JP6948722B2 - Detection device and detection program - Google Patents

Description

The present invention is based on the measurement data of a plurality of factors items obtained by measurements on biological, detecting device for detecting the state related diseases, related detection program for realizing及beauty the detection device.

According to various research results, the process of exacerbation of many diseases, especially complex diseases, reaches a so-called turning point when a certain critical threshold is exceeded, similar to systems such as climate systems, ecosystems, and economic systems. Then, a state transition suddenly occurs, and the state changes rapidly from a healthy state to a diseased state (see, for example, Non-Patent Documents 1 to 5). In the study of the dynamic mechanism of such complex diseases, the progression process of disease exacerbation (for example, asthma attack, cancer onset) is modeled as a time-dependent nonlinear dynamic system, and the modeled system is observed. It has already been found that the disease is rapidly exacerbated by the phase transition at the bifurcation point (see Non-Patent Documents 1 and 6).

FIG. 1 is an explanatory diagram conceptually showing the disease progression process. FIG. 1a schematically shows the disease progression process. In FIGS. 1, b, c, and d, the stability of the modeled system described above is shown as a potential function in the process of the progress process, the time indicating the progress is taken on the horizontal axis, and the value of the potential function is taken on the vertical axis. It is a graph conceptually shown. As shown in FIG. 1a, the process of disease exacerbation can be represented as a normal state (health state), a pre-disease state, and a disease state. Under normal conditions, the system is stable and the value of the potential function is at its minimum, as shown by the black circles in FIG. 1b. In the pre-disease state, the system has high potential function values, as shown by the black circles in c in FIG. Therefore, it is in a state of being easily affected by a disturbance, and is located near a branch point where a phase transition occurs only by receiving a small disturbance, that is, at the limit of a normal state. However, the pre-disease state can be easily restored to the normal state by appropriate treatment. On the other hand, in the diseased state, the system is stable and the value of the potential function is the global minimum value as shown by the position of the black circle in d in FIG. Therefore, it is difficult to recover from this disease state caused by phase transition from the normal state to the normal state. (In the figure, it is described as unrecoverable for convenience)

Therefore, if the pre-disease state can be detected and the patient can be informed that the pre-disease state is transitioning before the transition to the disease state, appropriate measures can be taken to bring the patient out of the pre-disease state. There is a high possibility that it can be restored to the normal state.

That is, if the branch point (critical threshold value) can be detected, the critical transition can be predicted and early diagnosis of the disease can be realized.

In addition, a biomarker has been conventionally used as a method for diagnosing a disease state. Conventionally used ordinary biomarkers are body fluids such as serum and urine collected from living organisms, and molecular-level DNA, RNA, proteins, metabolites and the like contained in tissues, which are living organisms in living organisms. It is an index that can quantitatively grasp the scientific changes. The conventional method for diagnosing a disease using a biomarker is to compare a biomarker extracted from a normal sample (a sample collected in a healthy state) with a biomarker extracted from an abnormal sample (a sample collected in a diseased state). Is a method of diagnosing.

Venegas, JG et al., "Self-organized patchiness in asthma as a prelude to catastrophic shifts.", (UK), Nature. , Nature Publishing Group, 2005, 434, p.777-782 McSharry, PE, Smith, LA, Tarassenko, L, Prediction of Tantrum: Is the Nonlinear Method Appropriate? epileptic seizures: are nonlinear methods relevant. ”, (UK), Nature Medicine, Nature Publishing Group, 2003, Vol. 9, p. 241-242 Roberto, PB, Eliseo, G., Joseph, C., "Transition models for change" -point estimation in logistic regression.) ”, (USA), Statistics in Medicine, Wiley-Blackwell, 2003, Vol. 22, p. 1141-1162 "Hearing preservation after gamma knife stereotactic radiosurgery of vestibular schwannoma.", (USA), Cancer, Wiley-Black, et al., Paek, S., et al. Well (Wiley-Blackwell), 2005, Volume 1040, p. 580-590 Liu, JK, Rovit, RL, Couldwell, WT, Pituitary Apoplexy, USA, Seminars Seminars in neurosurgery, Thieme 2001, Vol. 12, p. 315-320 Tanaka, G., Tsumoto, K., Tsuji, S., Aihara, K., Intermittent Hormone Therapy for Prostate Cancer (Bifurcation analysis on a hybrid systems model of intermittent hormonal therapy for prostate cancer.) ”, (USA), Physical Review, American Physics Society, 2008, 237, p. .. 2616-2627

However, in the case of complex diseases, it is extremely difficult to predict the critical transition. The reason is as follows.

First, the pre-disease state is the boundary of the normal state and is unlikely to undergo significant changes before reaching a turning point. Therefore, it is difficult to distinguish between the normal state and the pre-disease state by diagnosis by a conventional method such as biomarker or snapshot measurement.

Second, while various studies have been conducted, a reliable disease model capable of accurately detecting a warning signal for early diagnosis for predicting a turning point has not yet been developed. In particular, model-based diagnostic methods have a low probability of success because the process of progression of disease exacerbation varies from individual to individual, even for the same disease.

Third, the pre-disease condition is detected in patients, and the number of samples that can be obtained from a single patient is usually limited, making it difficult to obtain sufficient samples for prediction over a long period of time. Is.

Moreover, the conventional method of diagnosing a disease using a biomarker is a method of making a diagnosis by comparing a normal state with a disease state, and at the time of diagnosis, the patient has already fallen into a disease state and has returned to the previous normal state. It's hard to go back.

The present invention has been made in view of such circumstances, and an object of the present invention is to provide a detection device capable of detecting a pre-disease state, which is a process of transition from a normal state to a disease state.

Another object of the present invention is to provide a detection program for realizing the detection device.

In order to solve the above problems, the detection device described in the present application is a detection device that detects a pre-disease state, which is a process in which a living body to be measured changes from a normal state to a disease state, and has a different time from the living body. A storage means for storing measurement data of a plurality of factor items collected at points, a screening means for detecting a differential factor item from the measurement data of the factor items, and the differential detection means detected by the screening means. From the classification means for classifying the measurement data of the factor item into a plurality of clusters and the plurality of clusters obtained by the classification means, the correlation between the measurement data of the factor item in the cluster is significantly increased, and the relevant A cluster in which the standard deviation of the factor item measurement data in the cluster is significantly increased and the correlation between the factor item measurement data in the cluster and the factor item measurement data in other clusters is significantly reduced. It is characterized by comprising a selection means for selection and a display means for displaying that the living body is in the pre-disease state by characters, figures or images when the selection means selects the cluster. ..

Further, in the detection device, the selection means is characterized in that the significance is tested based on the comparison result between the measurement data of the factor item and the preset reference data.

Further, in the detection device, the selection means is characterized in that a cluster in which the index I calculated by the following formula 1 exceeds a predetermined threshold value is selected.
I = SDd × PCCd / OPCCd ・ ・ ・ Equation 1
However,
PCCd: Average value of absolute values of Pearson correlation coefficients between the measurement data of factor items in the cluster OPCCd: Pearson phase relationship between the measurement data of the factor items in the cluster and the measurement data of the factor items in other clusters. Average value of absolute numbers SDd: Average value of standard deviation of measurement data of factor items in the cluster

Further, in the detection device, the factor item is any one of a measurement item related to genes, a measurement item related to proteins, a measurement item related to metabolites, and a measurement item related to images, voices and sounds obtained from a living body. It is a feature.

Further, the detection program described in the present application is a disease in which the living body is in the process of transitioning from a normal state to a diseased state by a computer that processes measurement data of a plurality of factor items collected at different time points from the living body to be measured. A detection program that detects a pre-state, a screening step in which a computer detects a differential factor item from measurement data of a plurality of factor items collected from the living body at different time points, and the screening step. The correlation between the classification step of classifying the measurement data of the differential factor item detected in the above into a plurality of clusters and the measurement data of the factor item in the cluster from the plurality of clusters obtained in the classification step is Significantly increased, and the standard deviation of the measured data of the factor items in the cluster is significantly increased, and the correlation between the measured data of the factor items in the cluster and the measured data of the factor items in other clusters. A selection step for selecting a cluster in which the amount of data is significantly reduced, and a display step for displaying that the living body is in the pre-disease state by characters, figures, or images when the selection step selects the cluster. It is characterized by being executed.

A detection device or the like having the above characteristics can detect a pre-disease state, which is a process in which a living body to be measured changes from a normal state to a disease state.

Detection instrumentation置及beauty detection program according to the present invention, based on the measurement data collected from a living body, the living body to detect a disease state before a process of transition from the normal state to the disease state. As a result, the present invention exerts excellent effects such as being able to take various measures before transitioning to a diseased state.

FIG. 1 is a schematic diagram illustrating a disease progression process. FIG. 2 is a schematic diagram illustrating the dynamic characteristics of the DNB according to the detection method described in the present application. FIG. 3 is a flowchart showing an example of a DNB detection method according to the embodiment. FIG. 4 is a flowchart showing an example of the selection process of the differential biomolecule in the embodiment. FIG. 5 is a flowchart showing an example of the selection process of DNB candidates in the embodiment. FIG. 6 is a flowchart showing an example of the DNB determination process in the embodiment. FIG. 7 is a diagram showing an example of a diagnosis schedule by the method for early diagnosis of a disease by DNB in the embodiment. FIG. 8 is a flowchart showing an example of an early diagnosis method of a disease by DNB in the embodiment. FIG. 9 is an example of a figure displaying the disease risk proportional to the comprehensive index I. FIG. 10 is an example of a figure displaying the disease risk proportional to the comprehensive index I. FIG. 11 is a block diagram showing a configuration example of the detection device described in the present application. FIG. 12 is a flowchart showing an example of DNB detection processing by the detection device described in the present application. FIG. 13 is a table showing diagnostic data in the first verification example. FIG. 14A is a graph showing an example of time-series changes in the average value of the standard deviations of the detected DNB candidates in the first verification example. FIG. 14B is a graph showing an example of a time-series change in the average value of the absolute values of the Pearson correlation coefficients among the members of the detected clusters of DNB candidates in the first verification example. FIG. 14C is a graph showing an example of a time-series change in the average value of the absolute values of the Pearson correlation coefficients between the members of the detected cluster of DNB candidates and other genes in the first verification example. .. FIG. 14D is a graph showing an example of time-series changes in the average value of the comprehensive indexes of the detected DNB candidates in the first verification example. FIG. 15 is a map showing an example of the dynamic characteristics of DNB in the network composed of the genes of the case group over time in the first verification example. FIG. 16 is a table of diagnostic data in the second verification example. FIG. 17A is a graph showing an example of time-series changes in the average value of the standard deviations of the detected DNB candidates in the second verification example. FIG. 17B is a graph showing an example of a time-series change in the average value of the absolute values of the Pearson correlation coefficients among the members of the detected clusters of DNB candidates in the second verification example. FIG. 17C is a graph showing an example of a time-series change in the average value of the absolute values of the Pearson correlation coefficients between the members of the detected cluster of DNB candidates and other genes in the second verification example. .. FIG. 17D is a graph showing an example of time-series changes in the average value of the comprehensive indexes of the detected DNB candidates in the second verification example.

The inventors of the present invention utilize genomic high-throughput technology capable of obtaining thousands of pieces of information, that is, high-dimensional data, from one sample, and based on the branching process theory, the time development of complex diseases. We constructed a mathematical model and studied the mechanism of disease progression at the molecular network level. As a result, we elucidated the existence of a dynamic network biomarker (DNB) that can detect the immediately preceding branching (sudden deterioration) state before the occurrence of critical transition in the pre-disease state. If the dynamic network biomarker is used as a warning signal for a pre-disease state, a disease model is not required, and early diagnosis of complex diseases can be realized with only a few samples. Hereinafter, embodiments of the present application based on the dynamic network biomarker will be described.

<Theoretical basis>
First, the rationale that is the basis of the embodiment of the present application will be described. Usually, the system related to the disease progression process (hereinafter referred to as system (1)) can be expressed by the following equation (1).

Z (k + 1) = f (Z (k); P) ... Equation (1)

Here, Z (k) = (z1 (k), ..., Zn (k)) is the dynamic state of the system (1) observed at time k (k = 0, 1, ...). It is a variable representing the above, and can be used as information such as gene expression level, protein expression level, and metabolite expression level. In detail, it is information such as the concentration and the number of molecules related to genes, proteins and the like. P is a slowly changing parameter that drives the state transition of the system (1), and can be used as information on genetic factors such as SBP and CNV, and non-genetic factors such as methylation and acetylation. can. f = (f1, ..., fn) is a non-linear function of Z (k).

The normal state and the disease state can be represented by the attractors of the equation of state Z (k + 1) = f (Z (k); P), respectively. The function f is a non-linear function with thousands of variables, as the process of progression of complex diseases has very complex dynamic properties. Moreover, the element P that drives the system (1) is difficult to identify. Therefore, it is very difficult to construct and analyze a system model of normal state and disease state.

On the other hand, the inventors of the present invention focused on the critical transition state of the system immediately before the transition from the normal state to the disease state, that is, the pre-disease state. Generally, the system (1) has an equilibrium point having the following characteristics.

1. 1. If Z ^* is a fixed point of the system (1), then Z ^* = f (Z ^* ; P),
2. If Pc is the threshold value at which the system branches, then when P = Pc, ^{the absolute value of one real eigenvalue of the Jacobian matrix ∂f (Z; Pc) / ∂Z | Z = Z *} or the eigenvalue of a pair of complex conjugates. Becomes 1,
3. 3. When P ≠ Pc, the absolute value of the eigenvalues of the system (1) is generally not 1.

From the above characteristics, the inventors have theoretically clarified that the following peculiar characteristics appear when the system (1) is in the critical transition state. That is, when the system (1) is in the critical transition state, the control group (sub) consisting of some nodes in the network (1) in which each of the variables z1, ..., Zn of the system (1) is configured as a node. Network) appears. Ideally, the dominant group that appears in the critical transition state has the following strong characteristics.

(I) When zi and zj are nodes belonging to the controlling group,
PCC (zi, zj) → ± 1;
SD (zi) → ∞;
SD (zj) → ∞.
(II) When zi is a node belonging to the controlling group, but zj is not a node belonging to the controlling group.
PCC (zi, zj) → 0;
SD (zi) → ∞;
SD (zj) → Boundary value.
(III) When zi and zj are not nodes belonging to the controlling group PCC (zi, zj) → α, α ∈ (-1,1) \ {0};
SD (zi) → Boundary value;
SD (zj) → Boundary value.

Here, PCC (zi, zj) is the Pearson correlation coefficient between zi and zj, and SD (zi) and SD (zj) are the standard deviations between zi and zj.

That is, in the network (1), the appearance of the dominant group having the above-mentioned peculiar characteristics (I) to (III) is regarded as a sign that the system (1) is transitioning to the critical transition state (pre-disease state). be able to. Therefore, by detecting the control group, the critical transition of the system (1) can be detected. That is, the control group can be a critical transition, that is, a warning signal indicating a pre-disease state immediately before the exacerbation of the disease. Then, no matter how complicated the system (1) is, even if the driving element is unknown, if only the control group that becomes the warning signal is detected, the mathematical model of the system (1) is not directly handled. It is possible to identify the pre-illness state. By identifying the pre-illness state, it becomes possible to realize proactive measures and early treatment for the disease.

In the embodiments described in the present application, the control group that serves as a warning signal indicating a pre-disease state is referred to as a dynamic network biomarker (hereinafter, abbreviated as DNB).

<DNB characteristics and judgment conditions>
As described above, the DNB is a dominant group having unique characteristics (I) to (III), and is a subnetwork consisting of a plurality of nodes when the system (1) is in a pre-disease state. It appears in 1). In the network (1), if each node (z1, ..., Zn) is a factor item to be measured for a biomolecule such as a gene, a protein, or a metabolite, DNB has the above-mentioned peculiar characteristic (I). )-(III) is a group (sub-network) consisting of factor items related to some biomolecules.

Furthermore, the conditions for determining DNB can be determined as follows based on the above-mentioned peculiar characteristics (I) to (III).

-Condition (I): A group consisting of some biomolecules (genes, proteins or metabolites) existing in the network (1), and the absolute value of the Pearson correlation coefficient PCC between the biomolecules in the group. The average value of is significantly increased.
-Condition (II): The average value of the absolute values of the Pearson correlation coefficient OPCC between biomolecules in the group and other biomolecules is significantly reduced.
-Condition (III): The average value of the standard deviation SD of biomolecules in the group is significantly increased.

A group consisting of biomolecules that simultaneously satisfy the above conditions (I) to (III) is recognized as DNB.

In order to intuitively explain the characteristics of DNB, next, the dynamic characteristics of DNB in the network will be described by taking a network having six nodes as an example. FIG. 2 is a schematic diagram illustrating the dynamic characteristics of the DNB according to the detection method described in the present application. FIG. 2a shows a normal state, a pre-disease state, and a disease state as a disease progression process. B, c and d of FIG. 2 show the stability of the modeled system (1) in the normal state, the pre-disease state and the disease state as a potential function for the process of the progress process, and the progress is shown on the horizontal axis. It is a graph conceptually shown by taking time and taking the value of the potential function on the vertical axis. Further, e, f and g in FIG. 2 conceptually show an example of the network state of the system (1) corresponding to each of the normal state, the pre-disease state and the disease state. Furthermore, h in FIG. 2 shows an example of time-dependent changes in molecular concentration, which is a factor item of DNB in the pre-disease state.

Nodes z1 to z6 represent factor items related to different types of biomolecules such as genes, proteins, and metabolites, and the connection line between nodes z1 to z6 shows the correlation between the nodes, and the thickness of the line is large. Indicates the magnitude of the Pearson correlation coefficient PCC, and the drawing method of the circles surrounding z1 to z6 indicates the magnitude of the standard deviation SD of each node. As a method of drawing the connection line, it is shown that the standard deviation SD is the smallest when the inside of the circle is blank, and the standard deviation SD becomes larger as the diagonal line is in one direction and the diagonal line is in two directions.

In the normal state, as shown in E of FIG. 2, each node has an equal correlation and a standard deviation of each node, and is at a medium level. However, in the pre-disease state, groups (z1 to z3) appear that exhibit significantly unique characteristics as compared to other nodes. As shown in FIG. 2f, the nodes z1 to z3 in the group have a significantly increased Pearson correlation coefficient with each other, and the Pearson correlation coefficient with other nodes z4 to z6 has decreased significantly. There is. Further, the standard deviation of the nodes z1 to z3 in the group is large. The reason is that the concentrations of the nodes z1 to z3 in the group change drastically at different times (t = 1, t = 2, t = 3) as shown in h in FIG.

However, after the transition to the disease state, as shown in g of FIG. 2, the standard deviation of each node z1 to z3 in the group is slightly larger, but the Pearson correlation coefficient between each node is even. It is back to the middle level. That is, the above-mentioned peculiar characteristics of the group (z1 to z3) have disappeared.

As conceptually shown in FIG. 2, in the pre-disease state, a dominant group appears that includes some nodes with good characteristics. In the present application, such a controlling group called DMB shall be used as a biomarker for early diagnosis of a disease as a warning signal capable of indicating that the patient is in a pre-disease state and foretelling a transition to the disease state. Can be done. DNB is also a subnetwork that appears in networks with changing properties, unlike static biomarkers used in normal disease diagnosis. Therefore, in the present application, the controlling group is referred to as DNB (Dynamic Network Biomarker).

<Warning signal>
As mentioned above, DNB can be used for early diagnosis of disease as a warning signal indicating a pre-disease condition. As a measure of the strength of the warning signal, the average value of the absolute values of the Pearson correlation coefficient PCC between the nodes in the DNB and the absolute value of the Pearson correlation coefficient OPCC between the node in the DNB and other nodes are measured. The mean value of, and the standard deviation SD of DNB can be used. Furthermore, a comprehensive index I that can comprehensively reflect the characteristics of DNB can be introduced. In the embodiment of the present application, as an example, the comprehensive index I represented by the following equation (2) is introduced.

I = SDd × PCCd / OPCCd ... Equation (2)

In equation (2), PCCd indicates the average value of the absolute values of the Pearson correlation coefficients between the nodes in the DNB, and OPCCd indicates the average value of the absolute values of the Pearson correlation coefficients between the nodes in the DNB and other nodes. Shown, SDd indicates the average value of the standard deviations of the nodes in the DNB. As can be seen from the equation (2), when SDd and PCCd increase and OPCCd decreases, the total index I increases significantly. Therefore, the characteristics of DNB can be detected with high sensitivity. Furthermore, the distance from the disease state can be grasped to some extent from the value of the comprehensive index I.

<DNB detection method>
FIG. 3 is a flowchart showing an example of a DNB detection method according to the embodiment. In the detection method described in the present application, it is first necessary to obtain measurement data by measurement on a living body. By using a high-throughput technology such as a DNA chip, it is possible to measure more than 20,000 genes from one biological sample. In order to statistically analyze, in the embodiment of the present application, a plurality of (6 or more) biological samples are collected from the measurement target at different time points, and the measurement data obtained by measuring the collected biological samples are aggregated. Get statistical data. As shown in FIG. 3, the methods for detecting DNB in the embodiment of the present application are mainly high-throughput data acquisition processing (s1), differential biomolecule selection processing (s2), and clustering processing (s3). And the DNB candidate selection process (s4) and the DNB determination process (s5) by significance analysis. Next, each of these processes will be described in detail.

In the high-throughput data acquisition process of step s1, the sample to be detected is used as a case sample, the reference sample is used as a control sample, and each high-throughput physiological data, that is, the expression level of the biomolecule is measured from each sample. This is a process for acquiring data (for example, microarray data). The reference sample is a sample collected in advance from the patient to be inspected, a sample collected first at the time of collection, or the like, and is used as a control sample having a purpose such as calibration of a measuring device. Control samples are not always necessary, but are useful for eliminating error factors and improving measurement reliability.

The process of selecting the differential biomolecule in step s2 is a process of selecting a biomolecule showing a significant change in the expression level. FIG. 4 is a flowchart showing an example of the selection process of the differential biomolecule in the embodiment. FIG. 4 shows in detail the biological treatment of the differential biomolecule in step s2 shown in FIG.

As shown in FIG. 4, first, the statistical data based on the high-throughput data (expression level of biomolecules) measured from each of the n case samples is designated as D1c, and the data measured from the control sample is designated as Dr (s21). Next, Student's t-test is performed on each case sample biomolecule D1c, and the biomolecule D2c showing a significant change in the expression level as compared with the high-throughput data Dr of the control sample is selected (s22). In step s22, Student's t-test is exemplified as a method for selecting the biomolecule D2c showing a remarkable change in the expression level, but the method is not particularly limited. For example, it is possible to apply other test methods such as the Mann-Whitney U test, and such a nonparametric test is particularly effective when the population D1c does not follow a normal distribution. be. Further, even in the case of Student's t-test, the value of the significance level α can be appropriately set to a value such as 0.05 or 0.01.

Next, the misexpression rate FDR is used to correct multiple comparisons or t-tests of multiple students for each case sample biomolecule D2c, and each corrected case sample gene or protein data D3c. Is elected (s23). Next, using the two-fold change method, Dc in which the standard deviation SD changes relatively significantly from each corrected case sample gene or protein data D3c is selected as a differential biomolecule. .. The differential biomolecule Dc selected here not only shows a remarkable difference as compared with the biomolecule Dr of the control sample, but also deviates greatly from its own average value. Also in step s23, the test method is not limited to Student's t-test.

Next, a clustering process (s3 in FIG. 3) is performed. The clustering process referred to here is a process of classifying a plurality of biomolecules into groups having a high correlation with each other, and each group in which the biomolecules are classified is referred to as a cluster. That is, the differential biomolecules Dc selected in step s24 shown in FIG. 4 are classified into n clusters so that the biomolecules having a high correlation with each other are formed into one cluster. All resulting clusters are potential control groups, ie, candidates for DNB to be detected.

Next, the selection process (s4) of the DNB candidates shown in FIG. 3 is performed. FIG. 5 is a flowchart showing an example of the selection process of DNB candidates in the embodiment. FIG. 5 shows in detail the selection process of the DNB candidate in step s4 shown in FIG. That is, the DNB candidate selection process is performed based on the flowchart of the DNB candidate selection process shown in FIG. In the circular loop shown in FIG. 5, for all clusters (k) (k = 1, ..., n), the average value of the absolute values of the Pearson correlation coefficients between the nodes in the cluster, PCCd (k), and the cluster. Calculate the mean OPCCd (k) of the absolute values of the Pearson correlation coefficients between the nodes in the cluster, the mean SDd (k) of the standard deviations of the nodes in the cluster, and the overall index I (k). (S41 to s46). Then, from all the clusters, the cluster having the largest total index I value is selected as a candidate for DNB (s47).

Next, the DNB determination process (s5 in FIG. 3) is performed by the significance analysis shown in FIG. FIG. 6 is a flowchart showing an example of the DNB determination process in the embodiment. FIG. 6 shows in detail the DNB determination process by the superiority analysis of step s5 shown in FIG. That is, based on the above-mentioned DNB determination conditions (I) to (III), it is determined whether or not the cluster (m) selected as a DNB candidate in step s47 is a DNB. It can be determined by various significance analyzes, but as an example, the process is performed based on the flowchart of the DNB determination process shown in FIG.

As shown in FIG. 6, first, the average value PCCdr of the absolute value of the Pearson correlation coefficient between each node and the average value SDdr of the standard deviation of each node of the data acquired from the control sample are calculated (s51, s52, respectively). ). Then, the average value PCCd (m) of the absolute value of the Pearson correlation coefficient between each node in the cluster (m) selected in step s47 is significantly increased as compared with the average value PCCdr of the Pearson correlation coefficient of the control sample. It is determined whether or not it has been done (s53). When it is determined that the increase is not significant (NO), the result that DNB does not exist is output, and the process is terminated (s57). On the other hand, if it is determined that the increase is significantly (YES), the process proceeds to the next step s54. In step s54, was the average value OPCCd (m) of the Pearson correlation coefficient between each node in the cluster (m) and other nodes significantly reduced as compared with the average value PCCdr of the Pearson correlation coefficient of the control sample? It is determined whether or not (s54). If it is determined that the reduction is not significant (NO), the result that DNB does not exist is output (s57), and the process is terminated. On the other hand, if it is determined that the reduction is significantly reduced (YES), the process proceeds to the next step s55. In step s55, it is determined whether or not the standard deviation average value SDd (m) of the nodes in the cluster (m) is significantly increased as compared with the standard deviation average value SDr of the control sample. If it is determined that there is no significant increase (NO), it is determined that DNB does not exist (s57), and the process is terminated. On the other hand, if it is determined that the cluster (m) has increased significantly, the cluster (m) is recognized as DNB (s56), and the process is terminated.

<Early diagnosis method of diseases by DNB>
As a diagnosis schedule, it is desirable to perform multiple diagnoses at regular intervals and take several samples for each diagnosis. FIG. 7 is a diagram showing an example of a diagnosis schedule for early diagnosis of a disease using DNB in the embodiment. As shown in FIG. 7, samples are collected in a plurality of stages (stage 1, stage 2, ..., Stage T). As for the number of samples to be collected in each stage, it is usually desirable to collect 6 or more samples in one stage in order to ensure the accuracy of the data. The interval between two consecutive stages can be as long as days, weeks, months, or years, depending on the disease situation, but each stage takes samples at different time points over a short period of time. It is desirable to do. For example, 6 samples are taken at 6 time points of the day. The interval between each time point can be, for example, several minutes or several hours, depending on the situation.

FIG. 8 is a flowchart showing an example of an early diagnosis method of a disease by DNB in the embodiment. As shown in FIG. 8, the methods for early diagnosis of diseases by DNB mainly include sampling processing (s100), selection of differential biomolecules (s200), selection processing of DNB candidates (s300), and processing. It includes a DNB determination process (s400) by significance analysis and a diagnosis result output process (s500). Next, these processing contents will be described in detail.

Sample collection process (s100): Similar to a general disease diagnosis method, a sample for collecting necessary physiological data is collected according to the disease to be diagnosed. For example, in the case of liver damage, samples such as blood and liver tissue are taken.

In addition to using the sample collected from the diagnosis target as a case sample when diagnosing, as a reference sample, a sample collected from a healthy person other than the diagnosis target or a sample collected first from the diagnosis target is used as a control sample. Can be.

Selection of differential biomolecule (s200): The differential biomolecule is selected from the sample collected in step s100 according to the flow chart for selecting the differential biomolecule shown in FIG.

DNB candidate selection process (s300): According to the DNB candidate selection flowchart shown in FIG. 5, a dominant group that is a DNB candidate is selected from the differential biomolecules selected in step s200.

DNB determination process by significance analysis (s400): It is determined whether or not the DNB candidate selected in step s300 is DNB according to the flowchart showing the method of determining DNB by significance analysis shown in FIG. ..

Diagnosis result output (s500): When it is determined that the DNB does not exist in the step s400, the data of the DNB candidate selected in the step s300 is saved in the storage device as the reference data for the next diagnosis, and an abnormality is recognized. Output the diagnosis result that there is no such thing. On the other hand, when it is determined in step s400 that there is a cluster certified as DNB, the biomolecular data of the certified cluster is stored as a member of DNB, and the diagnosis result indicating that the patient is in a pre-disease state is output. It is also possible to output the diagnosis result associated with the detected DNB. Although it is shown here as a diagnosis result, in detail, for example, it is a result that can be used as a reference for a diagnosis by a doctor, and the diagnosis result output in step s500 is not the diagnosis itself by the doctor but the diagnosis by the doctor. This is output data that assists in diagnosis and serves as a reference for diagnosis in order to support diagnosis.

For example, it is possible to output a comprehensive index I that comprehensively reflects the characteristics of DNB. The larger the total index I is, the closer it is to the branch point. Therefore, if the disease risk proportional to the total index I is output in the form of an intuitively visible figure or image, the warning effect becomes higher.

9 and 10 are examples of figures displaying the disease risk proportional to the comprehensive index I. In FIG. 9, the entire arrow indicates the pre-disease state (early onset), the flow in the direction indicated by the arrow indicates the time course of the disease state (onset), and the diamond mark located on the left side of the arrow is for diagnosis. It is a disease risk pointer whose position changes according to the value of the obtained comprehensive index I. The larger the value of the total index I, the closer the diamond mark is to the right side of the arrow.

In addition, if an early diagnosis of a disease has been made using DNB before, the disease risk proportional to the comprehensive index I is displayed together with the comprehensive index obtained in the previous diagnosis, as shown in FIG. be able to. In FIG. 10, the dotted diamond mark shows the comprehensive index obtained by the diagnosis on July 1, 2011, and the solid diamond mark shows the comprehensive index obtained by the diagnosis on September 1, 2011. From the position change of the diamond mark, it is possible to intuitively judge that the disease state is approaching.

Further, as the information associated with the DNB, a map of the entire network including the detected DNB or a part of the network including the DNB (for example, FIG. 15 described later) can be output.

Furthermore, it is also possible to output a list of biomolecules that are members of DNB. As described above, DNB appears in the pre-disease state that transitions from the normal state to the disease state, but the biomolecule detected as DNB, that is, the gene, protein or metabolite itself is not necessarily a factor that exacerbates the disease. It is not necessarily a pathogenic gene, protein or metabolite. However, some DNB members have been found to be associated with the disease.

Therefore, if a substance (gene, protein or metabolite) related to a specific disease contained in the detected DNB members is extracted, for example, a diagnosis target such as a examiner may develop by a doctor's diagnosis. Be able to grasp sexual diseases to some extent.

Therefore, following the above-mentioned "output of diagnosis result" (s500 in FIG. 8), using a database that stores the correspondence between the gene, protein or metabolite and the disease, the detected DNB is associated with the disease. A certain gene, protein or metabolite can be extracted and output as a diagnostic result that can be used as a reference for diagnosis.

Then, for example, when blood is collected from a person undergoing a medical examination and DNB is detected from the data of genes, proteins or metabolites obtained from the blood, the genes, proteins or metabolites contained in the DNB are detected. Diseases related to can be identified to some extent, and potential diseases possessed by the examiner can be diagnosed at an early stage.

<Detector>
The DNB detection method described in detail above can be embodied as a detection device using a computer. FIG. 11 is a block diagram showing a configuration example of the detection device described in the present application. The detection device 1 shown in FIG. 11 is realized by using a personal computer, a client computer connected to a server computer, and various other computers. The detection device 1 includes various mechanisms such as a control unit 10, a recording unit 11, a storage unit 12, an input unit 13, an output unit 14, and an acquisition unit 15.

The control unit 10 is a mechanism that is configured by using a circuit such as a CPU (Central Processing Unit) and controls the entire detection device 1.

The recording unit 11 is a non-volatile auxiliary recording mechanism such as a magnetic recording mechanism such as an HDD (Hard Disk Drive) and a non-volatile semiconductor recording mechanism such as an SSD (Solid State Disk). Various programs and data such as the detection program 11a described in the present application are recorded in the recording unit 11.

The storage unit 12 is a volatile main storage mechanism such as SDRAM (Synchronous Dynamic Random Access Memory) and SDRAM (Static Random Access Memory).

The input unit 13 is an input mechanism including hardware such as a keyboard and a mouse, and software such as a driver.

The output unit 14 is an output mechanism including hardware such as a monitor and a printer, and software such as a driver.

The acquisition unit 15 is a mechanism for acquiring various data from the outside. Specifically, it is various hardware such as a LAN (Local Area Network) port that takes in data via a communication network, a port that connects to a dedicated line such as a parallel cable that can be connected to a measuring device, and software such as a driver. ..

Then, by storing the detection program 11a recorded in the recording unit 11 in the storage unit 12 and executing the detection program 11a under the control of the control unit 10, the computer executes various procedures related to the detection program 11a, which is described in the present application. Functions as the detection device 1. For convenience, the recording unit 11 and the storage unit 12 are separated, but both have the same function of recording various information, and which mechanism should be used for recording according to the specifications of the device, the operation mode, and the like. Can be determined as appropriate.

FIG. 13 is a flowchart showing an example of DNB detection processing by the detection device 1 described in the present application. The process of the detection device 1 described in the present application executes the above-mentioned DNB detection process. The control unit 10 of the detection device 1 acquires measurement data for a plurality of factor items obtained by measurement on a living body by the acquisition unit 15 (Sc1). Step Sc1 corresponds to the high-throughput data acquisition process shown as step s1 in FIG. In addition, since it is expressed as a factor item because it is expressed as a target of computer processing here, the factor item referred to here is a measurement item related to a gene that can be a node of DNB, a measurement item related to a protein, and a metabolite. Indicates measurement items such as measurement items. It is also possible to use measurement items related to an image obtained from an internal image output by a measuring device such as a CT scan.

The control unit 10 tests whether or not each measurement data of the acquired factor items changes significantly over time, and selects a differential biomolecule based on the test result (Sc2). Step Sc2 corresponds to the process of selecting the differential biomolecule shown as step s2 in FIG.

Therefore, in the process of step Sc2, the control unit 10 tests the significance based on the measurement data of each factor item and the comparison result with the factor item and the reference data preset for each time series ( Sc21), including a process (Sc22) of selecting factor items that have been tested as having significant changes over time. That is, various processes shown in FIG. 4 are executed. The data processed by the detection device 1 as reference data is a control sample. For example, based on a setting such that the sample acquired first is used as a control sample, the detection device 1 refers to the reference data for the sample. It is treated as.

The control unit 10 classifies a plurality of factor items into a plurality of clusters based on the correlation of time-series changes in the measurement data of each selected factor item (Sc3). Step Sc3 corresponds to the clustering process shown as step s3 in FIG.

The control unit 10 corresponds to the selection conditions set in advance from each of the classified clusters based on the correlation of the time-series change of the measurement data of each factor item and the time-series change of the measurement data between each factor item. Select a cluster (Sc4). Step Sc4 corresponds to the DNB candidate selection process shown as step s4 in FIG.

Therefore, in the process of step Sc4, the control unit 10 calculates, for each cluster, the average value of the values indicating the correlation of the measurement data of each factor item in the cluster as the first index (Sc41), and in the cluster. The average value of the values indicating the correlation between the measurement data of the factor item and the measurement data of the factor item outside the cluster is calculated as the second index (Sc42), and the standard deviation of the measurement data for each factor item in the cluster is calculated. The process (Sc43) of calculating the average value as the third index is included. Further, in the process of step Sc4, the control unit 10 calculates a total index based on the product of the first index, the second index, and the reciprocal of the third index (Sc44), and the calculated total index is the maximum. The process of selecting a cluster (Sc45) is included. That is, various processes shown in FIG. 5 are executed. The first index, the second index, the third index, and the total index include, for example, the average value of the absolute values of the Pearson correlation coefficients between the nodes in the cluster, PCCd (k), the nodes in the cluster, and other indexes. The mean OPPCd (k) of the absolute values of the Pearson correlation coefficients with the nodes, the mean SDd (k) of the standard deviations of the nodes in the cluster, and the overall index I (k) are used, respectively.

The control unit 10 detects factor items included in the selected cluster as biomarker candidates (Sc5). Step Sc5 corresponds to the DNB determination process shown as step s5 in FIG.

Therefore, in the process of step Sc5, the control unit 10 calculates the reference standard deviation indicating the average value of the standard deviations of the corresponding reference data for each factor item (Sc51), and averages the values indicating the correlation between the factor items. A reference correlation value indicating the value is calculated (Sc52). Then, the first index increases with significance compared to the reference standard deviation, the second index decreases with significance compared to the reference correlation value, and the third index has significance with respect to the reference standard deviation. It includes a process (Sc53) of detecting an item contained in the cluster as a biomarker when it is increased. That is, various processes shown in FIG. 6 are executed. As the reference standard deviation and the reference correlation value, for example, the average value PCCdr of the absolute value of the Pearson correlation coefficient between each node and the average value SDdr of the standard deviation of each node are used.

Then, the control unit 10 outputs the factor item detected as a candidate for the biomarker from the output unit 14 (Sc6), and ends the process.

<First verification example>
In order to verify the diagnostic accuracy of the early diagnosis method of the disease by DNB described in the present application, the diagnosis is performed by the diagnostic method described in the present application using the experimental data of the mouse in which the lung disorder is caused, and the diagnosis result is the actual medical condition. The effectiveness of the diagnostic method described in the present application was verified in comparison with the progress. Next, the verification example will be described in detail. The experimental data shows that multiple experimental CD-1 male mice are divided into a case group and a control group, and the case group is in a normal air environment and the control group is in an air environment containing phosgen, which is a toxic gas. It was obtained from an experiment in which the health status of the mice in both groups was observed and the molecular level mechanism of acute lung injury due to inhalation of phosgen was investigated. Using the experimental data, the health status of mice in the phosgene-exposed case group was diagnosed using the diagnostic method described in the present application. Normally, mice develop phosgene-induced lung injury when they inhale a certain amount of phosgene.

FIG. 13 is a table showing diagnostic data in the first verification example. As shown in FIG. 13, the diagnosis target is a mouse having phosgen-induced lung injury (CD-1 male mouse), and the sample collection target is a mouse in the case group to be diagnosed and the control group to be referred to. The sampling point is the time point at which 0, 0.5, 1, 4, 8, 12, 24, 48, 72 hours have passed since the start of the experiment, and the sampling point of the gene used for DNB detection. The number is 22690.

Specifically, the following processing was performed using the diagnostic method described in the present application.

First, the differentially expressed gene is selected from the high-throughput gene data measured from each sample. Six case samples and six control samples are provided at each sampling point. At the first sampling point (0h), the case sample data is the same as the control sample data.

At each sampling point, using Student's t-test with significance level p <0.05, the differential gene with A = [0, 53, 184, 1325, 1327, 738, 980, 1263, 915], respectively. Was elected.

Then, for each sampling point, B = [0, 29, 72, 195, respectively, by using false expression rate (FDR) and doubling modification screening for the selected differential gene set A. 269, 163, 173, 188, 176] genes were selected.

For the selected gene set B, clustering was performed to combine those with high correlation into one cluster at each sampling point, and 40 clusters were obtained for each.

In addition, at each sampling point, data normalization is performed for all genes in the obtained 40 clusters. Then, at each sampling point, the average value SDd (third index) of the standard deviation of each cluster in the normalized control group and the case group, and the average value of the absolute values of Pearson's correlation coefficients among the members of the cluster. The PCCd (second index), the average OPCCd (first index) of the absolute values of the Pearson correlation coefficients between the members of the cluster and other genes, and the total index I are calculated.

Then, at each sampling point, the cluster with the largest comprehensive index I is selected as a DNB candidate from each cluster in the calculated case group, and the average value of the standard deviations of the control group is selected for the DNB candidate. Whether or not it is DNB is determined by significance analysis based on the average value PCCc of the absolute values of SDc and the Pearson correlation coefficient between genes. As a result, the number of clusters that became DNB at each sampling point was [0, 0, 0, 0, 1, 0, 0, 0, 0].

That is, DNB was detected at the 5th sampling point (8h), and the DNB is the 111th cluster having 220 genes.

FIG. 14A is a graph showing an example of time-series changes in the average value SDd of the standard deviations of the detected DNB candidates in the first verification example. FIG. 14B is a graph showing an example of the time-series change of the average value PCCd of the absolute values of the Pearson correlation coefficients among the members of the detected clusters of DNB candidates in the first verification example. FIG. 14C is a graph showing an example of the time-series change of the average value OPCCd of the absolute values of the Pearson correlation coefficients between the members of the detected cluster of DNB candidates and other genes in the first verification example. be. FIG. 14D is a graph showing an example of time-series changes in the comprehensive index I of the detected DNB candidates in the first verification example. In FIGS. 14A-14D, the horizontal axis represents the time stage t, and the vertical axis represents the average value SDd of the standard deviation (FIG. 14A) and the average value of the absolute values of the Pearson correlation coefficients between the members of the cluster PCCd (FIG. 14A). FIG. 14B) shows the mean OPCCd (FIG. 14C) of the absolute values of the Pearson correlation coefficients between the members of the cluster and other genes, and the overall index I (FIG. 14D). The broken line shows the time course of various indices of DNB candidates detected from the case group, and the solid line shows the time course of various indices of one cluster selected from the control group.

As can be seen from FIGS. 14A to 14D, from the fourth time stage (that is, after 4 hours), the first index PCCd, the third index SDd, and the total index I of the DNB candidates began to increase significantly, and the fifth It peaks at the time stage (ie, 8 hours have passed). On the other hand, the third index OPCCd, which is a candidate for DNB, decreases from the second time stage and shows a minimum value in the same fifth time stage (that is, after 8 hours).

In addition, in order to intuitively show the dynamic characteristics of DNB, the dynamic characteristics of the entire gene network including DNB are shown in FIG. FIG. 15 is a map showing an example of the dynamic characteristics of DNB in the network composed of the genes of the case group over time in the first verification example. In FIG. 15, the gene network of the case group (3452 genes, 9238 links) is displayed at each sampling point of 0.5, 1, 4, 8, 12, 24, 48, and 72h in order. , The node indicated by "○" is a gene belonging to the DNB candidate, and the node indicated by "□" is another gene near the node of the DNB candidate, and is between the nodes. The line shows the correlation between the two nodes. Further, the color density of "○" indicates the magnitude of the standard deviation SD of the gene, and the density of the connection line connecting both nodes indicates the magnitude of the absolute value of the correlation coefficient PCC of both nodes. All maps are constructed using Cytoscape (http://www.cytoscape.org/).

As shown in FIG. 15, the characteristics of DNB candidates (SD, PCC) change with the passage of time, and gradually evolve from a normal cluster that behaves like other genes to DNB. At the fifth stage (when 8h has passed) shown in E of FIG. 15, the characteristic as DNB is the most remarkable, and the warning signal which is the pre-disease state (8h) is clearly shown. However, after transitioning to a diseased state (24h, 48h, 72h), DNB members again behave in the same way as other genes.

From the results, it is found that the pre-disease state exists near the fifth time stage, and after the fifth time stage, the system transitions to the disease state.

Therefore, when diagnosing by the early diagnosis method of the disease by DNB of the embodiment of the present application, since a warning signal of becoming a disease is slightly visible in the diagnosis at the fourth time stage, the diagnosis result that the disease will worsen in the near future is obtained. Can be put out. Then, in the diagnosis at the fifth time stage, since the warning signal for getting sick is clearly visible, it is possible to give a diagnosis result that the person will get sick soon.

On the other hand, according to the results of actual mouse experiments, the mice in the case group developed pulmonary edema 8 hours after inhalation of phosgene, 50% to 60% died 12 hours later, and another 60% died 24 hours later. ~ 70% died.

Therefore, it can be said that the diagnosis result of the early diagnosis by DNB of the embodiment of the present application is completely in agreement with the actual disease exacerbation situation of the mouse.

<Second verification example>
In the first verification example described above, the effectiveness of the method for early diagnosis of disease by DNB according to the embodiment of the present application was verified using the data of animal experiments. In this verification example, clinical data of B-cell lymphoma is used to further verify the diagnostic accuracy of the method for early diagnosis of disease by DNB according to the embodiment of the present application.

FIG. 16 is a table listing the diagnostic data in the second verification example. As shown in FIG. 16, based on clinical features, lesions and flow cytometry, the samples were sampled in resting phase (P1), active phase (P2), marginal phase (P3), metastatic phase (P4), invasion phase (P5). ) Is divided into five stages. At each stage, the number of samples is 5, 3, 6, 5, 7, respectively. At each stage, the splenomegaly is "None", "None", "+/-", "+", "+++", respectively. At each stage, the flow cytometry is "normal rest", "normal activity", "abnormal", "mixed", "B-1 clone", respectively. Further, the sample collected in the resting period (P1) is used as a control sample, and the sample collected in another stage (P2 to P5) is used as a case sample.

From the gene expression data measured from the above 26 samples, 13712 genes were diagnosed by the above-mentioned method for early diagnosis of diseases by DNB. As a result, FIGS. 17A to 17D show each index of DNB candidates detected from the genes of the case group. FIG. 17A is a graph showing an example of time-series changes in the average value SDd of the standard deviations of the detected DNB candidates in the second verification example. FIG. 17B is a graph showing an example of time-series changes in PCCd, which is the average value of the absolute values of the Pearson correlation coefficients among the members of the detected clusters of DNB candidates in the second verification example. FIG. 17C is a graph showing an example of the time-series change of the average value OPCCd of the absolute values of the Pearson correlation coefficients between the members of the detected cluster of DNB candidates and other genes in the second verification example. be. FIG. 17D is a graph showing an example of time-series changes in the comprehensive index I of the detected DNB candidates in the second verification example.

In FIGS. 17A to 17D, the horizontal axis indicates the number of the stage (P1 to P4), and the vertical axis represents the average value SDd of the standard deviation (FIG. 17A) and the Pearson correlation coefficient between the members of the cluster, respectively. The mean PCCd of absolute values (FIG. 17B), the mean OPCCd of absolute values of the Pearson correlation coefficient between cluster members and other genes (FIG. 17C), and the overall index I are shown (FIG. 17D). ..

As is clear from FIGS. 17A to 17D, in the second stage (P2), that is, in the active phase, the overall index I of the DNB candidates has reached its peak, and the warning signal indicating the disease state is the strongest. .. The results of this diagnosis are in perfect agreement with the actual lesion. Even in the actual clinical data, the condition worsens from the marginal stage immediately after the active stage, the splenomegaly becomes "+/-", and the flow cytometry becomes "abnormal". Therefore, the DNB-specific analysis results in this verification example are in perfect agreement with the actual clinical data.

Further, in the normal diagnosis, as shown in FIG. 16, since the splenomegaly in the active phase is "None" and the flow cytometry is "normal activity", the diagnosis result is that no abnormality is found. On the other hand, in the case of diagnosing by the early diagnosis method of the disease by DNB of the embodiment of the present application, since the warning signal (DNB) indicating the pre-disease state was detected in the active phase, the diagnosis result that "a sign of abnormality is seen". Can be communicated to the patient. Therefore, patients stop the exacerbation of their condition by taking therapeutic measures at an early stage.

Therefore, it has been verified that the early diagnosis of a disease by DNB according to the embodiment of the present application is very effective for the early diagnosis of a complicated disease such as lymphoma.

In addition, 22 genes and TFs were contained in the DNB detected in this verification example, of which 13 genes were clearly associated with B-cell lymphoma, and further, the 13 genes. Eight of the genes have been found to be growth master regulators. Therefore, the DNB in the present invention can not only convey signs of abnormality to the patient at an early stage as a warning signal indicating a pre-disease state, but can also specifically identify a gene associated with the disease. It seems to be very useful for the treatment of complex diseases and pharmaceuticals.

The embodiment merely discloses a part of innumerable examples of the present invention, and the design can be appropriately changed in consideration of various factors such as the type of disease and the purpose to be detected. .. In particular, as the factor item, various measurement data can be used as long as the information is obtained by the measurement related to the living body. For example, it is not limited to the measurement data related to the above-mentioned genes, proteins, and metabolites, but can be used as measurement data by quantifying various situations of each site based on an in-vivo image output by a measurement device such as a CT scan. It can be used. Furthermore, in addition to images, it is also possible to measure voice or sound emitted from the body, quantify it, and use it as measurement data.

1 Detection device 10 Control unit 11 Recording unit 12 Storage unit 13 Input unit 14 Output unit 15 Acquisition unit 11a Detection program

Claims (5)

It is a detection device that detects the pre-disease state, which is the process of transition from the normal state to the disease state of the living body to be measured.
A storage means for storing measurement data of a plurality of factor items collected from the living body at different time points, and
A screening means for detecting a differential factor item from the measurement data of the factor item, and
A classification means for classifying the measurement data of the differential factor item detected by the screening means into a plurality of clusters, and a classification means.
Among the plurality of clusters obtained by the classification means, the correlation between the measurement data of the factor items in the cluster is significantly increased, and the standard deviation of the measurement data of the factor items in the cluster is significantly increased. In addition, a selection means for selecting a cluster in which the correlation between the measurement data of the factor item in the cluster and the measurement data of the factor item in the other cluster is significantly reduced, and
A detection device including a display means for displaying that the living body is in the pre-disease state by means of characters, figures, or images when the selection means selects the cluster. The detection device according to claim 1, wherein the selection means tests significance based on a comparison result between the measurement data of the factor item and the preset reference data. The detection device according to claim 1 or 2, wherein the selection means selects a cluster in which the index I calculated by the following formula 1 exceeds a predetermined threshold value.
I = SDd × PCCd / OPCCd ・ ・ ・ Equation 1
However,
PCCd: Average value of absolute values of Pearson correlation coefficients between the measurement data of factor items in the cluster OPCCd: Pearson phase relationship between the measurement data of the factor items in the cluster and the measurement data of the factor items in other clusters. Average value of absolute numbers SDd: Average value of standard deviation of measurement data of factor items in the cluster The factor items are any one of a measurement item related to genes, a measurement item related to proteins, a measurement item related to metabolites, and a measurement item related to images, voices, and sounds obtained from a living body. The detection device according to any one of claims 3. It is a detection program that causes a computer that processes measurement data of multiple factor items collected from the living body to be measured at different time points to detect the pre-disease state, which is the process of transition from the normal state to the disease state of the living body. hand,
On the computer
A screening step for detecting a differential factor item from the measurement data of a plurality of factor items collected from the living body at different time points.
A classification step for classifying the measurement data of the differential factor item detected in the screening step into a plurality of clusters, and a classification step.
Among the plurality of clusters obtained in the classification step, the correlation between the measurement data of the factor items in the cluster is significantly increased, and the standard deviation of the measurement data of the factor items in the cluster is significantly increased. And the selection step to select a cluster whose correlation between the measurement data of the factor item in the cluster and the measurement data of the factor item in the other cluster is significantly reduced.
A detection program characterized in that when the selection step selects the cluster, a display step of displaying that the living body is in the pre-disease state by characters, figures, or images is executed.

Images