CN107887023A - A kind of microbial diseases Relationship Prediction method based on similitude and double random walks - Google Patents

Abstract

The invention discloses a kind of microbial diseases Relationship Prediction method based on similitude and double random walks, function of diseases similitude is built first with disease gene relation and gene function affinity information；Disease Gaussian kernel similitude is built further according to known microbial diseases relation；The final similitude of disease is integrated on the basis of function of diseases similitude and Gaussian kernel similitude.The similitude of microorganism derives from the Gaussian kernel similitude of known microorganisms disease relationship.Again by microorganism affinity information, disease affinity information, it is known that microbial diseases relation information be integrated into a double-deck heterogeneous network.Microbial diseases Relationship Prediction is carried out in heterogeneous network using double random walk methods.The present invention can be predicted to the relation between microbial diseases, and basic basis are provided for Bioexperiment, save its manpower and materials cost.

Description

A Microbe-Disease Relationship Prediction Method Based on Similarity and Double Random Walk

technical field

The invention belongs to the field of systems biology and relates to a microorganism-disease relationship prediction method based on similarity and double random walks.

Background technique

More and more studies have shown that microbes play a very important role in many complex human diseases. With the rapid development of the current next-generation DNA sequencing technology, the discovery of the relationship between microorganisms and diseases in the human body has been promoted, such as microbiota and various cancer diseases, cardiovascular diseases, metabolic syndrome (such as obesity and diabetes), central nervous system Nervous system diseases and autoinflammatory diseases, etc. These studies not only contribute to the understanding of the disease mechanism, but also contribute to the development of new treatment and diagnosis schemes for the disease. For example, fecal microbiota transplantation has been identified as a safe and effective treatment option for Clostridium infection by reintroducing normal flora into the donor stool, correcting imbalances and re-establishing normal bowel function. Therefore, a systematic understanding of the relationship between organisms and diseases is becoming more and more urgent.

At present, it is commonly used to discover the relationship between microorganisms and diseases through conventional experiment-based methods, which are time-consuming and expensive, and are also limited by the experimental environment. For example, some bacteria cannot be cultivated in the existing planting laboratory environment. . At the same time, the prediction of the relationship between microorganisms and diseases through computational models has not been vigorously applied. So far, few methods have emerged for predicting microbe-disease relationships through computational models. The KATZHMDA method is currently the first model to predict the relationship between new microorganisms and diseases through the known relationship between microorganisms and diseases. It integrates disease representation similarity, Gaussian kernel similarity and microbial Gaussian kernel similarity, using KATZ degree information to predict novel microbe-disease relationships. Most of the current research on microbes in computing is focused on the classification of microbes, while the relationship between them and diseases is seriously insufficient. The current level of development and prediction results of computational prediction models for the relationship between microorganisms and diseases are not enough to allow biological experimenters to recognize the validity of computational model predictions, and further use them as the basis for subsequent experimental research.

Constrained by the efficiency of biological experimental verification, insufficient attention and progress in predicting the relationship between microorganisms and diseases through computational models, and the need for further improvement in the prediction results, the current systematic understanding of the relationship between microorganisms and diseases is still limited. There is an urgent need to propose a more effective prediction model, make full use of existing biological information, discover new relationships between microorganisms and diseases in a more scientific way, lay the foundation for subsequent research on their relationship prediction, and further provide important insights for biological experimental research. basic basis. In addition, with the development of current computational biology and next-generation DNA sequencing technology, the importance of microbes to diseases is becoming more and more important, which in turn puts forward an urgent need for the development of predictive models for the relationship between microbes and diseases. Therefore, in order to further confirm the importance of the relationship between microorganisms and diseases, and provide assistance for the development of subsequent relationship prediction models and biological experiment verification, it is necessary to design an effective method for predicting the relationship between microorganisms and diseases.

Contents of the invention

The technical problem to be solved by the present invention is to provide a method for predicting the relationship between microorganisms and diseases based on similarity and double random walks, which can accurately predict the relationship between microorganisms and diseases, and provide a basis for the follow-up on the relationship between microorganisms and diseases. The development of predictive models provides the basis and further effectively avoids a large amount of manpower and material resources consumed by biochemical experiments.

The technical solution of the invention is as follows:

A method for predicting the relationship between microorganisms and diseases based on similarity and double random walks, comprising the following steps:

Step 1: construct disease functional similarity matrix D _funsim , disease Gaussian kernel similarity matrix K _GIP,d and microbial Gaussian kernel similarity matrix K _GIP,m respectively;

Step 2: Integrate the disease functional similarity matrix D _funsim and the disease Gaussian kernel similarity matrix K _GIP,d to obtain the final disease similarity matrix S _d ; use the microbial Gaussian kernel similarity matrix K _GIP,m as the microbial final similarity matrix S _m ;

Step 3: According to the known microbe-disease relationship, microbe final similarity matrix S _m and disease final similarity matrix S _d , construct a two-layer heterogeneous network, and use double random walk method to associate microbe-disease pairs Score prediction; the larger the association score, the more likely the corresponding microbe-disease pair is related.

Further, in step 1, the functional similarity between two diseases is first calculated based on the known disease-gene relationship and the functional similarity of genes, and then the disease function is constructed from the functional similarity between all diseases similarity matrix D _funsim ;

For any two diseases A and B, the functional similarity calculation formula is as follows:

Among them, G _A ＝{g _A1 ,g _A2 ,...,g _Am } is the gene set associated with disease A, similarly, G _B ＝{g _B1 ,g _B2 ,... , g _Bn } is the gene set associated with disease B, m and n are the gene numbers in the gene set _GA and G _B respectively; is the functional similarity value of gene g _Ai and gene set G _B , is the functional similarity value of gene g _Bj and gene set _GA , and the corresponding calculation formula is as follows:

Among them, F(g _Ai , g _Bj ) is the semantic similarity value between genes g _Ai and g _Bj . The HumanNet database provides the calculated semantic similarity value based on the logarithmic likelihood function. The specific calculation method is as follows:

F(g _Ai ,g _Bj )＝LLS(g _Ai ,g _Bj ).

Where LLS represents the logarithmic likelihood function (in the HumanNet database, using the logarithmic likelihood function to calculate the gene semantic similarity value is the prior art).

Further, in the step 1, according to the known microorganism-disease relationship, the disease Gaussian kernel similarity matrix K _GIP,d and the microbial Gaussian kernel similarity matrix K _GIP,m are respectively constructed, and the process is as follows:

First, define is the collection of microorganisms, N _m is the number of microorganisms; is the collection of all diseases, N _d is the number of diseases; the adjacency matrix Y∈N _m ×N _d indicates whether there is a known relationship between each microorganism and the disease; if there is a known relationship between microorganism m _i and disease d _j Then the value of y _ij is 1, otherwise the value is 0;

Then, the Gaussian kernel similarity between all diseases is calculated; for any two diseases d ₁ and d ₂ , the Gaussian kernel similarity is calculated as follows:

K _GIP,d (d ₁ ,d ₂ )＝exp(-γ _d ||yd ₁ -yd ₂ || ² )

in, γ _d is the adjustment parameter to control the kernel width, γ' _d is the disease bandwidth parameter, which is set to 1 according to (Gaussian kernel use) experience;

Then calculate the Gaussian kernel similarity between all microorganisms; for any two microorganisms m ₁ and m ₂ , the calculation method of the Gaussian kernel similarity is as follows:

K _GIP,m (m ₁ ,m ₂ )＝exp(-γ _m ||ym ₁ -ym ₂ || ² ).

in, γ _m is an adjustment parameter to control the width of the nucleus, and γ' _m is a microbial bandwidth parameter, which is set to 1 according to experience;

Finally, the disease Gaussian kernel similarity matrix K _{GIP,d is} constructed from the Gaussian kernel similarity between all diseases, and the microbial Gaussian kernel similarity matrix K _GIP,m is constructed from the Gaussian kernel similarity between all microorganisms.

Further, in the step 2, the disease functional similarity matrix D _funsim and the disease Gaussian kernel similarity matrix K _GIP,d are integrated to obtain the final disease similarity matrix S _d , and the specific integration method is calculated as follows:

That is, the final similarity of disease is the average of functional similarity and Gaussian kernel similarity.

Further, in the step 5, according to the final similarity of microorganisms S _m , the final similarity of diseases S _d , the known microorganism-disease data adjacency matrix Y is integrated into a two-layer heterogeneous network, and the double random walk method is used to continue Forecasting, the forecasting process is as follows:

First, column normalization is performed on the final microbial similarity matrix S _m data to obtain a random walk microbial similarity matrix MM∈N _m ×N _m , which is calculated as follows:

Similarly, column normalization is performed on the final disease similarity matrix S _d data to obtain a random walk disease similarity matrix MD∈N _d ×N _d , which is calculated as follows:

Then, walking in this two-layer heterogeneous network at the same time, the process is as follows:

Iterative left walk in the microbial network:

Iteratively perform right walks in the disease network:

Among them, t is the number of current iterations, P _t ∈ _{N m} ×N _d represents the microorganism-disease association score matrix predicted by the t-th iteration, P _t (i,j) represents the association score between microorganism i and disease j (association degree); L_P _t represents the new microorganism-disease association score matrix obtained from the t-th iteration prediction on the microbial network, and R_P _t represents the microorganism-disease association score matrix obtained from the t-th iteration prediction on the disease network; P ₀ is the normalization matrix of the adjacency matrix Y∈N _m ×N _d ,

α is the attenuation parameter, I _l and I _r are the parameters of the maximum number of iterations of the microbial network and the disease network respectively, and the values of α, I _l and I _r are determined based on experience or cross-validation (set the value of the attenuation parameter to 0.1, I _l and I The values of _r are 2 and 1 respectively); L _num and R _num are respectively the number of iterations completed by the microbial network and the disease network,

When P _t converges (P _t+1 -P _t is less than a small threshold (such as 10 ^-10 ), it is considered that the walk reaches a steady state) or the iterative walk in the microbial network and the disease network both reach the maximum iteration When the number of times is , the iteration ends, and the final _Pt is the predicted microbe-disease correlation score matrix.

Beneficial effect:

The present invention proposes a microorganism-disease relationship prediction method based on similarity and double random walks to predict new relationships between microorganisms and diseases. First, calculate the functional similarity of the disease through the relationship between the disease gene and the similarity of gene function information, and then calculate the Gaussian kernel similarity of the disease based on the known microorganism-disease relationship, and further integrate to obtain the final similarity of the disease. Likewise, the microbial Gaussian kernel similarity was calculated based on the known microbial-disease relationship and served as the final microbial similarity. Microbial final similarity, disease final similarity, and known microbial-disease relationships are then integrated into a two-layer heterogeneous network. Finally, different walking steps are set in the microbial similarity network and disease similarity network, and the final microbial-disease relationship-to-association score is predicted by iterating to a certain steady state. Comparison of prediction results with other methods through five-fold crossover and leave-one-out validation shows that the present invention can more effectively predict the relationship between microorganisms and diseases. It can provide an important foundation for the subsequent development of computational models for predicting the relationship between microorganisms and diseases, provide basic guidance for biomedical experiments, and save human and material costs.

In the present invention, column normalization processing is performed on the microbial similarity matrix and the disease similarity matrix in the process of constructing the double-layer heterogeneous network. In the process of random walk, the iterative step limit of random walk is set respectively for the microbial network and the disease network. The final association score was predicted by combining similar diseases associated with similar microbes and similar microbes associated with similar diseases in both networks. And the same five-fold cross-validation and leave-one-out validation methods as in the KATZHMDA method were adopted to compare the prediction performance, and the prediction performance of the present invention was shown through the analysis of the AUC index.

Aiming at the field of microorganism-disease relationship, the present invention provides an effective method for predicting the relationship between microorganisms and diseases through a calculation model, which can provide an important basis for subsequent research on calculation models in this field and an overall understanding of disease mechanisms To help and further advance the development of drugs and the diagnosis and treatment of complex diseases.

Description of drawings

Fig. 1 is the overall flow chart of the present invention;

Fig. 2 is the five-fold cross-validation comparison diagram of the present invention on the data set;

Fig. 3 is that the present invention leaves one-out cross-validation comparative figure on data set;

Detailed ways

The present invention will be described in further detail below in conjunction with accompanying drawing and specific embodiment:

Example 1:

First, disease functional similarity is calculated using disease gene relationship and gene similarity information; disease Gaussian kernel similarity and microbial Gaussian kernel similarity are calculated based on known microorganism-disease relationship; disease functional similarity and Gaussian kernel similarity are used to integrate disease finally Similarity, the specific integration method is to take the mean value of disease Gaussian kernel similarity and disease function similarity. The microbial Gaussian kernel similarity was used as the final similarity of microorganisms. Then the microbial similarity information, disease similarity information and known microbial-disease relationship information are integrated into a two-layer heterogeneous network. Based on the starting point that similar microorganisms are associated with similar diseases and similar diseases are associated with similar microorganisms, the double random walk method is used to predict the microorganism-disease relationship in heterogeneous networks. The key process of the random walk method is to set different walking steps in the microbial similarity network and disease similarity network, and iterate to a certain steady state to obtain the final association score of the microorganism-disease relationship pair.

The known microorganism-disease relationship used in the present invention comes from the HMDAD (http://www.cuilab.cn/hmdad) database, including 39 kinds of diseases and 292 kinds of microorganisms in total, and its known microorganism-disease relationship number is 483 . After deduplication processing, the final number of relations is 450, and the number of diseases and microorganisms are 39 and 292, respectively. Disease gene relationship data comes from the DisGeNET database.

The whole process of microorganism-disease relationship prediction based on similarity and double random walk is shown in Figure 1, which can be divided into the following steps:

(1) The specific process of calculating the disease functional similarity D _funsim is:

First, for the disease pair A and B, define its functional similarity calculation formula as follows:

Among them, G _A ＝{g _A1 ,g _A2 ,...,g _Am } is the gene set associated with disease A, similarly, G _B ＝{g _B1 ,g _B2 ,... , g _Bn } is the gene set associated with disease B, m and n are the gene numbers in the gene set _GA and G _B respectively; is the functional similarity value of gene g _Ai and gene set G _B , is the functional similarity value of gene g _Bj and gene set _GA , and the corresponding calculation formula is as follows:

Among them, F(g _Ai , g _Bj ) is the semantic similarity value between genes g _Ai and g _Bj . The HumanNet database provides the calculated semantic similarity value based on the logarithmic likelihood function. The specific calculation method is as follows:

F(g _Ai ,g _Bj )＝LLS(g _Ai ,g _Bj ).

In the HumanNet database, the functional similarity value of genes 6188 and 6209 is 0.9697. According to the genes associated with the disease and the functional similarity of the genes, the functional similarity value of the diseases Gastric and duodenal ulcer and Gastro-oesophageal reflux is 0.1655.

(2) According to the known microorganism-disease relationship, the process of constructing the microbial Gaussian kernel similarity is as follows:

First, define is the collection of microorganisms, N _m is the number of microorganisms; is the collection of all diseases, N _d is the number of diseases; the adjacency matrix Y∈N _m *N _d indicates whether there is a known relationship between each microorganism and the disease. If there is a known relationship between microorganism m _i and disease d _j , then the value of y _ij is 1, otherwise the value is 0. For example, the Gaussian kernel similarity calculation method of microorganisms m ₁ and m ₂ is defined as follows:

K _GIP,m (m ₁ ,m ₂ )＝exp(-γ _m ||ym ₁ -ym ₂ || ² ).

in, γ _m is an adjustment parameter to control the width of the kernel, and its calculation method is as follows:

Among them, γ' _m is set to 1 according to the Gaussian kernel experience. According to the above calculation formula, the Gaussian kernel similarity value of Actinobacteria and Actinobacteria is 0.0390.

Similarly, the Gaussian kernel similarity for defining diseases _d1 and _d2 is calculated as follows:

K _GIP,d (d ₁ ,d ₂ )＝exp(-γ _d ||yd ₁ -yd ₂ || ² )

Among them, γ' _d is also set to 1 according to experience. ; According to the above calculation formula, the Gaussian kernel similarity value of the disease Allergic asthma and the disease Atopicdermatitis is 0.4274.

(3) According to the calculated disease functional similarity D _funsim and disease Gaussian kernel similarity K _GIP,d integrate the final disease similarity process, the specific integration method is calculated as follows:

The final disease similarity was the average of functional similarity and Gaussian kernel similarity.

(4) The microbial Gaussian kernel similarity matrix K _GIP,m is used as the final microbial similarity matrix S _m :

S _m =K _GIP,m

Microbes have only one Gaussian kernel similarity, so their final similarity S _m is Gaussian kernel similarity.

(5) According to the final similarity S _m of microorganisms and the final similarity S _d of diseases, the known microorganism-disease data is integrated into a two-layer heterogeneous network, and the double random walk method is used to continue to predict. The prediction process is as follows:

First, column normalization is performed on the final microbial similarity matrix S _m data to obtain a random walk microbial similarity matrix MM∈N _m *N _m , which is calculated as follows:

Similarly, perform column normalization on the final disease similarity matrix S _d data to obtain a random walk disease similarity relationship matrix MD∈N _d *N _d , which is calculated as follows:

We define the matrix P ∈ _{N m} * N _d to denote the predicted microbe-disease relationship, and P(i,j) denotes the association score (degree of association) between microbe i and disease j. The prediction process of the random walk model is to walk in this two-layer heterogeneous network at the same time, so we set the maximum iteration parameters I _l and I _r for the microbial network and the disease network respectively. The walking process in a heterogeneous network is as follows:

Left walk in microbial network:

Right walks in disease networks:

Among them, t is the number of current iterations, P _t ∈ _{N m} ×N _d represents the microorganism-disease association score matrix predicted by the t-th iteration, P _t (i,j) represents the association score between microorganism i and disease j (association degree); L_P _t represents the new microorganism-disease association score matrix obtained from the t-th iteration prediction on the microbial network, and R_P _t represents the microorganism-disease association score matrix obtained from the t-th iteration prediction on the disease network; P ₀ is the normalization matrix of the adjacency matrix Y∈N _m ×N _d ,

α is the attenuation parameter, I _l and I _r are respectively the parameters of the maximum number of iterations of the microbial network and the disease network, and the values of α, I _l and I _r are determined according to experience or cross-validation (in this embodiment, according to experience and cross-validation, set The value of the attenuation parameter is 0.1, and the values of the maximum iteration number parameters I _l and I _r are 2 and 1 respectively); L _num and R _num are the number of iterations that the microbial network and the disease network have completed, respectively,

When P _t converges (P _t+1 -P _t is less than a small threshold (such as 10 ^-10 ), it is considered that the walk reaches a steady state) or the iterative walk in the microbial network and the disease network both reach the maximum iteration When the number of times is , the iteration ends, and the final _Pt is the predicted microbe-disease correlation score matrix. The association scores in the matrix are sorted from large to small, and the higher the ranking of microorganism-disease pairs, the more likely there is an association relationship.

In order to verify the effectiveness of the present invention, we refer to the verification standards of other algorithms and adopt two verification methods: (1) 5-fold cross-validation; (2) leave-one-out verification. In the five-fold cross-validation, the known microorganism-disease relationship is randomly divided into 5 parts, and one part is selected in turn as the test set, and the remaining 4 parts are used as the training set, and the number of test verifications is 100. In the leave-one-out validation, one of the known microorganism-disease relationships is sequentially selected as the test set, and the rest are used as the training set. The evaluation index used is AUC (the areas under ROC curves) value.

Figure 2 shows the AUC plot integrating disease functional similarity and Gaussian kernel similarity in five-fold cross-validation. It can be seen from the figure that the AUC value of the present invention (Predicting microbe-disease interactions based on similarities and bi-random walk on the heterogeneous network, referred to as BRWH-MDI) is 0.8676, which is better than the other three disease-based Gaussian kernel similarity , Disease Characterization Similarity and Microbial Gaussian Kernel Similarity (KATZHMDA: 0.8567, HGBI: 0.7762, NBI: 0.5622). Especially when the error rate (FPR value) is low, the correct rate (TPR value) is higher, which proves that the microorganism-disease relationship ranked first in the prediction results of the present invention is more correct.

Figure 3 depicts the performance comparison of various methods in leave-one-out validation by integrating disease functional similarity and Gaussian kernel similarity. It can also be seen from the figure that the AUC value of the BRWH of the present invention is 0.8780, which is also due to the performance of the other three methods (KATZHMDA: 0.8644, HGBI: 0.7866, NBI: 5553). Also when the error rate (FPR value) is low, the correct rate (TPR value) is higher, which also shows the higher accuracy of the top microbial-disease relationship in the prediction results of the present invention.

Through the performance of the above application cases, the present invention can more accurately predict new microorganism-disease relationships, provide guidance for subsequent biomedical experiments, and improve the level of disease diagnosis and treatment.

Claims (5)

1. A method for predicting a microorganism-disease relationship based on similarity and biradical walk, comprising the steps of:

step 1: separately constructing disease function similarity matrix D_funsimDisease Gaussian kernel similarity matrix K_GIP,dAnd the Gaussian kernel similarity matrix K of the microorganisms_GIP,m；

Step 2: integrating disease functional similarity matrix D_funsimAnd the disease Gaussian kernel similarity matrix K_GIP,dObtaining the final similarity matrix S of the diseases_d(ii) a Subjecting microorganisms to Gaussian nuclear phaseSimilarity matrix K_GIP,mAs a final similarity matrix S for the microorganisms_m；

And step 3: according to the known microorganism-disease relationship, the final similarity matrix S of the microorganism_mAnd final disease similarity matrix S_dConstructing a double-layer heterogeneous network, and predicting the relevance score of the microorganism-disease pair by using a double random walk method; the greater the association score, the greater the likelihood that the corresponding microorganism-disease pair is associated.

2. The method for predicting microorganism-disease relationship based on similarity and biradical walk according to claim 1, wherein in the step 1, the functional similarity between two diseases is calculated according to the known disease-gene relationship and the functional similarity of genes, and then a disease functional similarity matrix D is constructed according to the functional similarity between every two diseases_funsim；

For any two diseases a and B, the functional similarity calculation is as follows:

<mrow> <msub> <mi>D</mi> <mrow> <mi>f</mi> <mi>u</mi> <mi>n</mi> <mi>s</mi> <mi>i</mi> <mi>m</mi> </mrow> </msub> <mrow> <mo>(</mo> <mi>A</mi> <mo>,</mo> <mi>B</mi> <mo>)</mo> </mrow> <mo>=</mo> <mfrac> <mrow> <munder> <mo>&amp;Sigma;</mo> <mrow> <mn>1</mn> <mo>&amp;le;</mo> <mi>i</mi> <mo>&amp;le;</mo> <mi>m</mi> </mrow> </munder> <msub> <mi>FS</mi> <msub> <mi>G</mi> <mi>B</mi> </msub> </msub> <mrow> <mo>(</mo> <msub> <mi>g</mi> <mrow> <mi>A</mi> <mi>i</mi> </mrow> </msub> <mo>)</mo> </mrow> <mo>+</mo> <munder> <mo>&amp;Sigma;</mo> <mrow> <mn>1</mn> <mo>&amp;le;</mo> <mi>j</mi> <mo>&amp;le;</mo> <mi>n</mi> </mrow> </munder> <msub> <mi>FS</mi> <msub> <mi>G</mi> <mi>A</mi> </msub> </msub> <mrow> <mo>(</mo> <msub> <mi>g</mi> <mrow> <mi>B</mi> <mi>j</mi> </mrow> </msub> <mo>)</mo> </mrow> </mrow> <mrow> <mi>m</mi> <mo>+</mo> <mi>n</mi> </mrow> </mfrac> <mo>,</mo> </mrow>

wherein G is_A＝{g_A1,g_A2,......,g_AmIs the set of genes associated with disease A, likewise, G_B＝{g_B1,g_B2,......,g_BnThe gene set associated with disease B, and m and n are the gene sets G_AAnd G_BThe number of genes in (a);is gene g_AiAnd gene set G_BIs determined by the functional similarity value of (a),is gene g_BjAnd gene set G_AThe corresponding calculation formula is as follows:

<mrow> <msub> <mi>FS</mi> <msub> <mi>G</mi> <mi>B</mi> </msub> </msub> <mrow> <mo>(</mo> <msub> <mi>g</mi> <mrow> <mi>A</mi> <mi>i</mi> </mrow> </msub> <mo>)</mo> </mrow> <mo>=</mo> <munder> <mrow> <mi>m</mi> <mi>a</mi> <mi>x</mi> </mrow> <mrow> <mn>1</mn> <mo>&amp;le;</mo> <mi>j</mi> <mo>&amp;le;</mo> <mi>n</mi> </mrow> </munder> <mrow> <mo>(</mo> <mi>F</mi> <mo>(</mo> <mrow> <msub> <mi>g</mi> <mrow> <mi>A</mi> <mi>i</mi> </mrow> </msub> <mo>,</mo> <msub> <mi>g</mi> <mrow> <mi>B</mi> <mi>j</mi> </mrow> </msub> </mrow> <mo>)</mo> <mo>)</mo> </mrow> <mo>,</mo> </mrow>

<mrow> <msub> <mi>FS</mi> <msub> <mi>G</mi> <mi>A</mi> </msub> </msub> <mrow> <mo>(</mo> <msub> <mi>g</mi> <mrow> <mi>B</mi> <mi>j</mi> </mrow> </msub> <mo>)</mo> </mrow> <mo>=</mo> <munder> <mrow> <mi>m</mi> <mi>a</mi> <mi>x</mi> </mrow> <mrow> <mn>1</mn> <mo>&amp;le;</mo> <mi>i</mi> <mo>&amp;le;</mo> <mi>m</mi> </mrow> </munder> <mrow> <mo>(</mo> <mi>F</mi> <mo>(</mo> <mrow> <msub> <mi>g</mi> <mrow> <mi>A</mi> <mi>i</mi> </mrow> </msub> <mo>,</mo> <msub> <mi>g</mi> <mrow> <mi>B</mi> <mi>j</mi> </mrow> </msub> </mrow> <mo>)</mo> <mo>)</mo> </mrow> <mo>,</mo> </mrow>

wherein F (g)_Ai,g_Bj) Is gene g_AiAnd g_BjThe semantic similarity value calculation value based on the log likelihood function is provided by the HumanNet database, and the specific calculation mode is as follows:

F(g_Ai,g_Bj)＝LLS(g_Ai,g_Bj).

where LLS represents a log-likelihood function (in the human net database, calculating a genetic semantic similarity value using a log-likelihood function is prior art).

3. The method for predicting microorganism-disease relationship based on similarity and biradical walk according to claim 1, wherein in the step 1, the disease Gaussian kernel similarity matrix K is respectively constructed according to the known microorganism-disease relationship_GIP,dAnd the Gaussian kernel similarity matrix K of the microorganisms_GIP,mThe process is as follows:

first, defineIs a collection of microorganisms, N_mIs the number of microorganisms;is a collection of all diseases, N_dIs the number of diseases; adjacency matrix Y ∈ N_m×N_dIndicating whether a known relationship exists between each microorganism and the disease; if the microorganism m_iAnd disease d_jThere is a known relationship of y_ijThe value is 1, otherwise the value is 0;

then, calculating the Gaussian nuclear similarity between every two diseases; for any two diseases d₁And d₂The Gaussian kernel similarity calculation mode is as follows:

K_GIP,d(d₁,d₂)＝exp(-γ_d||yd₁-yd₂||²)

<mrow> <msub> <mi>&amp;gamma;</mi> <mi>d</mi> </msub> <mo>=</mo> <msub> <msup> <mi>&amp;gamma;</mi> <mo>&amp;prime;</mo> </msup> <mi>d</mi> </msub> <mo>/</mo> <mrow> <mo>(</mo> <mfrac> <mn>1</mn> <msub> <mi>N</mi> <mi>d</mi> </msub> </mfrac> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>j</mi> <mo>=</mo> <mn>1</mn> </mrow> <msub> <mi>N</mi> <mi>d</mi> </msub> </munderover> <mo>|</mo> <mo>|</mo> <msub> <mi>yd</mi> <mi>j</mi> </msub> <mo>|</mo> <msup> <mo>|</mo> <mn>2</mn> </msup> <mo>)</mo> </mrow> <mo>,</mo> </mrow>

wherein,γ_dadjustment parameter, γ ', for controlling the core width'_dFor disease bandwidth parameters, set to 1 according to (gaussian kernel usage) experience;

then calculating the Gaussian nuclear similarity between every two microorganisms; for any two microorganisms m₁And m₂The Gaussian kernel similarity calculation mode is as follows:

K_GIP,m(m₁,m₂)＝exp(-γ_m||ym₁-ym₂||²).

<mrow> <msub> <mi>&amp;gamma;</mi> <mi>m</mi> </msub> <mo>=</mo> <msub> <msup> <mi>&amp;gamma;</mi> <mo>&amp;prime;</mo> </msup> <mi>m</mi> </msub> <mo>/</mo> <mrow> <mo>(</mo> <mfrac> <mn>1</mn> <msub> <mi>N</mi> <mi>m</mi> </msub> </mfrac> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <msub> <mi>N</mi> <mi>m</mi> </msub> </munderover> <mo>|</mo> <mo>|</mo> <msub> <mi>ym</mi> <mi>i</mi> </msub> <mo>|</mo> <msup> <mo>|</mo> <mn>2</mn> </msup> <mo>)</mo> </mrow> <mo>,</mo> </mrow>

wherein,γ_madjustment parameter, γ ', for controlling the core width'_mThe bandwidth parameter for the microorganism is empirically set to 1; a

Finally, constructing a disease Gaussian kernel similarity matrix K by the Gaussian kernel similarity between every two diseases_GIP,dConstructing a microorganism Gaussian nucleus similarity matrix K by the Gaussian nucleus similarity between every two microorganisms_GIP,m。

4. The method for predicting microorganism-disease relationship based on similarity and biradical walk according to claim 1, wherein in the step 2, a disease function similarity matrix D is integrated_funsimAnd the disease Gaussian kernel similarity matrix K_GIP,dObtaining the final similarity matrix S of the diseases_dThe specific integration mode is calculated as follows:

<mrow> <msub> <mi>S</mi> <mi>d</mi> </msub> <mo>=</mo> <mfrac> <mrow> <msub> <mi>D</mi> <mrow> <mi>f</mi> <mi>u</mi> <mi>n</mi> <mi>s</mi> <mi>i</mi> <mi>m</mi> </mrow> </msub> <mo>+</mo> <msub> <mi>K</mi> <mrow> <mi>G</mi> <mi>I</mi> <mi>P</mi> <mo>,</mo> <mi>d</mi> </mrow> </msub> </mrow> <mn>2</mn> </mfrac> </mrow>

i.e., the final similarity of disease is the average of the functional similarity and the gaussian kernel similarity.

5. The method for predicting microorganism-disease relationship based on similarity and biradical walk according to claim 1, wherein in the step 5, the final similarity S is determined according to the microorganism_mUltimate similarity of disease S_dThe known microbe-disease data adjacency matrix Y is integrated into a double-layer heterogeneous network, and the prediction is continued by using a double random walk method, wherein the prediction process comprises the following steps:

first, a final similarity matrix S is generated for the microorganisms_mThe data is subjected to column normalization processing to obtain a microorganism similarity relation matrix MM ∈ N which is randomly migrated_m×N_mThe calculation method is as follows:

<mrow> <mi>M</mi> <mi>M</mi> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> <mo>=</mo> <mfenced open = "{" close = ""> <mtable> <mtr> <mtd> <mfrac> <mrow> <msub> <mi>S</mi> <mi>m</mi> </msub> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> </mrow> <mrow> <msubsup> <mo>&amp;Sigma;</mo> <mrow> <mi>k</mi> <mo>=</mo> <mn>1</mn> </mrow> <mrow> <mi>k</mi> <mo>=</mo> <msub> <mi>N</mi> <mi>m</mi> </msub> </mrow> </msubsup> <msub> <mi>S</mi> <mi>m</mi> </msub> <mrow> <mo>(</mo> <mi>k</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> </mrow> </mfrac> </mtd> <mtd> <mtable> <mtr> <mtd> <mrow> <mi>i</mi> <mi>f</mi> </mrow> </mtd> <mtd> <mrow> <msubsup> <mo>&amp;Sigma;</mo> <mrow> <mi>k</mi> <mo>=</mo> <mn>1</mn> </mrow> <mrow> <mi>k</mi> <mo>=</mo> <msub> <mi>N</mi> <mi>m</mi> </msub> </mrow> </msubsup> <msub> <mi>S</mi> <mi>m</mi> </msub> <mrow> <mo>(</mo> <mi>k</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> <mo>&gt;</mo> <mn>0</mn> </mrow> </mtd> </mtr> </mtable> </mtd> </mtr> <mtr> <mtd> <mn>0</mn> </mtd> <mtd> <mrow> <mi>o</mi> <mi>t</mi> <mi>h</mi> <mi>e</mi> <mi>r</mi> <mi>w</mi> <mi>i</mi> <mi>s</mi> <mi>e</mi> <mo>.</mo> </mrow> </mtd> </mtr> </mtable> </mfenced> </mrow>

likewise, the final similarity matrix S for disease_dData is processed by column normalizationObtaining a random walk disease similarity relation matrix MD belonging to N_d×N_dThe calculation method is as follows:

<mrow> <mi>M</mi> <mi>D</mi> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> <mo>=</mo> <mfenced open = "{" close = ""> <mtable> <mtr> <mtd> <mfrac> <mrow> <msub> <mi>S</mi> <mi>d</mi> </msub> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> </mrow> <mrow> <msubsup> <mo>&amp;Sigma;</mo> <mrow> <mi>k</mi> <mo>=</mo> <mn>1</mn> </mrow> <mrow> <mi>k</mi> <mo>=</mo> <msub> <mi>N</mi> <mi>d</mi> </msub> </mrow> </msubsup> <msub> <mi>S</mi> <mi>d</mi> </msub> <mrow> <mo>(</mo> <mi>k</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> </mrow> </mfrac> </mtd> <mtd> <mtable> <mtr> <mtd> <mrow> <mi>i</mi> <mi>f</mi> </mrow> </mtd> <mtd> <mrow> <msubsup> <mo>&amp;Sigma;</mo> <mrow> <mi>k</mi> <mo>=</mo> <mn>1</mn> </mrow> <mrow> <mi>k</mi> <mo>=</mo> <msub> <mi>N</mi> <mi>d</mi> </msub> </mrow> </msubsup> <msub> <mi>S</mi> <mi>d</mi> </msub> <mrow> <mo>(</mo> <mi>k</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> <mo>&gt;</mo> <mn>0</mn> </mrow> </mtd> </mtr> </mtable> </mtd> </mtr> <mtr> <mtd> <mn>0</mn> </mtd> <mtd> <mrow> <mi>o</mi> <mi>t</mi> <mi>h</mi> <mi>e</mi> <mi>r</mi> <mi>w</mi> <mi>i</mi> <mi>s</mi> <mi>e</mi> <mo>.</mo> </mrow> </mtd> </mtr> </mtable> </mfenced> </mrow>

then, the two layers of heterogeneous networks are walked simultaneously, and the process is as follows:

left walking is performed iteratively in the microbial network:

<mrow> <mi>L</mi> <mo>_</mo> <msub> <mi>P</mi> <mi>t</mi> </msub> <mo>=</mo> <mfenced open = "{" close = ""> <mtable> <mtr> <mtd> <mrow> <mi>&amp;alpha;</mi> <mo>&amp;times;</mo> <mi>M</mi> <mi>M</mi> <mo>&amp;times;</mo> <msub> <mi>P</mi> <mrow> <mi>t</mi> <mo>-</mo> <mn>1</mn> </mrow> </msub> <mo>+</mo> <mrow> <mo>(</mo> <mn>1</mn> <mo>-</mo> <mi>&amp;alpha;</mi> <mo>)</mo> </mrow> <mo>&amp;times;</mo> <mi>Y</mi> </mrow> </mtd> <mtd> <mrow> <mi>t</mi> <mo>&amp;le;</mo> <msub> <mi>I</mi> <mi>l</mi> </msub> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <mi>L</mi> <mo>_</mo> <msub> <mi>P</mi> <msub> <mi>I</mi> <mi>l</mi> </msub> </msub> </mrow> </mtd> <mtd> <mrow> <mi>t</mi> <mo>&gt;</mo> <msub> <mi>I</mi> <mi>l</mi> </msub> </mrow> </mtd> </mtr> </mtable> </mfenced> </mrow>

the right walk is iterated in the disease network:

<mrow> <mi>R</mi> <mo>_</mo> <msub> <mi>P</mi> <mi>t</mi> </msub> <mo>=</mo> <mfenced open = "{" close = ""> <mtable> <mtr> <mtd> <mrow> <mo>(</mo> <mn>1</mn> <mo>-</mo> <mi>&amp;alpha;</mi> <mo>)</mo> <mo>&amp;times;</mo> <msub> <mi>P</mi> <mrow> <mi>t</mi> <mo>-</mo> <mn>1</mn> </mrow> </msub> <mo>&amp;times;</mo> <mi>M</mi> <mi>D</mi> <mo>+</mo> <mi>&amp;alpha;</mi> <mo>&amp;times;</mo> <mi>Y</mi> </mrow> </mtd> <mtd> <mrow> <mi>t</mi> <mo>&amp;le;</mo> <msub> <mi>I</mi> <mi>r</mi> </msub> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <mi>R</mi> <mo>_</mo> <msub> <mi>P</mi> <msub> <mi>I</mi> <mi>r</mi> </msub> </msub> </mrow> </mtd> <mtd> <mrow> <mi>t</mi> <mo>&gt;</mo> <msub> <mi>I</mi> <mi>r</mi> </msub> </mrow> </mtd> </mtr> </mtable> </mfenced> </mrow>

<mrow> <msub> <mi>P</mi> <mi>t</mi> </msub> <mo>=</mo> <mfrac> <mrow> <mi>L</mi> <mo>_</mo> <msub> <mi>P</mi> <mi>t</mi> </msub> <mo>&amp;times;</mo> <msub> <mi>L</mi> <mrow> <mi>n</mi> <mi>u</mi> <mi>m</mi> </mrow> </msub> <mo>+</mo> <mi>R</mi> <mo>_</mo> <msub> <mi>P</mi> <mi>t</mi> </msub> <mo>&amp;times;</mo> <msub> <mi>R</mi> <mrow> <mi>n</mi> <mi>u</mi> <mi>m</mi> </mrow> </msub> </mrow> <mrow> <msub> <mi>L</mi> <mrow> <mi>n</mi> <mi>u</mi> <mi>m</mi> </mrow> </msub> <mo>+</mo> <msub> <mi>R</mi> <mrow> <mi>n</mi> <mi>u</mi> <mi>m</mi> </mrow> </msub> </mrow> </mfrac> </mrow>

where t is the number of current iterations, P_t∈N_m×N_dRepresenting the microorganism-disease association score matrix obtained by the t-th iteration prediction, P_t(i, j) represents the association score (degree of association) of microorganism i and disease j; l _ P_tRepresenting a new microorganism-disease association score matrix, RpP, from the t-th iterative prediction on the microorganism network_tRepresenting a microorganism-disease association score matrix obtained by carrying out the t iterative prediction on a disease network; p₀For the adjacency matrix Y ∈ N_m×N_dThe normalized matrix of (a) is determined,α is the attenuation parameter, I_lAnd I_rThe maximum iteration number parameters of the microbial network and the disease network are α and I respectively_lAnd I_rIs determined empirically or by cross-validation (setting the value of the attenuation parameter to 0.1, I)_lAnd I_rThe values of (a) and (b) are 2 and 1, respectively); l is_numAnd R_numThe number of iterations that the microbial network and disease network have completed respectively,

when P is present_tConvergence (P)_t+1-P_tLess than a small threshold (e.g., 10)^-10) Thought migration reaches steady state) or when iterative migration in both the microbial network and the disease network reaches the maximum number of iterations, the iteration is ended, and the final P is_tNamely the microorganism-disease association score matrix obtained by prediction.