A biological network construction algorithm based on different comparison modes between molecular characteristics

Abstract

The present application relates to a kind of biological network construction algorithm based on different comparison mode between molecular characteristics, including the input hepatocarcinoma's omics data, the frequency and frequency of each case of arbitrary two molecular characteristics in adjacent two class samples under horizontal comparison relationship and vertical comparison relationship, and calculate its contained complexity and maximum complexity, calculate molecular feature pair in adjacent two class samples under horizontal comparison relationship score S h (i,j) and vertical comparison relationship score S v (i,j), Score (i,j)=S h (i,j) and greater than or equal to threshold value, the edge of two nodes f i ,f j In network diagram is expressed in a color, Score (i,j)=S v (i,j) and greater than or equal to threshold value, the edge of two nodes f i ,f j In network diagram is expressed in another color, biological network is G=(V (G),E (G),W (G)); Effectively eliminate the influence of individual difference on data analysis, make the biological network more real and effective reflect the change of molecular correlation, improve the clinical early diagnosis effect of hepatocarcinoma.

Description

Technical Field

[0001] This invention relates to the field of biological data analysis technology, and in particular to a biological network construction algorithm based on different comparison methods between molecular features. Background Technology

[0002] Liver cancer seriously threatens human health and quality of life, imposing a heavy economic burden on society and families. Clinical research on liver cancer has become a hot topic in the field of bioinformatics, especially the screening of indicators for early warning and personalized treatment, which is crucial for improving patient survival rates. However, the development of liver cancer involves complex interactions among multiple factors, including genetics, viruses, and the environment, and its pathogenic mechanism is not yet fully understood. Therefore, liver cancer faces clinical challenges such as difficulty in early diagnosis and poor treatment outcomes. Furthermore, the body's metabolism is influenced by internal genetic factors and external environmental factors, resulting in diverse metabolic phenotypes. These individual differences directly affect the clinical efficacy of early diagnosis and drug treatment for liver cancer. With the development of modern science and technology, early and accurate diagnosis of liver cancer based on individual patient characteristics and combined with multi-omics bioinformatics, such as genomics and metabolomics, has become a key focus in translational medicine research. Multi-omics data analysis methods can comprehensively analyze the pathogenic mechanism of liver cancer, discover novel and highly effective early warning biomarkers integrating gene and metabolic processes, and improve cancer prognosis and postoperative outcomes through early diagnosis and intervention.

[0003] In systems biology, molecules at different levels, such as the genome and metabolome, are interconnected and interact to ensure the normal functioning of various physiological activities in the body. Organically integrating genomics and metabolomics, and fully utilizing the biological information at different levels within the organism, can more comprehensively reveal the essence of life science problems and promote solutions to clinical challenges in liver cancer. Because molecules at different levels are closely linked and work together to maintain a robust biochemical reaction environment within cells, living organisms typically possess strong robustness. This means that the organism can maintain its functions even under certain external disturbances or internal noise, which is crucial for the survival and development of most organisms. When a living organism encounters severe stimuli, the robust mechanisms of its internal systems are often disrupted, leading to the occurrence and even worsening of liver cancer. Systematically exploring the differences in biochemical reaction activities and robust mechanisms among different omics molecules under disease states helps to further understand the pathogenesis of liver cancer, promotes the effective screening of clinical biomarkers, and thus improves the early clinical diagnosis of liver cancer.

[0004] Changes in lipid metabolism influence many biochemical processes, such as tumor proliferation, division, and metastasis, playing a crucial role in the development of carcinoma. Unlike healthy cells that use exogenous fatty acids to meet their normal proliferative needs, tumor cells utilize endogenous lipid synthesis to satisfy their uncontrolled proliferative activities. Most lipids and lipoproteins are synthesized in the liver, a vital organ for energy metabolism. The concentrations of these key lipids and lipoproteins, such as fatty acid synthases, ATP citrate lyase, and stearoyl-CoA desaturase, are dynamically changing during the development of liver cancer and can serve as clinical indicators for different stages of liver cancer. Therefore, exploring abnormal changes in lipid metabolism during liver cancer development helps to deepen our understanding of the progression of liver cancer, discover more effective early warning information, and improve the effectiveness of early clinical diagnosis of liver cancer.

[0005] In living organisms, intermolecular relationships are complex and diverse, encompassing both linear and nonlinear relationships. Different types of relationships contribute differently to research on the occurrence and development of liver cancer under various circumstances. Considering the inherent advantages of different types of intermolecular relationships, this invention explores algorithms for constructing biological networks based on different types of molecular feature relationships. This eliminates the impact of differences in the magnitude of the distinguishing power of different types of relationships on data analysis. An evaluation criterion is developed that can be used to measure both linear and nonlinear relationships among molecular features. The biological network constructed based on this evaluation criterion can accurately characterize changes in different types of molecular feature relationships during the occurrence and development of liver cancer. Key network nodes with strong predictive capabilities can be screened as early diagnostic biomarkers for liver cancer, thereby improving the clinical early diagnosis of liver cancer. Summary of the Invention

[0006] This invention provides a biological network construction algorithm based on different comparison methods between molecular features. It explores the changes in intermolecular relationships during the development of liver cancer from both linear and nonlinear perspectives, effectively eliminating the influence of individual differences on data analysis. This allows the constructed biological network to more realistically and effectively reflect changes in intermolecular relationships, accurately characterize the changes in different types of relationships between molecular features during the development of liver cancer, and screen key network nodes with strong predictive capabilities as early diagnostic biomarkers for liver cancer, thereby improving the clinical early diagnosis effect of liver cancer.

[0007] To achieve the above objectives, the present invention employs the following technical solution:

[0008] A biological network construction algorithm based on different comparison methods between molecular features includes the following steps:

[0009] 1) Input omics data for different stages of liver cancer development;

[0010] 2) Any pair of molecular characteristics (f) in the statistical data i ,f j The frequency and relative occurrence of each case under horizontal and vertical comparison relationships between two adjacent classes of samples;

[0011] 3) Calculate molecular feature pairs (f i ,f j The information of difference contained in horizontal and vertical comparison relationships is presented using molecular feature pairs (f i ,f j Complexity Comp ij To measure;

[0012] 4) Calculate molecular feature pairs (f i ,f j The maximum complexity of Comp ij,max The maximum complexity is achieved by dividing the frequency of the d-th event (1≤d≤r) in each class of samples by the value of r, thus eliminating the influence caused by different values ​​of r, where r is the molecular feature pair (f i ,f j The number of events in each type of sample;

[0013] 5) Calculate the molecular feature pairs (f) respectively. i ,f j The score S represents the level comparison relationship between two adjacent classes of samples. h (i,j) and vertical comparison score S v (i,j), collectively referred to as S(i,j), determine the comparison relationship with stronger discriminative power, and obtain the molecular feature pairs (f i ,f j The final score of ) is Score(i,j) = max{S h (i,j),S v (i,j)};

[0014] 6) If Score(i,j) = S h If (i,j) is greater than or equal to the threshold ε, then two nodes f in the network graph... i ,f j The edges connecting points are represented by a single color; conversely, if Score(i,j) = S... v If (i,j) is greater than or equal to the threshold ε, then two nodes f in the network graph i ,f j The edges are represented by a different color;

[0015] 7) Constructing a biological network: Construct a biological network G = (V(G), E(G), W(G)), i.e., a weighted undirected network, where V(G) = F is the set of nodes, and E(G) = {(f i ,f j )|f i ,f j Let F, Score(i,j)≥ε be the set of edges, and W(G) be the set of edge weights, where W(G)={w(f i ,f j )=Score(i,j)|(f i ,f j )∈E(G)},w(f i ,f j ) represents node f in the biological network graph G. i With node f j The weight value of the connecting edge, if w(f i ,f j ) = S h If (i,j)≥ε, then node f i and node f j The edges between them are represented by a single color; if w(f i ,f j ) = S v If (i,j)≥ε, then node f i and node f j The edges connecting them are represented by a different color.

[0016] Furthermore, the calculation formula for step 3) is as follows:

[0017]

[0018] Among them, Comp ij For molecular characteristic pairs (f i ,f j The complexity values ​​under horizontal and vertical comparison relationships. For molecular characteristic pairs (f i ,f j The frequency of the d-th event in class c1 samples. For molecular characteristic pairs (f i ,f j The frequency of the d-th event in class c2 samples. Let be the frequency of the d-th event in all samples. For molecular characteristic pairs (f i ,f j The frequency of the d-th event in class c1 samples. For molecular characteristic pairs (f i ,fj The frequency of the d-th event in the c2-class samples.

[0019] Furthermore, step 4) involves molecular feature pairs (f) i ,f j The maximum complexity of Comp ij,max The calculation formula is as follows:

[0020]

[0021] Where, fre ij (c1) represents the molecular characteristic pair (f) i ,f j The frequency of all events occurring in class c1 samples, fre ij (c2) represents the molecular characteristic pair (f) i ,f j ) is the frequency of all events in class c2 samples, and num is the total number of samples.

[0022] Furthermore, the molecular characteristics are related to (f) i ,f j The score S(i,j) under horizontal and vertical comparison relationships is calculated using the following formula:

[0023] S(i,j)=(Comp ij,max -Comp ij ) / Comp ij,max (3)

[0024] Furthermore, it also includes network analysis methods. Using network graph topology analysis, the star-shaped subgraph formed by the node with the highest degree and its directly connected nodes in the constructed biological network graph is extracted as an important network signal, and the node with the highest degree is selected as a potential early warning marker.

[0025] Compared with the prior art, the beneficial effects of the present invention are:

[0026] 1) This invention comprehensively considers a pair of molecular characteristics (f) in the same sample from both horizontal and vertical comparison methods. i ,f j The magnitude relationship between the values ​​is expressed, and the molecular feature pairs (f) are statistically analyzed. i ,f jThe frequency and intensity of each situation under different comparison methods were analyzed, thereby effectively eliminating the influence of individual differences on data analysis. This made the constructed biological network more realistic and effective in reflecting changes in intermolecular relationships. Based on the constructed dynamic biological network, the topological structure analysis method was used to screen key biological information located at pathway response hubs as potential biomarkers for early clinical diagnosis of liver cancer, in order to improve the early diagnosis effect of liver cancer and reduce patient mortality.

[0027] 2) In view of the complex and diverse relationships between different omics molecules in living organisms, this invention uses horizontal and vertical comparison methods to deeply analyze the dynamic changes of the relationships between different omics molecules from both linear and nonlinear perspectives to construct a biological network. This allows for a more comprehensive characterization of the changes in related biochemical reactions during the development of liver cancer, providing assistance for a comprehensive and precise study of the pathogenesis of liver cancer and providing an important targeted research objective for the early diagnosis of liver cancer. Attached Figure Description

[0028] Figure 1 This is a gene network subgraph SG of an organism as it progresses from a healthy stage to stage I liver cancer in an embodiment of the present invention. I The topology.

[0029] Figure 2 This is a gene network subgraph SG of an organism progressing from stage I to stage II liver cancer in an embodiment of the present invention. I The topology.

[0030] Figure 3 This is a gene network subgraph SG of an organism progressing from stage II to stage III liver cancer in an embodiment of the present invention. I The topology.

[0031] Figure 4 This is a gene network subgraph SG of an organism progressing from stage II to stage IV liver cancer in an embodiment of the present invention. I The topology.

[0032] Figure 5 This is the topology of the metabolic network subgraph SDN4 in this embodiment of the invention, when liver cancer progresses from time point T1 to time point T2.

[0033] Figure 6 This is the topology of the metabolic network subgraph SDN4 in this embodiment of the invention, when liver cancer progresses from time point T2 to time point T3.

[0034] Figure 7 This is the topology of the metabolic network subgraph SDN4 in this embodiment of the invention, when liver cancer progresses from time point T3 to time point T4.

[0035] Figure 8 This is the topology of the metabolic network subgraph SDN4 in this embodiment of the invention, when liver cancer progresses from time point T4 to time point T5.

[0036] Figure 9 This is the topology of the metabolic network subgraph SDN4 in this embodiment of the invention, when liver cancer progresses from time point T5 to time point T6.

[0037] Figure 10 This is the topology of the metabolic network subgraph SDN4 in this embodiment of the invention, when liver cancer progresses from time point T6 to time point T7. Detailed Implementation

[0038] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings:

[0039] This invention presents a biological network construction algorithm based on different comparison methods between molecular features. Using hepatocellular carcinoma (HCC) omics data as the research object, this invention aims to discover potential early warning biomarkers for HCC clinically. It employs biological network construction and analysis as research methods to explore the mechanisms of HCC development and progression, thereby contributing to the research and application of early clinical diagnosis of HCC. In living organisms, the relationships between different omics molecules are complex and diverse. Different types of relationships contribute differently to clinical research on HCC under various circumstances. Systematically and comprehensively exploring the differential changes in pathway responses between different omics molecules during the development of HCC helps to determine early warning information for HCC, thereby improving the effectiveness of early clinical diagnosis. This invention proposes a biological network construction method, MPELNV, based on different comparison methods between molecular features. It utilizes in-depth comparisons of horizontal and vertical relationships to analyze the differential changes in linear and nonlinear relationships between molecules during the process of carcinogenesis in the body, screening network information with strong predictive capabilities as early diagnostic biomarkers for HCC.

[0040] Let F = {f1, f2, ..., f m} is defined as the set of molecular features, where m represents the number of molecular features; X = {x1, x2, ..., x...} n} is defined as a sample set, where n represents the number of samples; C = {c1, c2, ..., c z} is defined as a set of class labels, where z represents the number of class labels.

[0041] 1) Input omics data of different stages of liver cancer development.

[0042] 2) Any pair of molecular characteristics (f) in the statistical data i ,f j The frequency and relative frequency of each case under horizontal and vertical comparison relationships between two adjacent classes of samples.

[0043] 3) To avoid loss of generality, assume that under different comparison relationships, a pair of molecular characteristics (f i ,f j In each class of samples, there are r events. For example, r = 2 for horizontal comparisons and r = 4 for vertical comparisons. This invention uses a complexity level of Comp. ij Calculate a pair of molecular characteristics (f i ,f j The difference information contained under different comparison methods is calculated as shown in formula (4):

[0044]

[0045] Among them, Comp ij For molecular characteristic pairs (f i ,f j Complexity values ​​under horizontal and vertical comparison relationships. For molecular characteristic pairs (f i ,f j The frequency of the d-th event in class c1 samples. For molecular characteristic pairs (f i ,f j The frequency of the d-th event in class c2 samples. Let be the frequency of the d-th event in all samples. For molecular characteristic pairs (f i ,f j The frequency of the d-th event in class c1 samples. For molecular characteristic pairs (f i ,f j The frequency of the d-th event in the c2-class samples.

[0046] 4) From the perspective of horizontal comparison relationship analysis, i.e. linear association, there are two cases for each class of samples: f i ≥f j and f i <f j See Table 1.

[0047] Table 1 Molecular characteristic pairs (f i ,f j ) Horizontal comparison method

[0048]

[0049] From the perspective of vertical comparison relationship analysis, i.e., non-linear correlation relationship, there are 4 cases in each class of samples: f i ≥μ i &f j ≥μ j f i≥μ i &f j <μ j f i <μ i &f j ≥μ j and f i <μ i &f j <μ j As shown in Table 2,

[0050] Table 2 Molecular characteristic pairs (f i ,f j Vertical comparison method

[0051]

[0052] Where fre represents the frequency of the event, and μ represents the mean of the molecular feature f across all samples;

[0053] Since the r value is not the same under different comparison relationships, a pair of molecular characteristics (f i ,f j ) of Comp ij The values ​​differ by magnitude under different comparison relationships. This invention uses the maximum complexity method to solve this problem, as shown in Table 3. The maximum complexity method divides the frequency (fre) of the d-th event (1≤d≤r) in each class of samples by the value of r, thereby eliminating the influence caused by different values ​​of r. Molecular feature pairs (f i ,f j The maximum complexity of Comp ij,max The calculation method is shown in formula (5):

[0054]

[0055] Where, fre ij (c1) represents the molecular characteristic pair (f) i ,f j The frequency of all events occurring in class c1 samples, fre ij (c2) represents the molecular characteristic pair (f) i ,f j The frequency of all events in class c2 samples, where num is the total number of samples.

[0056] Table 3: Statistics on Maximum Complexity

[0057]

[0058] 5) Calculate molecular feature pairs (f i ,f jThe score S(i,j) under horizontal and vertical comparison relationships is calculated using the following formula:

[0059] S(i,j)=(Comp ij,max -Comp ij ) / Comp ij,max (6)

[0060] The level comparison score S between the two adjacent samples was calculated using formulas (4), (5), and (6). h (i,j) and vertical comparison score S v (i,j), determine the more discriminative comparative relationship, and obtain the molecular feature pairs (f i ,f j The final score of ) is Score(i,j) = max{S h (i,j),S v (i,j)}.

[0061] 6) If Score(i,j) = S h If (i,j) is greater than or equal to the threshold ε, then two nodes f in the network graph... i ,f j The edges connecting points are represented by a single color; conversely, if Score(i,j) = S... v If (i,j) is greater than or equal to the threshold ε, then two nodes f in the network graph... i ,f j The edges are represented by a different color.

[0062] 7) Constructing a biological network: Construct a biological network G = (V(G), E(G), W(G)), i.e., a weighted undirected network, where V(G) = F is the set of nodes, and E(G) = {(f i ,f j )|f i ,f j Let F, Score(i,j)≥ε be the set of edges, and W(G) be the set of edge weights, where W(G)={w(f i ,f j )=Score(i,j)|(f i ,f j )∈E(G)},w(f i ,f j ) represents node f in the biological network graph G. i With node f j The weight value of the connecting edge, if w(f i ,f j ) = S h If (i,j)≥ε, then node f i and node f jThe edges between them are represented by a single color; if w(f i ,f j ) = S v If (i,j)≥ε, then node f i and node f j The edges connecting them are represented by a different color.

[0063] The constructed network can realistically and comprehensively characterize the differences in pathway response activities of the body under different states. Topological analysis of the constructed biological network can reveal important network signals closely related to the research question. Based on the constructed biological network, it can be used to screen network early warning information and potential biomarkers for liver cancer, which can help in the early diagnosis of liver cancer patients. It is also suitable for fields such as disease omics data analysis and translational medicine.

[0064] In the constructed biological network diagram, the node with the highest degree indicates that the molecule participates in multiple pathway reactions during the development and progression of liver cancer, occupying a central and pivotal position in biochemical reactions and playing a crucial biological role in the carcinogenesis of living organisms. The star-shaped subgraph formed by the node with the highest degree and its directly connected nodes serves as an important network signal. Changes in its network topology can intuitively reflect alterations in the intermolecular relationships during liver cancer lesions, thus contributing to the research on the pathogenesis of liver cancer. This invention screens the node with the highest degree as a potential early warning biomarker, aiming to effectively improve the early clinical diagnosis of liver cancer.

[0065] The following embodiments are implemented based on the technical solution of the present invention, providing detailed implementation methods and specific operation processes. However, the scope of protection of the present invention is not limited to the following embodiments. Unless otherwise specified, the methods used in the following embodiments are conventional methods.

[0066]

Example

[0067] Screening for early diagnostic biomarkers for liver cancer based on genomic and metabolomics data related to lipid metabolism mechanisms.

[0068] (1) Collection of genomic data related to lipid metabolism mechanisms

[0069] The training set of liver cancer genomics data in this experiment was obtained from the Cancer Genome Atlas database, containing 50 control group samples (N) and 347 model group samples (M). The model group consisted of 171 liver cancer stage I samples, 86 liver cancer stage II samples, 85 liver cancer stage III samples, and 5 liver cancer stage IV samples. According to functional enrichment analysis, a total of 533 genes were closely related to lipid metabolism activities. Therefore, these 533 genes were used for subsequent network construction and analysis. The validation set of liver cancer genomics data in this experiment was obtained from the Gene Expression Comprehensive Database, consisting of datasets GSE101685, GSE62232, GSE121248, and GSE174570, and was used to verify the effectiveness of potential early diagnostic biomarkers for liver cancer selected based on the training set.

[0070] (2) Collection of metabolomics data related to lipid metabolism mechanisms

[0071] The training set of liver cancer metabolomics data in this experiment consisted of 10 control rats and 7 model rats, containing 7 time points, namely T1, T2...T7, with a total of 119 samples. Time point T1 represents the typical hepatitis stage (H), time points T2 to T4 represent the cirrhosis stage (CIR), and time points T5 to T7 represent the liver cancer stage (HCC), with time point T4 representing the typical cirrhosis stage and time point T5 representing the early stage of liver cancer. The training set contained 338 lipid metabolites. In addition, the validation set of liver cancer metabolomics data in this experiment consisted of another 6 model rats, containing 6 time points and a total of 36 serum samples.

[0072] (3) Construction of dynamic biological networks based on liver cancer genomics data

[0073] In the liver cancer genomics data, the training set model includes control group samples, liver cancer stage I samples, liver cancer stage II samples, liver cancer stage III samples, and liver cancer stage IV samples, based on molecular feature pairs (f i ,f j A dynamic network is constructed based on the Score(i,j) values ​​between two adjacent classes of samples. If Score(i,j) = S h If (i,j) is greater than or equal to 0.6, then the two corresponding molecular features are connected by a green border; if Score(i,j) = S v If (i,j) is greater than or equal to 0.6, then the two corresponding molecular features are connected by a red edge, and four network graphs are output.

[0074] (4) For the dynamic network of liver cancer genes, G I A biological network constructed based on control group samples and stage I liver cancer samples can be used more effectively to discover prospective early warning gene information for liver cancer and extract G... IThe star subgraph SG formed by the 7 nodes with the highest moderate density and their directly connected nodes. I As a predictive network for liver cancer early warning signals, the seven nodes with the highest degree were selected as potential early diagnostic gene markers for clinical liver cancer.

[0075] Figure 1-4 SG, a sub-map of the gene network of a living organism at different stages of liver cancer. I The topology.

[0076] (5) Construction of dynamic biological networks based on liver cancer metabolomics data

[0077] In the liver cancer metabolomics data, the training set model includes samples from 7 time points to characterize the occurrence and development of liver cancer, based on molecular feature pairs (fi, f). j A dynamic network is constructed using the Score(i,j) values ​​at two adjacent time points. If Score(i,j) = S... h If (i,j) is greater than or equal to 0.85, then the two corresponding molecular features are connected by a green border; if Score(i,j) = S v If (i,j) is greater than or equal to 0.85, then the two corresponding molecular features are connected by a red edge, and 6 network graphs are output.

[0078] (6) For the metabolic dynamic network of liver cancer, DN4 is a biological network constructed based on T4 typical liver cirrhosis samples and T5 early stage liver cancer samples. It can be used more effectively to discover prospective early warning metabolic information of liver cancer. The star subgraph SDN4 composed of the 7 nodes with the highest degree in DN4 and the nodes directly connected to them is extracted as the early warning signal of liver cancer prospective network, and the 7 nodes with the highest degree are used as potential early diagnostic metabolic markers of clinical liver cancer.

[0079] Figure 5-10 The topological changes of the metabolic network subgraph SDN4 at different stages of liver cancer.

[0080] (7) The area under the curve was used to verify the early diagnostic effect of the potential gene markers and metabolic markers for clinical liver cancer screened by the present invention. Tables 4, 5, 6 and 7 show the comparison results of the present invention with other methods on genomics and metabolomics data. The comparative experiments show that the potential markers screened by the present invention have the strongest discrimination ability on gene, metabolomics training set and test set.

[0081] Table 4 Comparison results of different methods on the genomics data training set

[0082]

[0083] Table 5 Comparison results of different methods on the validation set of genomics data.

[0084]

[0085]

[0086] Table 6 Comparison results of different methods on the metabolomics data training set

[0087]

[0088] Table 7 Comparison results of different methods on the metabolomics data validation set.

[0089]

Claims

1. A biological network construction algorithm based on different comparison methods between molecular features, characterized in that, Includes the following steps: 1) Input omics data for different stages of liver cancer development; 2) Any pair of molecular characteristics (f) in the statistical data i , f j The frequency and relative occurrence of each case under horizontal and vertical comparison relationships between two adjacent classes of samples; 3) Calculate molecular feature pairs (f i , f j The information of difference contained in horizontal and vertical comparison relationships is presented using molecular feature pairs (f i , f j Complexity Comp ij To measure; 4) Calculate molecular feature pairs (f i , f j The maximum complexity of Comp ij,max The maximum complexity is achieved by dividing the frequency of the d-th event (1 ≤ d ≤ r) in each class of samples by the value r, thus eliminating the influence caused by different r values, where r is the molecular feature pair (f i , f j The number of events in each type of sample; 5) Calculate the molecular feature pairs (f) respectively. i , f j The score S represents the level comparison relationship between two adjacent classes of samples. h (i, j) and vertical comparison score S v (i, j), collectively referred to as S(i, j), determine the comparison relationship with stronger discriminative power, and obtain the molecular feature pairs (f i , f j The final score of ) is Score(i, j) = max{S h (i, j), S v (i, j)}; 6) If Score(i, j) = S h If (i, j) are greater than or equal to the threshold ε, then the two nodes f in the network graph are... i , f j The edges are represented by a color if Score(i, j) = S v If (i, j) are greater than or equal to the threshold ε, then the two nodes f in the network graph are... i ,f j The edges are represented by a different color; 7) Constructing a biological network: Construct a biological network G = (V(G), E(G), W(G)), i.e., a weighted undirected network, where V(G) = F is the set of nodes, and E(G) = {(f i , f j ) | f i , f j Let F ∈ Score(i, j) ≥ ε be the set of edges, and W(G) be the set of edge weights, where W(G) = {w(f i , f j = Score(i, j) | (f i , f j ) ∈ E(G)},w(f i , f j ) represents node f in the biological network graph G. i With node f j The weight value of the connecting edge, if w(f i , f j ) = S h If (i, j) ≥ ε, then node f i and node f j The edges between them are represented by a single color; if w(f i , f j ) = S v If (i, j) ≥ ε, then node f i and node f j The edges connecting them are represented by a different color.

2. The biological network construction algorithm based on different comparison methods between molecular features according to claim 1, characterized in that, The calculation formula for step 3) is as follows: （1） Among them, Comp ij For molecular characteristic pairs (f i , f j The complexity value under horizontal and vertical comparison relationships, fre ( ) is a molecular characteristic pair (f i , f j )exist The frequency of the d-th event in the class sample, fre ( ) is a molecular characteristic pair (f i , f j )exist The frequency of the d-th event in the class sample, fre Let be the frequency of the d-th event in all samples. For molecular characteristic pairs (f i , f j )exist The frequency of the d-th event in the class sample For molecular characteristic pairs (f i , f j )exist The frequency of the d-th event occurring in class samples; Step 4) Molecular feature pairs (f i , f j The maximum complexity of Comp ij,max The calculation formula is as follows: （2） Where, fre ij ( ) is a molecular characteristic pair (f i , f j )exist The frequency of all events occurring in the class sample, fre ij ( ) is a molecular characteristic pair (f i , f j )exist The frequency of all events occurring in a class of samples, where num is the total number of samples; The molecular feature pair (f i , f j The score S(i, j) under horizontal and vertical comparison relationships is calculated using the following formula: S(i, j) = (Comp ij,max - Comp ij ) / Comp ij,max (3)。 3. The biological network construction algorithm based on different comparison methods between molecular features according to claim 1, characterized in that, It also includes network analysis methods, using network graph topology analysis to extract the star-shaped subgraph formed by the node with the highest degree and its directly connected nodes in the constructed biological network graph as important network signals, and to screen the node with the highest degree as a potential early warning marker.

Images