WO2015110018A1

WO2015110018A1 - Detection device and detection method for detecting pre-disease state

Info

Publication number: WO2015110018A1
Application number: PCT/CN2015/071237
Authority: WO
Inventors: 陈洛南; 刘锐; 合原一幸
Original assignee: 中国科学院上海生命科学研究院
Priority date: 2014-01-21
Filing date: 2015-01-21
Publication date: 2015-07-30
Also published as: CN104794321B; CN104794321A

Abstract

A detection device (1) and detection method for detecting a pre-disease state. The device (1) comprises a sample acquisition unit (10), wherein the sample acquisition unit (10) acquires comparison sample data and single sample data of a detection object; a DNB setting unit (11), wherein the DNB setting unit (11) sets DNB members; index calculation units (13, 14, 15, 16), wherein the index calculation units (13, 14, 15, 16) obtain composite indexes by calculating the distance between the probability distribution of the comparison sample data and the single sample data; and a state judgement unit (12), wherein the state judgement unit (12) judges that the detection object is in a pre-disease state when the composite indexes exceed a threshold value. Therefore, the DNB theory is effectively applied to single sample conditions, so that the pre-disease state can be simply and accurately detected.

Description

Detection device and detection method for detecting pre-disease state

Technical field

The present invention relates to a detecting device and a detecting method for detecting a pre-disease state.

Background technique

With the environmental pollution, the increase of population and the accelerating pace of life, people's pressure is increasing day by day. Many people have formed bad eating habits and living habits under pressure. In this case, the number of people suffering from various complicated diseases (cancer, diabetes, cardiovascular and cerebrovascular diseases, etc.) has increased, such as the number of people suffering from liver malignant tumors and diseases such as diabetes. Among these complex diseases, some diseases are relatively flat, such as chronic inflammation, which can usually be controlled by drug intervention and health care; but many diseases have sudden deterioration, such as liver cancer, its condition The deterioration is very fast, there is generally no discomfort before the onset of the disease, and once the symptoms appear to go to the hospital for treatment, often the patient is in the middle and late stage, and the survival time after the onset is not much. This type of disease with a sudden deterioration of the disease has a very similar feature, that is, there is a "critical point" or a key node in the course of the disease. (Refer to Non-Patent Documents 1-5) Before the arrival of this critical point, the condition is not particularly obvious, which often causes the patient to ignore the condition and delay the optimal timing of treatment; and after the critical point, the condition does not develop smoothly. However, it suddenly deteriorates from the stable period in a short period of time and becomes a serious illness. It is for this reason that the diagnosis of such diseases is often not timely, which makes the treatment in the critical illness period difficult, the curative effect is poor, and the survival time after the onset is short, so it is very harmful. The key to timely diagnosis of such complex diseases in a timely manner is to find early warning characteristics or signals before the sudden deterioration of the disease, and to predict the conditions of "critical point" and sudden deterioration, which has become a study in biological theory and clinical medical research. Hot Issues.

As shown in (a) of FIG. 3, in general, the "pre-disease state" is a critical state before the "critical point" of disease progression reaches. Appropriate treatment at this stage can return the disease to the "normal state", so it is called the reversible phase. But when the disease progresses faster than the critical point and quickly reaches the "disease state", the treatment is very difficult, and it is difficult to return the disease to a relatively normal state, so it is called the irreversible phase. Pre-disease state The period is the key time node, the molecules that drive the disease state are the key factors, and their regulatory networks are also the key network leading to rapid disease deterioration. Obviously, early prediction and diagnosis of the pre-existing disease state is particularly important in the development of the disease, which is the last chance for patients with many diseases to be effectively controlled. However, unlike the disease state, the normal state is not significantly different from the pre-disease state. Therefore, for many complex diseases, early prediction or pre-diagnosis is a very difficult problem, and there is no effective method. But the increasingly mature high-throughput bio-big data provides a valuable opportunity to fully understand biological processes and their anomalous mechanisms. We can conduct more extensive research on the pathological processes of complex diseases, especially by developing new theories and new methods based on biological big data to identify early warning signals (ie key time nodes or pre-disease states) of complex disease pathological processes. Identify key factors in disease development and extract key networks. This not only clarifies the molecular mechanism of the development of complex diseases, but also helps fight complex diseases and provides new methods and potential drug targets for the prevention, diagnosis and treatment of complex diseases.

In fact, not only complex disease processes, but also in many biological processes, such as cell differentiation, cell proliferation, and disease progression, involve "jumping" state transitions, ie, dramatic changes in the state of the system or qualitative changes. Adipocyte differentiation is such a process. A pluripotent stem cell maintains the potential to differentiate into a variety of cells before becoming a "pre-adipocyte". Once it becomes a pre-adipocyte, it undergoes rapid clonal expansion and subsequent terminal differentiation, thereby producing mature adipocytes. The same is true of the disease progression process. The system gradually transforms from a normal state to a pre-disease state, and then the disease progresses further and rapidly develops into an early state or a disease state. In general, this dramatic change can be described as a bifurcation from a mathematical point of view. Therefore, how to detect key nodes and their key factors from small samples has very important scientific significance in the biological and medical fields.

The results of modern medical and biological research show that the dynamic synergy of various functional modules or biomolecules in various organs of an organism determines the function and state of organs. Therefore, we develop complex diseases and malignant transformation processes. It can be regarded as a dynamic process of time evolution of a complex dynamic system. The external factors affecting disease are regarded as parameters in the dynamic system, and the molecular concentration involved in disease evolution is regarded as a state variable in the system, and the sudden deterioration of the disease It corresponds to the sudden change of the system. The key nodes in the course of the disease correspond to the critical point of the parameters in the dynamic system, especially before the malignant transformation of the disease The period can be regarded as the critical state of the power system. An early warning signal to obtain a malignant transformation becomes a question of how to define a "tipping point", how to detect and identify biological signals in the early stages of malignant transformation, and how to determine whether a complex dynamic dynamic system is at a critical state.

As shown in Figures 1 and 3, the development of the disease can be divided into the following three states.

Normal state, which describes a normal phase or a slightly slower phase of the disease than the disease period, including the latent stage of the disease, the chronic inflammation stage before the cancer, or the disease is effectively controlled and in a relatively healthy stage, which is a relatively stable state. (Figure 1, Figure 3).

Pre-disease state, when the system is in a normal state, if it is continuously driven by external stimuli or some internal factors, then the system enters the pre-disease state, which is a critical stage before the mutation point of disease deterioration arrives (actually normal) a limit of state). The system at this stage is very sensitive to external disturbances. Appropriate treatment can return the disease to a relatively normal period, but if it is not treated in time, the disease can easily cross the mutation point to reach the disease stage (Figure 1, Figure 3).

The state of the disease, which indicates that the condition has deteriorated into a critical illness, or chronic inflammation has been malignantly transformed into cancer. The system is in a steady state again. In general, when the disease reaches this stage, the treatment is very difficult, and it is difficult to return the disease to a relatively normal state (Fig. 1, Fig. 3).

In Fig. 1, (a) shows that the three stages of complex disease development experience normal state, pre-disease state and disease state, respectively, and (b) is normal state, which is the state in which the system is at a local minimum. During this period, the system is in a stable state and gradually or steadily changes. The system in this state has strong resistance to external interference, and (c) is the pre-disease state, which is a critical state and is relatively normal. The limit, that is, is a state before the upcoming fierce transition. This state is still reversible and can be returned to normal under appropriate system parameter disturbances. The system in this state has higher potential energy, so the system is sensitive to external interference when the system is in this state, the external disturbance can drive the system to enter the disease state beyond the critical point, and (d) is the disease state, which is another stable state, the system (e) shows a network at a normal state, (e) shows a network at a normal state, where the color of the node represents the degree to which the gene expression deviates from the mean, the edge represents the correlation between the two genes, and (f) shows a network in a pre-disease state in which there is a set of genes (Z1, Z2, Z3) in the network. The expression deviates from the mean value, and there is a strong correlation between the genes in this group, and the correlation with other genes becomes weak. (g) shows the network under the disease state. In this state, the gene The degree of deviation from the mean value falls back to a smaller extent, and the correlation between the two genes becomes similar to the normal state. It can be seen from (h) that in the pre-disease state, the Z1, Z2, and Z3 expressions are very intense. But the correlation is very high.

Therefore, early prediction and diagnosis of pre-disease status is particularly important, which is an important opportunity for many patients to be effectively controlled. However, there are many difficulties in predicting the pre-disease state. The first difficulty is that the pre-disease state corresponds to the state where the system parameters are close to the critical point. At this time, the system does not undergo a phase change, so the state of the system does not change significantly compared with the normal state. Therefore, it is a very difficult nonlinear problem to accurately predict the early stage of malignant transformation. The second difficulty comes from the complex disease itself, because many complex diseases are the result of a combination of factors such as gene level, transcription level, and protein level. Therefore, although some research on these complex diseases has made some progress. However, to date, no accurate and reliable dynamic models have been constructed for complex diseases to characterize and study the phenomenon of malignant transformation. The third difficulty is from the collection of data. Research on ecosystems, financial systems, etc. can be sampled for a long time and at high density, but this method of data collection cannot be done for studying complex diseases because people will not Go to the hospital frequently before the body feels really unwell. It is based on these questions that the early prediction of the malignant transformation of complex diseases or the diagnosis of “pre-disease state” is a complex nonlinear problem that can only be realized based on small sample data or even single sample data. Such problems are very difficult to solve, so most of the theoretical and experimental work in the past has focused on research on "disease status" or "early disease status." The diagnosis of the disease state is mainly based on molecular biomarkers, such as genes, proteins and metabolic molecules, which can identify the phenotype of the disease, and can distinguish between normal state and disease state by observing its gene expression or protein expression. However, molecular biomarker-based prediction and diagnostic methods are powerless in dealing with early or pre-existing disease states, since pre-disease states are only a limiting stage of relative normal state, and cannot be distinguished at the level of expression and the like. Disease state and normal state.

In this regard, the inventors of the present invention have proposed a pair to show from a normal state to a disease state A method of detecting a candidate marker of a biomarker for an early warning signal of a pre-existing state before a state transition (Non-Patent Document 6). According to this method, early detection of disease can be achieved by detecting a dynamic network marker (DNB) that occurs immediately after a disease state is transferred.

However, the detection method described in Non-Patent Document 6 can only perform early prediction of diseases for multi-sample data. Early predictions of complex diseases often face difficulties in data collection, that is, the study of complex diseases, especially for the clinical application of most complex diseases, cannot be sampled for long periods of time, because people do not Go to the hospital frequently before the body feels really unwell. Therefore, the early prediction of the malignant transformation of complex diseases or the diagnosis of the "pre-disease state" is a complex nonlinear problem that can only be realized based on small sample data or even single-sample data. In this case, the above conventional DNB-based prediction method cannot be used because there is only one sample.

Therefore, in order to effectively apply the method based on DNB detection in a single sample situation, it is necessary to provide a new method for testing a single sample based on DNB.

current technology

[Non-Patent Document 1] Venegas, J.G., et al., "Self-organized plaque in asthma as a prelude to catastrophe", UK, Nature, Nature Publishing Group, 2005, Vol. 434, pp. 777-782

[Non-Patent Document 2] McSharry, PE, Smith, LA, Tarassenko, L, "Predicting seizures: Is it appropriate to use a nonlinear method", UK, Nature Medicine, Nature Publishing Group, 2003, Vol. 9 , pp. 241-242

[Non-Patent Document 3] Roberto, PB, Eliseo, G., Josef, C., "Conversion Model for Evaluating the Change Point of Logistic Regression", USA, Medical Statistics, Wiley Blackwell Publishers, 2003, Vol. 22, pp. 1141-1162

[Non-Patent Document 4] Paek, S., et al., "Hearing preservation after gamma knife surgery on vestibular schwannomas", USA, Cancer, Wiley Blackwell, 2005, Vol. 1040 , pp. 580-590

[Non-Patent Document 5] Liu, J.K., Rovit, R.L., Couldwell, W.T., "Pituitary Stroke", USA, Proceedings of Neurosurgery, Thieme Press, 2001, Vol. 12, pp. 315-320

[Non-Patent Document 6] Chen Luonan, Liu Rui, Liu Zhiping, Li Meiyi, and He Yuanyi, "Detecting Early Warning Signals for Sudden Deterioration of Complex Diseases by Dynamic Network Markers", Science Report, March 29, 2012 Day, Internet (website: http://www.natureasia.com/ja-jp/srep/abstracts/35129)

Summary of the invention

The present invention has been completed in view of the above problems, and based on the DNB theory, a detection device and a detection method for predicting a malignant mutation of a complex disease based on a single sample (high-throughput data) have been developed, and the DNB theory can be effectively applied to a single-sample case. It is impossible to do with traditional prediction methods.

A first aspect of the present invention provides a detecting apparatus for detecting a pre-disease state, comprising: a sample acquiring unit that acquires collation sample data and single-sample data of a detection object; and a DNB setting unit, The DNB setting unit sets a DNB member; an index calculating unit that obtains a composite index by calculating a distance between a probability distribution of the reference sample data and the single sample data; and a state discriminating unit that When the composite index exceeds the threshold, it is determined that the test subject is in the pre-disease state.

In another preferred example, the index calculation unit includes: a first index calculation unit that uses a distance between the comparison sample data and a probability distribution of the DNB members in the one-sample data as the first index; 2 index calculation unit, the second index calculation unit compares the distance between the probability distributions of the DNB members and the non-DNB members in the sample data as the second index; the third index calculation unit, the third index calculation unit compares the sample data And a distance between the probability distributions of the non-DNB members in the single sample data as the third index; and a correction value setting unit that sets the correction value, the index calculation unit is based on the first index, the first The 2 index, the 3rd index, and the correction value are combined.

In another preferred example, the index calculation unit uses a product of the first index, the second index, and the reciprocal of the sum of the third index and the correction value as a composite index.

In another preferred embodiment, the distance between the probability distributions is a KL distance.

In another preferred embodiment, the correction value is a positive number equal to or less than one.

In another preferred embodiment, the correction value is 0.01.

In another preferred embodiment, an output unit is further included, which outputs a probability distribution of the DNB members in an analog display manner.

A first aspect of the present invention provides a detection method for detecting a pre-disease state, comprising the steps of: a sample acquisition step of acquiring comparison sample data and single sample data of a detection object; a DNB setting step of setting a DNB a member; an index calculation step of obtaining a composite index by calculating a distance between a probability distribution of the reference sample data and the single sample data; and a state discriminating step of determining that the detected object is in a pre-disease state when the composite index exceeds the threshold.

The present invention provides a detection device and a detection method for predicting a malignant mutation of a complex disease based on a single sample (high-throughput data), and can effectively apply the DNB theory to a single-sample situation, which can be simple and accurate. The pre-disease state was detected.

DRAWINGS

Figure 1 is a schematic diagram showing three stages of the development of a complex disease.

Fig. 2 is a block diagram showing the configuration of a detecting device of the present invention.

3 is a schematic diagram showing mutations predicting complex diseases based on single sample data in accordance with an embodiment of the present invention.

4 is a schematic diagram showing numerical simulation results for verifying the correctness and reliability of an embodiment of the present invention.

FIG. 5 is a schematic diagram showing an example of detecting according to an embodiment of the present invention.

FIG. 6 is a schematic diagram showing another example of detecting according to an embodiment of the present invention.

detailed description

The contents of the present invention are further clarified below with reference to the accompanying drawings, but the following examples are not intended to limit the scope of the invention.

A first aspect of the invention is a detection device for detecting a pre-disease state. Fig. 3 is a block diagram showing the configuration of a detecting device of the present invention. Unlike the multi-samples in the prior art, single-sample (high-throughput data) is utilized in the present embodiment to predict mutations in complex diseases. As shown in Figure 2, detection The apparatus 1 includes: a sample acquisition unit 10 that acquires comparison sample data and single sample data of the detection object; a DNB setting unit 11 that sets a DNB member; and an index calculation unit that calculates a probability distribution between the comparison sample data and the single sample data The distance is obtained to obtain a composite index; and the state discriminating unit 12 determines that the detected object is in the pre-disease state when the composite index exceeds the threshold. The control sample data and the single sample data refer to the two sets of control and case samples, in which the control group (control group) contains a large number of samples, and the case group contains only one sample (single sample). The setting method of the DNB member is well known, and the setting method described in Non-Patent Document 6 can be used, and will not be described herein.

More specifically, the index calculation unit includes: a first index calculation unit 13 that takes the distance between the comparison sample data and the probability distribution of the DNB members in the one-sample data as the first index; the second index calculation unit 14 compares the sample data The distance between the probability distributions of the middle DNB members and the non-DNB members is taken as the second index; the third index calculating unit 15 takes the distance between the comparison sample data and the probability distribution of the non-DNB members in the single sample data as the third index; The correction value setting unit 16 sets the correction value, and the index calculation unit obtains the composite index based on the first index, the second index, the third index, and the correction value. It should be understood that it is also possible to use a single indicator calculation unit to calculate the distance between the probability distribution of the control sample data and the single sample data, and subdividing into the above four units is only an alternative.

In a preferred embodiment, the composite index may be a product of the first index, the second index, and a reciprocal of the sum of the third index and the correction value.

The calculation of the composite index will be described in detail below with reference to FIG. 3 is a schematic diagram showing mutations predicting complex diseases based on single sample data in accordance with an embodiment of the present invention. Since it is necessary to estimate the degree of similarity of the distribution of the two sets of samples, that is, to calculate the probability distance of the distribution of the two sets of samples, in the present invention, no matter which method of calculating the probability distance is used, The present invention is implemented, such as histogram intersection distance, chi-square test, quadratic distance, matching distance, Kolmogorov-Smirnov test (KS distance), earth moving distance, KL distance, and the like. It is preferable to use the K-L distance for calculation, because the calculation using K-L distance is the smallest, the most widely used, and the highest precision. K-L distance is also called relative entropy, KL divergence, information divergence, information gain, and is two probability points. A measure of the asymmetry of the difference in cloth. The significance is: in the same event space, the event space of the probability distribution P(k), if encoded by the probability distribution Q(k), averages how many bits per basic event (symbol) code length is increased.

The Kullback-Leibler distance (K-L distance) is defined as:

Where P and Q are respectively two discrete probability distributions, P(k)=Prob _P (x=x _k ), Q(k)=Prob _Q (y=y _k ), and

Kullback-Leibler distance can also be written

D _KL (P,Q)=H(P,Q)-H(P)

That is, the KL distance is a conditional entropy, which is mainly used to judge the distance between two probability distributions, and is also used to calculate the similarity between two samples A and B. If D _KL (P _A , P _B ) = 0, then the samples A and B contain the same information, and the similarity between the two samples is maximized.

The following composite index I is designed using the K-L distance:

This composite index I is called DNB-S score, where the KL distance between the Case data and the DNB members in the control data (D _KL (case _DNB , control _DNB )) corresponds to the above-mentioned first index, the DNB member in the Case data. The KL distance between the non-DNB member and the non-DNB member (D _KL (case _DNB , case _non-DNB )) corresponds to the KL distance between the non-DNB members in the second index, Case data, and control data (D _KL (case _{non- DNB} , control _non-DNB )) corresponds to the above third index, ε is a small positive number, in order to avoid the occurrence of the zero denominator correction value, taken as 0 to 1, preferably 0.01. In the composite index of the above design, the composite index is the product of the first index, the second index, and the reciprocal of the sum of the third index and the correction value, but this design is only an example, as long as it is based on the first index, the first The 2 index, and the 3rd index and the correction value can also be designed in other ways.

As shown in Figure 3, (a) shows that the development of a complex disease is divided into three states, and (b) shows the basis for a one-sample prediction of the disease system, that is, based on a case sample, many control samples (control group Samples, such as samples of healthy people), existing DNB networks. These samples may be, for example, gene expressions, based on the DNB theory, whether the case sample is in the pre-disease state, and (c) showing the distribution obtained using the KL distance, which can be seen in the case data when the system is in the pre-disease state. The DNB biomolecules have a bimodal distribution, while the DNB biomolecules in the control data have a unimodal distribution. Meanwhile, the non-DNB molecules in the case data have a unimodal distribution, which makes the single case sample taken from the pre-disease state. The difference between the distribution of DNB and non-DNB, the distribution of DNB in case data and control data respectively can provide a significant signal, and (d) shows the prediction using the new composite index I, DNB-S score, when the composite indicator When the threshold value is exceeded, it is judged that the detection object is in the pre-disease state, and the indicator can be calculated only by a single case sample. When the system is in a normal state or disease state, the DNB-S score has a lower value, and when the system is in the pre-disease state, the DNB-S score is at a higher value, which can provide a reliable signal for early warning of system mutation.

According to the nature of DNB, when the system is close to the pre-disease state, the DNB-S score has the following characteristics:

The KL distance (D _KL (case _DNB , control _DNB )) between the DKB members in the Case data and the control data is increased because the distribution of biomolecules in the case data and the control data in the DNB is significantly different, that is, in the case data. The DNB molecule exhibits a bimodal distribution, while the DNB molecules in the control data exhibit a unimodal distribution (see Figure 3).

The KL distance (D _KL (case _DNB , case _non-DNB )) between the DNB member and the non-DNB member in the Case data increases, because even in the same case data, the biomolecule in the DNB exhibits a bimodal distribution in the case, and The non-DNB molecule exhibited a unimodal distribution (see Figure 3).

These two properties allow the DNB-S score to accurately indicate the pre-disease state. On the other hand, when the system crosses the critical point and enters the disease state, the third item in the DNB-S score has the following properties:

When the system is in a normal state or a pre-disease state, there is no significant change in the KL distance ( _DKL (case _non-DNB , control _non-DNB )) between the Case data and the control data; however, when the system crosses At the critical point, after entering the disease state, the distribution of biomolecules in non-DNB in case data and control data is significantly different, that is, the expression level of non-DNB molecules in disease state (ie, in case data) is higher (or The mean of the low, and thus the mean of the distribution is larger (or smaller).

According to the characteristics of the above DNB-S score, we can see that the first two items can be used to distinguish whether a single sample is in a normal state or a pre-disease state, and the third item can be used to distinguish whether a single sample is in a pre-disease state or in a disease state. status.

Therefore, when it is judged that the detection subject is in the pre-disease state, the state discriminating unit 12 can issue an early warning signal.

Further, the detecting means may further include an output unit (not shown in Fig. 2) for outputting the probability distribution of the DNB members in an analog display manner (refer to (c) in Fig. 3).

A second aspect of the invention is a detection method for detecting a pre-disease state, the detection method comprising the following steps.

First, the comparison sample data and the single sample data of the detection object are acquired. In the detecting device of the first aspect of the invention, the sample acquiring unit 10 performs this step.

Next, set the DNB member. In the detecting device of the first aspect of the invention, the DNB setting unit 11 performs this step.

Then, the composite index is obtained by calculating the distance between the probability distribution of the control sample data and the single sample data. In the detecting device of the first aspect of the invention, the index calculating unit performs the step.

Finally, when the composite index exceeds the threshold, it is determined that the test subject is in the pre-disease state. In the detecting device of the first aspect of the invention, the state discriminating unit 12 performs this step.

In the execution of the index calculation step, the detection means of the first aspect of the present invention may be similar to the composite index, and the indicator calculation step further includes: a first index calculation step, and the control sample data and the single sample data in the DNB The distance between the probability distributions of the members is taken as the first index; the second index calculation step is to use the distance between the probability distributions of the DNB members and the non-DNB members in the comparison sample data as the second index; the third index calculation step, which will be compared The distance between the sample data and the probability distribution of the non-DNB members in the single sample data is taken as the third index; and the correction value setting step is performed to set the correction value.

In the index calculation step after the above-described subdivision, the composite index is obtained based on the first index, the second index, the third index, and the correction value.

Preferably, in the index calculation step, the product of the first index, the second index, and the reciprocal of the sum of the third index and the correction value is used as a composite index.

Similarly, in the calculation of the composite index, the present invention can be implemented regardless of the method of calculating the probability distance, but it is preferable to perform calculation using the K-L distance. The composite index I designed by the K-L distance is also the same as the detecting device in the first aspect of the present invention described above, and will not be described herein.

The correctness and reliability of the embodiment of the present invention are verified below with reference to FIG. 4 is a schematic diagram showing numerical simulation results for verifying the correctness and reliability of an embodiment of the present invention.

Numerical simulations were performed using a 16-dimensional network, which is the following Michaelis-Menten model:

The range of the parameter P is -0.1 to 0.32, and ζi(t) is white noise. We use the Euler method to simulate the process from the normal state (P>0) to the critical point (P=0) and into the disease state (P<0), as shown in Figure 4. This numerical simulation demonstrates the correctness and reliability of the composite index I, that is, the DNB-S score set in the above embodiment of the present invention.

Figure 4(a) shows a 16-dimensional Mie network in which nodes 1-7 belong to DNB members, nodes 9-16 belong to non-DNB members, and (b) show predictions of 5 sets of single-sample data. As a result, when the system approaches the critical point (parameter P=0), using the DNB-S score, an accurate warning signal can be found, that is, the five groups of BS scores all rise significantly, exceeding the threshold, and (c) shows the analog display. The DNB members (nodes 1-7) in the case data have significantly different distributions, ie, the DNBs in the case data exhibit a bimodal distribution (both sides), while the DNBs in the control data exhibit a unimodal distribution (central), (d ) shows that the simulation shows that even though they are in the case data, the DNB members (nodes 1-7) and the non-DNB members (nodes 8-16) have significantly different distributions, ie the DNB members exhibit a bimodal distribution (both sides), Non-DNB members exhibit a unimodal distribution (central).

The detection device and the detection method of the present invention, that is, the validity and accuracy of the composite index I, that is, the DNB-S score, are tested using real clinical data or test data. The specific use of the DNB-S score is based on published data: (1) clinical trial data on the incidence of influenza after injection of H3N2 virus in an individual (GSE30550); (2) acute injury to the lungs in mice exposed to toxic gases Gene data (GSE2565).

Test Example 1: Cold disease (influenza)

FIG. 5 is a schematic diagram showing an example of detecting according to an embodiment of the present invention. In Figure 5, it is shown that DNB has been successfully applied in a single-sample early detection of a particular disease, a clinical trial of individual morbidity after injection of a cold virus (high-throughput test data GSE30550). The data of 17 individuals were tested by applying the DNB-S score (each individual had only one sample at one sampling time point), and 9 of them developed symptomatic subjects after the injection of the cold virus. Eight individuals did not develop cold symptoms (asymptomatic subjects) after the injection of the cold virus. As shown in Figure 5, in (a), 9 individuals with morbidity were examined and passed before each individual became ill. The DNB-S score detected that it was in the pre-disease state (beyond the threshold) and signaled an early warning; in (b), eight individuals who did not develop the disease and found no significant change in the DNB-S score (not Exceeded the threshold). Therefore, it was demonstrated that the DNB-S score is valid and accurate for this clinical trial data. Specifically, the 9 affected individuals were tested before the 7th sampling time point (36 hours), and in the trial, all the affected individuals were at the 8th sampling time point (45 hours) or after. The symptoms were detected (see Figure 5). For the 8 non-infected individuals, the pre-disease status was not detected by the DNB-S score, so no warning signal was issued. This further application in clinical trial data demonstrates the validity and accuracy of the DNB-S score. In addition, (c) is a dynamic map of the entire biomolecular network depicting changes in the overall network structure from 0 hours to 45 hours. (d) Detailed records of the onset and morbidity of each individual, the time of onset, the time point at which the pre-existing disease state has been detected by the DNB-S score, and the individual on the onset of the disease can also be seen. This test is timely. More specifically, in Fig. 5, the data of the colds of 17 individuals were predicted using the DNB-S score, and (a) and (b) showed the curves of 17 individuals using the DNB-S score for testing. Nine individuals developed a symptomatic (symptomatic) after injection of the cold virus (H3N2virus), and the pre-existing disease state (beyond the threshold) was detected in the early stage of the lesion using the DNB-S score ((a)), and 8 individuals were injected with a cold. The virus (H3N2virus) did not develop lesions (asymptomatic), their data was insensitive to the DNB-S score (not exceeding the threshold) ((b)), and (c) was the dynamic network mapped to the first individual (1st subject) Development map, using the existing PPI network, to map the data to the network, wherein the color of the node represents the fluctuation of the expression of the biomolecule, and the molecules belonging to the DNB members are specially arranged in the lower left corner, and can be seen Under normal conditions (0-12 hours), there is no big change in the network structure of the system, but in the pre-disease state (36 hours), the network structure of the system (especially the DNB members in the lower left corner) changes greatly. Provides clear disease warning The signal, (d) is a table of information on the onset and morbidity of 17 individuals based on clinical data, the time of onset, and the time point at which the disease was detected using the DNB-S score. It can be seen that the DNB-S score is indeed before the onset of the individual. An early warning signal is issued to accurately detect the pre-existing disease state, and there is no response to the individual who does not develop the disease.

Test Example 2: acute lung injury

FIG. 6 is a schematic diagram showing another example of detecting according to an embodiment of the present invention. In Figure 6, it is shown that the DNB-S score was successfully applied in a single-sample early detection of another specific disease, phosgene inhalation acute lung injury (high-throughput test data GSE2565). The data was taken from an acute lung injury test in which six individuals (mouse) were exposed to phosgene. There were a total of 9 sampling time points. At each sampling time point, each individual had a case sample and a Control (control) sample. The data was initially detected for disease using the DNB-S score (Figure 6). As shown in Fig. 6, (a) shows a curve of six individuals using the DNB-S score for testing. From this figure, it can be seen that six test subjects exposed to phosgene are at the fourth time point ( At 4 hours), a significant signal (beyond the threshold) that can be judged to be in the pre-disease state was detected, and according to the experimental observation, the six individuals found the lung at the fifth and sixth time points (8 hours–12 hours). Symptoms of injury. Therefore, using the DNB-S score, the pre-disease state was accurately judged, and the acute mutation of the system was successfully alerted, and (b) the curve of the individual using the DNB-S score was examined for the individuals in the six control groups. It can be seen that DNB The -S score did not respond to these 6 test subjects not exposed to phosgene and did not detect any warning signals (not exceeding the threshold). The successful application of DNB-S score in this disease also demonstrates the effectiveness and accuracy of embodiments of the present invention.

More specifically, in Fig. 6, the data of acute injury of lungs in 6 individuals were predicted by DNB-S score, and (a) the curves of 6 individuals using the DNB-S score were examined. It was observed that the six test subjects exposed to phosgene detected a pre-existing condition at the fourth time point (4 hours) and sent a clear signal (beyond the threshold), and according to experimental observations, Individuals were found to have symptoms of lung injury at the fifth and sixth time points (8 hours–12 hours). Therefore, the pre-disease status was accurately determined using the DNB-S score, and the system was successfully alerted to acute mutations. ) The curves of the individuals in the 6 control groups using the DNB-S score were shown. It can be seen that the DNB-S score did not respond to the 6 test subjects not exposed to phosgene, and no warning was detected. The signal (not exceeding the threshold), the successful application of the DNB-S score on the disease demonstrates the effectiveness and accuracy of embodiments of the present invention.

As described above, the present invention effectively applies the DNB theory to a single sample case, not only the structure and the square. The method is simple and efficient and accurate.

The above description is only a preferred embodiment of the present invention, and thus the scope of the present invention is not limited thereto, and those skilled in the art can make other use of the technical solutions and technical concepts of the present invention. Corresponding changes and modifications are intended to be included within the scope of the appended claims.

Description of the reference numerals

1 detection device

10 sample acquisition unit

11 DNB setting unit

12 state discriminating unit

13 1st index calculation unit

14 2nd index calculation unit

15 3rd index calculation unit

16 correction value setting unit

Claims

A detecting device for detecting a pre-disease state, comprising:

a sample acquisition unit that acquires comparison sample data and single sample data of the detection object;

DNB setting unit, the DNB setting unit sets a DNB member;

An index calculation unit that obtains a composite index by calculating a distance between a probability distribution of the comparison sample data and the single sample data;

The state discriminating unit determines that the detection target is in the pre-disease state when the composite index exceeds the threshold.
The detecting device according to claim 1, wherein

The indicator calculation unit includes:

a first index calculation unit that uses a distance between the comparison sample data and a probability distribution of the DNB members in the one-sample data as the first index;

a second index calculation unit that compares a distance between a probability distribution of a DNB member and a non-DNB member in the sample data as a second index;

a third index calculation unit that uses, as a third index, a distance between the probability distribution of the non-DNB members in the comparison sample data and the one-sample data;

a correction value setting unit that sets the correction value,

The index calculation unit obtains a composite index based on the first index, the second index, the third index, and the correction value.
The detecting device according to claim 2, wherein

The index calculation unit uses a product of the first index, the second index, and the reciprocal of the sum of the third index and the correction value as a composite index.
The detecting device according to any one of claims 1 to 3, characterized in that

The distance between the probability distributions is the KL distance.
The detecting device according to claim 2, wherein

The correction value is a positive number equal to or less than one.
A detecting device according to claim 5, wherein

The correction value is 0.01.
The detecting device according to claim 1, wherein

The state discriminating unit issues an early warning signal when it is judged that the detection subject is in the pre-disease state.
The detecting device according to claim 1, wherein

Also included is an output unit that outputs a probability distribution of the DNB members in an analog display.
A detection method for detecting a pre-disease state, characterized in that it comprises the following steps:

a sample acquisition step of obtaining comparison sample data and single sample data of the detection object;

DNB setting steps to set the DNB member;

An indicator calculation step of obtaining a composite index by calculating a distance between a probability distribution of the control sample data and the single sample data;

The state discriminating step determines that the detected subject is in the pre-disease state when the composite index exceeds the threshold.
The detecting method according to claim 9, wherein

The indicator calculation step includes:

a first index calculation step of using a distance between the comparison sample data and a probability distribution of the DNB members in the single sample data as the first index;

a second index calculation step of using a distance between a probability distribution of a DNB member and a non-DNB member in the sample data as a second index;

a third index calculation step of using a distance between the comparison sample data and a probability distribution of non-DNB members in the one-sample data as the third index;

Correction value setting step, setting the correction value,

In the index calculation step, a composite index is obtained based on the first index, the second index, the third index, and the correction value.
The detecting method according to claim 9, wherein

In the index calculation step, a product of the first index, the second index, and the reciprocal of the sum of the third index and the correction value is used as a composite index.
The detecting method according to any one of claims 9 to 11, wherein

The distance between the probability distributions is the KL distance.