which have the same noise as the measurement signal. If the adapted reference signals 161 and 162 are transformed into the distribution form or intensity distribution, the ideal distributions are deformed in accordance with the measured noise, in particular compressed. The measurement signal and the matching reference distribution 171 or 172 are thus closer for the cluster analysis and the cluster decision is more robust.

[0061] FIG. 8 shows a flow chart of an evaluation process 800 according to one exemplary embodiment. The evaluation process 800 is executable in conjunction with the device from FIG. 1 or a similar device. A first block 810 of the evaluation process 800 represents the measurement signal. A following second block 820 represents a scaling of the measurement signal and a measurement of a cycle-dependent noise η(t). A following third block 830 represents an update of the first reference signal and the second reference signal using the noise η(t). An in turn following fourth block 840 represents a transformation of the measurement signal and also the updated reference signals into a distribution form. A following fifth block 850 represents a cluster analysis using the transformed measurement signal and the transformed updated reference signal by means of a k-means algorithm with k=2. A final sixth block 860 represents a decision according to the cluster analysis as to whether or not the measurement signal indicates an amplification.

[0062] The evaluation process 800 can also be referred to as a robust decision process. The evaluation process 800 is suitable in particular for evaluating qPCR having high noise. A qPCR curve in the form of the measurement signal is used as the input data of the evaluation process 800. The measurement signal is scaled in a first step and the cycle-dependent noise is measured as shown in the second block 820. Next, two reference signals or generic curves are generated or updated using the measured noise, as shown in the third block 830. The first reference signal has a qPCR-typical sigmoid curve and the second reference signal has a linear curve. The measurement signal, the first reference signal, and the second reference signal, which were processed in the above-described way, are then converted into an intensity distribution, as shown in the fourth block 840. Subsequently, the converted signals are classified by means of cluster analysis into two clusters, as shown in the fifth block 850. The measurement signal is then further treated like the reference signal, with which the measurement signal was associated during the cluster analysis. The evaluation process 800 can be or become integrated directly into evaluation software of a microfluidic device, such as the device from FIG. 1.

[0063] FIG. 9 shows a flow chart of a method 900 or analysis method 900 according to one exemplary embodiment. The method 900 is executable to perform an analysis of biological material. The method 900 is executable here in conjunction with the device from FIG. 1 or a similar device.

[0064] In a step 910 of reading, in the method 900 for analysis, a measurement signal, a first reference signal, and a second reference signal are read. The measurement signal represents acquired optofluidic data of the biological material. The first reference signal represents first optofluidic model data, which correspond to a first property of the biological material. The second reference signal represents second optofluidic model data, which correspond to a second property of the biological material.

[0065] Subsequently, in the method 900 for analysis, in a step 920 of ascertaining, a noise of the read measurement signal is ascertained to generate noise data. In turn following, in a step 930 of applying, the noise data is applied to the first reference signal and to the second reference signal to generate an adapted first reference signal and an adapted second reference signal. Subsequently, in a step 940 of transforming, the measurement signal, the adapted first reference signal, and the adapted second reference signal are transformed into a frequency distribution form to generate a measurement signal distribution, a first reference distribution, and a second reference distribution.

[0066] Subsequently, in the method 900 for analysis, in a step 950 of performing, a cluster analysis is performed using or on the measurement signal distribution, the first reference distribution, and the second reference distribution, to establish as a function of a result of the cluster analysis whether the biological material has the first property or the second property. Although it is not shown in the illustration of FIG. 9, the method 900 for analysis can also include a step of outputting or providing an analysis signal which represents a result of the analysis.

[0067] According to one exemplary embodiment, the method 900 for analysis includes a step 905 of acquiring the optofluidic data of the biological material to provide the measurement signal. The step 905 of acquiring is executable here before the step 910 of reading.

[0068] Optionally, the method 900 for analysis includes a step 915 of scaling the read measurement signal by projecting absolute values on a predefined value interval. The step 915 of scaling is executable here before the step 920 of ascertaining. In the step 920 of ascertaining, the noise of the scaled measurement signal is ascertained here.

[0069] If an exemplary embodiment includes an “and/or” linkage between a first feature and a second feature, this is to be read to mean that the exemplary embodiment includes both the first feature and also the second feature according to one embodiment and includes either only the first feature or only the second feature according to a further embodiment.