A key gene prediction method based on network information entropy

Abstract

The present invention provides a key gene prediction method based on network information entropy, which specifically includes the following steps: S1, constructing a background network, S2, constructing a reference sample network, S3, constructing a perturbation sample network, S4, calculating network information entropy and difference network information entropy, and S5, using the difference network information entropy to calculate the global difference network information entropy, and selecting the key local difference network information entropy to predict the key gene. The present invention is based on basic theories such as dynamic network biomarkers, sample-specific networks, and information entropy, accurately monitors the volatility of complex biological systems under a dynamic background, and proposes a new method to reveal potential correlation information between nodes by analyzing network connections, that is, edges as the minimum analysis unit, and identify key network structures that cause significant fluctuations, which is conducive to early warning of specific genes and identification of key pathogenic gene groups.

Description

Technical Field

[0001] The invention belongs to the technical field of biosystems, and in particular relates to a key gene prediction method based on network information entropy. Background Art

[0002] For most complex biological systems, biological processes are far from being in a simple static or relatively balanced state. Instead, over time, the state of these biological systems can be slightly disturbed by the internal or external environment, triggering dramatic fluctuations and potentially initiating cascade reactions in the system's physicochemical networks, leading to large-scale, often irreversible changes in the biological system, which in turn trigger the emergence of disease states or the deterioration of the disease.

[0003] In the field of modern medicine, traditional biomarkers play an important role in disease diagnosis and treatment monitoring. By detecting specific molecules, proteins or genetic markers in blood, urine or other biological samples, doctors are able to assess a person's health status or the presence or absence of a disease. These markers provide an important basis for early detection and monitoring of diseases, but they are often static and cannot fully reflect the dynamic changes of biological systems during disease progression. Therefore, traditional biomarkers have their limitations in early warning and personalized treatment of complex diseases such as cancer or neurodegenerative diseases.

[0004] In recent years, with the rise of the concept of "systems biology", scientists have begun to pay attention to the complexity and dynamics of biological systems, especially in the critical state of disease, that is, the critical stage when the biological system changes from health to disease. At this stage, the biological system may show some small but decisive changes, which are usually sudden and difficult to predict with traditional methods. Further analysis of changes in these critical states is particularly critical for achieving early detection and personalized treatment of diseases. Summary of the invention

[0005] The purpose of the present invention is to provide a key gene prediction method based on network information entropy in response to the problems in the prior art.

[0006] To this end, the above-mentioned purpose of the present invention is achieved through the following technical solutions:

[0007] A key gene prediction method based on network information entropy, characterized in that it specifically comprises the following steps:

[0008] S1, constructing a background network: using specific relationship data as input data of the background network, and screening the relationship data according to a given threshold to construct the background network of the given threshold;

[0009] S2, constructing the reference sample network: using the omics data of a specific representation as the initial input data of the reference sample, calculating its correlation coefficient and standard deviation, and applying the correlation coefficient to the background network to construct the reference sample network;

[0010] S3, constructing a perturbation sample network: using the omics data of a specific representation as the initial input data of the perturbation sample, concatenating the initial input data of a single perturbation sample with the initial input data of the reference sample in the sample dimension, calculating the correlation coefficient and standard deviation between the concatenated data, and applying the correlation coefficient between the concatenated data to the background network to construct the perturbation sample network;

[0011] S4, calculate network information entropy and difference network information entropy: use the reference sample network and the perturbation sample network to calculate their network information entropy respectively, take the absolute value of the difference between the two multiplied by the difference of the standard deviation, and get the difference network information entropy;

[0012] S5, using the difference network information entropy to calculate the global difference network information entropy, sorting the difference network information entropy values ​​in step S4 from high to low, and selecting the edge of the difference network information entropy ranked in the top 5% as the key local difference network information entropy to predict the key genes;

[0013] Among them, in step S4, the calculation formula of the difference network information entropy is as follows:

[0014] ΔH E (s+1)=|SD E (s+1)-SD E (s)|·|H E (s+1)-H E (s)|

[0015] Where H E (s) is the network information entropy of a certain edge E in the case of s samples, SD E (s) is the standard deviation of a certain edge E in the case of s samples, (s+1) represents the case where a perturbation sample is added to the reference sample network, and the edge specifically refers to the adjacent nodes in the network. While adopting the above technical solution, the present invention can also adopt or combine the following technical solutions:

[0016] As a preferred technical solution of the present invention: the specific relationship data includes data in a protein interaction database, a biological pathway database, and a molecular interaction database;

[0017] The specific omics data include genome, transcriptome, metabolome, proteome and other omics data of human peripheral blood, tumor slice tissue of cancer patients, experimental mice, etc., as well as food chain data and disease transmission model data.

[0018] As a preferred technical solution of the present invention: the correlation coefficients in step S2 and step S3 include Pearson correlation coefficient, Spearman correlation coefficient, mutual information and MIC maximum information coefficient.

[0019] As a preferred technical solution of the present invention: the Pearson correlation coefficient measures the linear relationship between the initial input data, and the background network is applied to the relationship matrix generated by the correlation coefficient to obtain the reference sample network, wherein the calculation formula of the Pearson correlation coefficient is as follows:

[0020]

[0021] In the formula, x and y represent two nodes respectively, X n and Y n Represent the values ​​of the two nodes on the nth sample, and They represent the arithmetic mean of the values ​​of the two nodes in all samples, and s represents the number of samples.

[0022] That is, the Pearson correlation coefficient is used to measure the strength of the correlation between two variables. When genomics data is used to construct a correlation network, the network nodes represent genes and the edges represent the strength of the association between genes.

[0023] The formula for calculating the standard deviation is as follows:

[0024]

[0025] Where x represents a specific node, X n Represents the value of the node on the nth sample, Represents the arithmetic mean of the values ​​of the node in all samples, and s represents the number of samples.

[0026] As a preferred technical solution of the present invention: the calculation formula of the difference network information entropy is as follows:

[0027] ΔH E (s+1)=|SD E (s+1)-SD E (s)|·|H E (s+1)-H E (s)|

[0028] In the formula SD E (s) is the standard deviation of the edge, H E (s) is the network information entropy of the edge, s represents the number of reference samples, and s+1 represents the number of samples after adding a perturbation sample to the reference sample.

[0029] As a preferred technical solution of the present invention: the calculation formula of the standard deviation of the edge is as follows:

[0030]

[0031] The calculation formula of the edge network information entropy is as follows:

[0032]

[0033] In the formula and is the node standard deviation of the two connected nodes g1 and g2 corresponding to a specific edge E, and is the node network information entropy of the two connected nodes g1 and g2 corresponding to a specific edge E, α and , respectively represent the weights of the two nodes;

[0034] The node weights α and β include but are not limited to protein molecular weight, protein length, and arithmetic mean values. As a preferred technical solution of the present invention: the calculation formula of node network information entropy is as follows:

[0035]

[0036] In the formula, x and y represent two nodes respectively, s represents the number of samples, M represents the number of first-order neighbor nodes y corresponding to node x, and p x，y (s) represents the correlation probability between the values ​​of the two nodes x and y, and its calculation formula is as follows:

[0037]

[0038] Where x and y represent two connected nodes, s represents the number of samples, and M represents the number of first-order neighbor nodes y corresponding to node x.

[0039] As a preferred technical solution of the present invention: the global difference network information entropy can roughly determine the fluctuation of individuals at the overall level. According to the fluctuation at the overall level, one or more moments with higher global difference network information entropy values ​​can be found. According to the edges corresponding to the moments with higher values, some edges are screened out as key local difference networks by a certain method, so as to determine the key local difference network information entropy. The screening method of the key local difference network includes but is not limited to screening edges with higher difference network information entropy according to a certain proportion, screening edges with higher difference network information entropy according to a certain number, and using a certain difference network information entropy as a threshold to screen edges with the highest degree, etc.;

[0040] The calculation formula of global difference network information entropy is as follows:

[0041]

[0042] Where ΔH e (s+1) is the difference network information entropy of a single edge e, and WG is the global network, that is, the network composed of all edges.

[0043] The calculation formula of the key local difference network information entropy is as follows:

[0044]

[0045] Where ΔH e (s+1) is the difference network information entropy of a single edge e, and WL is the key local network, that is, the network composed of edges selected by a certain method.

[0046] As a preferred technical solution of the present invention: the screening method of the key local difference network includes but is not limited to screening edges with higher difference network information entropy according to a certain proportion, screening edges with higher difference network information entropy according to a certain number, and using a certain difference network information entropy as a threshold to screen edges with the highest degree.

[0047] As a preferred technical solution of the present invention: the specific relationship data of the background network includes but is not limited to protein interaction database, biological pathway database, molecular interaction database and other data; the omics data of specific phenomena include but are not limited to genome, transcriptome, metabolome, proteome and other omics data of human peripheral blood, tumor slice tissue of cancer patients, experimental mice, etc., as well as food chain data and disease transmission model data; the initial input data of the reference sample includes but is not limited to the omics data of specific phenomena from healthy humans, paracancerous tissues, control groups, etc.; the initial input data of the perturbation sample includes but is not limited to the omics data of specific phenomena from diseased humans, cancer tissues, experimental groups, etc.

[0048] Compared with the prior art, the key gene prediction method based on network information entropy of the present invention uses specific relationship data as background network input, uses omics data as initial input of reference and perturbation samples, calculates correlation coefficients and standard deviations of reference samples and perturbation samples, and further calculates network information entropy of each node based on this, and calculates network information entropy of edges between adjacent nodes on this basis. Based on the network information entropy of the edge, the difference between the reference sample and the reference sample with a single perturbation sample added is calculated and compared, so as to realize the quantification of volatility of complex biological networks, provide reference for disease warning and mining of key warning genes, and provide reference for accurate diagnosis of specific genes.

[0049] The present invention provides a key gene prediction method based on network information entropy, based on basic theories such as dynamic network biomarkers, sample-specific networks, and information entropy, to accurately monitor the volatility of complex biological systems in a dynamic context, and proposes a new method to reveal potential correlation information between nodes and identify key network structures that cause significant fluctuations by analyzing network connections, i.e., edges, as the smallest analysis unit, which is beneficial for early warning of specific genes and identification of key pathogenic gene groups, and provides a reference for medical workers to conduct targeted treatment strategy analysis at the gene network level, and provides a reference for achieving early intervention and timely treatment, which helps to prevent the development or deterioration of various acute and chronic diseases, and helps to achieve forward-looking disease warning and block the development of the disease as early as possible. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is a schematic diagram of the flow of a key gene prediction method based on network information entropy of the present invention. DETAILED DESCRIPTION

[0051] The present invention will be described in further detail with reference to the accompanying drawings and specific embodiments.

[0052] Example 1

[0053] like Figure 1 As shown, a key gene prediction method based on network information entropy specifically includes the following steps:

[0054] S1: Step of constructing a background network: using specific relationship data as input data of the background network, filtering the relationship data with a given threshold, thereby constructing a background network corresponding to the threshold. The specific relationship data of the background network includes, but is not limited to, data such as protein interaction database, biological pathway database, and molecular interaction database.

[0055] Specifically, the relationships with a score > 800 in the StringDB protein interaction database were used as the background network.

[0056] S2: Steps for constructing a reference sample network: Use the omics data of specific appearances as the initial input data of the reference sample, calculate the correlation coefficient and standard deviation between the initial input data, and apply the correlation coefficient between the initial input data to the background network to construct the reference sample network. The omics data of specific appearances include, but are not limited to, genome, transcriptome, metabolome, proteome and other omics data of human peripheral blood, tumor slice tissue of cancer patients, experimental mice, etc. This type of data is mainly gene expression data, mainly RNA-seq data, and other data that meet the standards, such as food chain data and disease transmission model data, are also acceptable.

[0057] Specifically, the Pearson correlation coefficient is used to calculate and evaluate the correlation between two variables, and the background network is applied to the relationship matrix generated by the correlation coefficient to obtain the reference sample network.

[0058] The specific calculation formula of Pearson correlation coefficient is as follows:

[0059]

[0060] In the formula, x and y represent two nodes respectively, X n and Y n Represent the values ​​of the two nodes on the nth sample, and They represent the arithmetic mean of the values ​​of the two nodes in all samples, and s represents the number of samples.

[0061] That is, the Pearson correlation coefficient is used to measure the strength of the correlation between two variables. When genomics data is used to construct a correlation network, the network nodes represent genes and the edges represent the strength of the association between genes.

[0062] The specific standard deviation calculation formula is as follows:

[0063]

[0064] Where x represents a specific node, X n Represents the value of the node on the nth sample, Represents the arithmetic mean of the values ​​of the node in all samples, and s represents the number of samples.

[0065] S3 steps to construct the perturbation sample network: using the omics data of a specific phenomenon as the initial input data of the perturbation sample, concatenating the initial input data of a single perturbation sample with the initial input data of the reference sample in the sample dimension, calculating the correlation coefficient and standard deviation between the concatenated data, and applying the correlation coefficient between the concatenated data to the background network to construct the perturbation sample network. The omics data, Pearson correlation coefficient and standard deviation of the specific phenomenon are consistent with those in S2.

[0066] S4 calculates the network information entropy and the difference network step: using the reference sample network and the perturbation sample network obtained in S2 and S3, respectively calculate the network information entropy of the reference sample network and the perturbation sample network, take the absolute value of the difference between the two multiplied by the difference in standard deviation, and obtain the difference network information entropy.

[0067] The specific calculation formula of difference network information entropy is as follows:

[0068] ΔH E (s+1)=|SD E(s+1)-SD E (s)|·|H E (s+1)-H E (s)|

[0069] The calculation formula for the standard deviation of a specific edge is as follows:

[0070]

[0071] The calculation formula of the network information entropy of a specific edge is as follows:

[0072]

[0073] In the formula and is the node standard deviation of the two connected nodes g1 and g2 corresponding to a specific edge E, and is the node network information entropy of the two connected nodes g1 and g2 corresponding to a specific edge E, and α and β represent the weights of the two nodes respectively.

[0074] The specific node weights α and β may be values ​​including but not limited to protein molecular weight, protein length, arithmetic mean, etc.

[0075] The calculation formula of specific node network information entropy is as follows:

[0076]

[0077] In the formula, x and y represent two nodes respectively, s represents the number of samples, M represents the number of first-order neighbor nodes y corresponding to node x, and p x，y (s) represents the correlation probability between the values ​​of the two nodes x and y.

[0078] The calculation formula for the correlation probability between specific connected nodes is as follows:

[0079]

[0080] Where x and β represent two connected nodes, s represents the number of samples, and M represents the number of first-order neighbor nodes y corresponding to node x.

[0081] S5 Global Difference Network Information Entropy

[0082] The global difference network information entropy is calculated using the difference network information entropy, and the difference network information entropy values ​​in step S4 are sorted from high to low, and the edges of the difference network information entropy ranked in the top 5% are selected as the key local difference network information entropy to predict the key genes. Specifically, the global difference network information entropy is used to roughly determine the fluctuation of the individual at the overall level. According to the fluctuation at the overall level, one or more moments with higher global difference network information entropy values ​​can be found. According to the edges corresponding to the moments with higher values, some edges are selected as key local difference networks by a certain method, so as to determine the key local difference network information entropy. The screening method of the key local difference network includes but is not limited to screening edges with higher difference network information entropy according to a certain ratio, screening edges with higher difference network information entropy according to a certain number, and screening edges with the highest degree by using a certain difference network information entropy as a threshold.

[0083] The specific calculation formula of global difference network information entropy is as follows:

[0084]

[0085] Where ΔH e (s+1) is the difference network information entropy of a single edge e, and WG is the global network, that is, the network composed of all edges.

[0086] The calculation formula of the specific key local difference network information entropy is as follows:

[0087]

[0088] Where ΔH e (s+1) is the difference network information entropy of a single edge e, and WL is the key local network, that is, the network composed of edges selected by a certain method.

[0089] For the initial specific phenomic data with multiple perturbation samples, repeat the above steps to obtain the volatility quantification values ​​and key local difference networks corresponding to all perturbation samples. Then, according to the frequency of occurrence of each edge in the key local difference network of multiple perturbation samples, high-frequency edges are screened out, and the key local difference network that can cover the characteristics of most perturbation samples is obtained, so as to achieve a more extensive characterization of the volatility and disease warning of complex biological systems.

[0090] The key gene prediction method based on network information entropy of the present invention can more comprehensively evaluate and monitor the health status of individuals, is conducive to the early warning of specific genes and the identification of key pathogenic gene groups, and provides a reference for medical workers to conduct targeted treatment strategy analysis at the gene network level, especially in the early stages of the disease, including refined critical point detection and individual-specific health assessment, aiming to make up for the shortcomings of traditional biomarkers in terms of dynamics and individualized treatment needs, so as to promote the development of personalized medicine and better prevent and treat diseases.

[0091] Although the above methods are illustrated and described as a series of actions for simplicity of explanation, it should be understood and appreciated that these methods are not limited by the order of the actions, because according to one or more embodiments, some actions may occur in a different order and / or concurrently with other actions from those illustrated and described herein or not illustrated and described herein but understandable to those skilled in the art.

[0092] Those skilled in the art will further appreciate that the various illustrative logic blocks, modules, circuits, and algorithm steps described in conjunction with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or a combination of the two. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, circuits, and steps are generally described above in terms of their functionality. Whether such functionality is implemented as hardware or software depends on the specific application and the design constraints imposed on the overall system. The technician may implement the described functionality in different ways for each specific application, but such implementation decisions should not be interpreted as resulting in a departure from the scope of the present invention.

[0093] The various illustrative logic blocks, modules, and circuits described in conjunction with the embodiments disclosed herein may be implemented or executed with a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. A general-purpose processor may be a microprocessor, but in an alternative, the processor may be any conventional processor, controller, microcontroller, or state machine. The processor may also be implemented as a combination of computing devices, such as a combination of a DSP and a microprocessor, a plurality of microprocessors, one or more microprocessors cooperating with a DSP core, or any other such configuration. The above specific embodiments are used to explain the present invention, and are only preferred embodiments of the present invention, rather than limiting the present invention. Any modifications, equivalent substitutions, improvements, etc. made to the present invention within the spirit of the present invention and the protection scope of the claims fall within the protection scope of the present invention.

Claims

1. A key gene prediction method based on network information entropy, characterized by: The steps include: S1, constructing a background network: using specific relationship data as input data of the background network, and screening the relationship data according to a given threshold to construct the background network of the given threshold; S2, constructing the reference sample network: using the omics data of a specific representation as the initial input data of the reference sample, calculating its correlation coefficient and standard deviation, and applying the correlation coefficient to the background network to construct the reference sample network; S3, constructing a perturbation sample network: using the omics data of a specific representation as the initial input data of the perturbation sample, concatenating the initial input data of a single perturbation sample with the initial input data of the reference sample in the sample dimension, calculating the correlation coefficient and standard deviation between the concatenated data, and applying the correlation coefficient between the concatenated data to the background network to construct the perturbation sample network; S4, calculate network information entropy and difference network information entropy: use the reference sample network and the perturbation sample network to calculate their network information entropy respectively, take the absolute value of the difference between the two multiplied by the difference of the standard deviation, and get the difference network information entropy; S5, using the difference network information entropy to calculate the global difference network information entropy, sorting the difference network information entropy values ​​in step S4 from high to low, and selecting the edge of the difference network information entropy ranked in the top 5% as the key local difference network information entropy to predict the key genes; Among them, in step S4, the calculation formula of the difference network information entropy is as follows: ΔH E (s+1)=|SD E (s+1)-SD E (s)|·|H E (s+1)-H E (s)| Where H E (s) is the network information entropy of a certain edge E in the case of s samples, SD E (s) is the standard deviation of a certain edge E in the case of s samples, (s+1) represents the case where a perturbation sample is added to the reference sample network, and the edge refers to the adjacent nodes in the network; The standard deviation of an edge is calculated as follows: The calculation formula of the network information entropy of the edge is as follows: In the formula and is the node standard deviation of the two connected nodes g1 and g2 corresponding to a specific edge E, and is the node network information entropy of the two connected nodes g1 and g2 corresponding to a specific edge E, α and β represent the weights of the two nodes respectively; The node weights α and β include but are not limited to protein molecular weight, protein length, node in-degree and out-degree, and arithmetic mean values; The calculation formula of node network information entropy is as follows: In the formula, x and y represent two connected nodes, s represents the number of samples, M represents the number of first-order neighbor nodes y corresponding to node x, and p x，y (s) represents the correlation probability of the values ​​between two connected nodes x and y, and its calculation formula is as follows: Where x and y represent two connected nodes, s represents the number of samples, and M represents the number of first-order neighbor nodes y corresponding to node x. The calculation formula of global difference network information entropy is as follows: Where ΔH e (s+1) is the difference network information entropy of a single edge e, WG is the global network, that is, the network composed of all edges, wherein the calculation formula of the key local difference network information entropy is as follows: Where ΔH e (s+1) is the difference network information entropy of a single edge e, and WL is the key local network, that is, the network composed of edges selected by a certain method.

2. The key gene prediction method based on network information entropy according to claim 1, characterized in that: In step S1, the specific relationship data includes data in protein interaction data, biological pathway data, and molecular interaction data.

3. The key gene prediction method based on network information entropy according to claim 1, characterized in that: In step S2, the omics data of the specific phenomena include multiple omics data such as human peripheral blood, tumor slice tissue of cancer patients, genome, transcriptome, metabolome, and proteome of experimental mice, as well as food chain data and disease transmission model data.

4. The key gene prediction method based on network information entropy according to claim 1 is characterized in that: The correlation coefficients in step S2 and step S3 include Pearson correlation coefficient, Spearman correlation coefficient, mutual information and MIC maximum information coefficient.

5. The key gene prediction method based on network information entropy according to claim 4 is characterized in that: The Pearson correlation coefficient measures the linear relationship between the initial input data, and the background network is applied to the relationship matrix generated by the correlation coefficient to obtain the reference sample network. The calculation formula of Pearson correlation coefficient is as follows: In the formula, x and y represent two nodes respectively, X n and Y n Represent the values ​​of the two nodes on the nth sample, and They represent the arithmetic mean of the values ​​of the two nodes in all samples, s represents the number of samples, That is, the Pearson correlation coefficient is used to measure the strength of the correlation between two variables. When using genomic data to construct a correlation network, the network nodes represent genes and the edges represent the strength of the association between genes. The standard deviation is calculated as follows: Where x represents a specific node, X n Represents the value of the node on the nth sample, Represents the arithmetic mean of the values ​​of the node in all samples, and s represents the number of samples.

6. The key gene prediction method based on network information entropy according to claim 1 is characterized in that: The screening methods of key local difference networks include but are not limited to screening edges with higher difference network information entropy according to a certain ratio, screening edges with higher difference network information entropy according to a certain number, and using a certain difference network information entropy as a threshold to screen edges with the highest degree.

7. The key gene prediction method based on network information entropy according to claim 1 is characterized in that: The initial input data of the reference samples include omics data from healthy humans, paracancerous tissues, and specific phenotypes of the control group; The initial input data for the perturbation samples include omics data from diseased humans, cancer tissues, and specific phenotypes of experimental groups.

Images