CN112863634A - Traditional Chinese medicine prescription recommendation method and system based on new crown protein heterogeneous network clustering - Google Patents

Abstract

The invention discloses a method and system for recommending traditional Chinese medicine prescriptions based on novel coronavirus protein heterogeneous network clustering, including: constructing edge connections between candidate traditional Chinese medicines and their compounds, and edge connections between compounds and novel coronavirus protein targets, and constructing traditional Chinese medicine-compounds according to the edge connections ‑Protein Heterogeneous Network; determine the meta-path of the traditional Chinese medicine-compound-protein heterogeneous network, and calculate the similarity between the heterogeneous nodes under the meta-path, so as to obtain the transition probability matrix of the traditional Chinese medicine-protein, and obtain the heterogeneous nodes according to the transition probability matrix Embedding vector; according to the Euclidean distance between Chinese medicine and protein calculated by the embedding vector, the protein is clustered, and the traditional Chinese medicine clustering combination centered on structural protein, non-structural protein and auxiliary protein is obtained; according to the traditional Chinese medicine database, the traditional Chinese medicine clustering combination is The candidate Chinese medicines were screened and recommended Chinese medicines were obtained. Through the TCM-compound-protein heterogeneous network, clustering based on SARS-CoV-2 was used to recommend potential TCMs for the treatment of COVID-19.

Description

Method and system for recommendation of traditional Chinese medicine prescription based on novel coronavirus protein heterogeneous network clustering

technical field

The invention relates to the technical field of drug recommendation, in particular to a method and system for recommending traditional Chinese medicine prescriptions based on novel coronavirus protein heterogeneous network clustering.

Background technique

The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

TCM drug repositioning is the study of expanding the clinical indications and discovering new targets or mechanisms of TCM monomers or TCM compounds that have been approved for clinical use. The genome of the new coronavirus is less than 30,000 bases, and it can transcribe 29 proteins, including 16 non-structural proteins, 4 structural proteins and 9 auxiliary proteins. Targets of action are crucial to drug development. Since the new coronavirus acts on the human body through a variety of proteins, if it is treated with a single target, it will not produce a good therapeutic effect. Therefore, researchers mostly use multi-target therapy to treat patients with new coronary pneumonia. Traditional Chinese medicine happens to have multiple components-multiple characteristics of the target. Using network pharmacology, the mechanism of drug action and its main components are analyzed on the basis of the "drug-compound-target" interaction network. Through the study of multiple targets, the therapeutic effect can be improved, thereby improving the success rate of drug repositioning.

In order to mine potential prescriptions for alternative treatment of COVID-19, in the document Identifying potentialtreatments of COVID-19 from Traditional Chinese Medicine (TCM) by using a data-driven approach, Ren used data mining and association networks to identify high-frequency medicinal materials and prescriptions in ancient prescriptions to dig. Through the method of molecular docking, the virtual screening of high-frequency CMs and compatible traditional Chinese medicines was explored. Furthermore, in the literature Analysis on herbal medicines utilized for treatment of COVID-19, Luo, by using complex system entropy and unsupervised hierarchical clustering, 8 core herbal combinations and 10 new formulations may become useful drug candidates for COVID-19. These methods for finding potential TCM formulas all use more traditional data mining and clustering methods to discover potential formulas based on frequency, but do not consider the interactions among herbs, compounds, and proteins.

SUMMARY OF THE INVENTION

In order to solve the above problems, the present invention proposes a method and system for recommending traditional Chinese medicine prescriptions based on novel coronavirus protein heterogeneous network clustering. A clustering combination of traditional Chinese medicines centered on non-structural proteins and auxiliary proteins, and the potential traditional Chinese medicines for the treatment of COVID-19 are recommended according to the clustering results.

In order to achieve the above object, the present invention adopts the following technical solutions:

In a first aspect, the present invention provides a method for recommending traditional Chinese medicine prescriptions based on novel coronavirus protein heterogeneous network clustering, including:

Construct edge connections between candidate traditional Chinese medicines and their compounds, as well as edge connections between compounds and new crown protein targets, and construct a traditional Chinese medicine-compound-protein heterogeneous network based on the above edge connections;

Determine the meta-path of the Chinese medicine-compound-protein heterogeneous network, calculate the similarity between the heterogeneous nodes under the meta-path, and obtain the transition probability matrix of the Chinese medicine-protein, and obtain the embedding vector of the heterogeneous nodes according to the transition probability matrix;

According to the Euclidean distance of the Chinese medicine and the protein calculated by the embedding vector, the protein is clustered, and the Chinese medicine clustering combination centered on the structural protein, the non-structural protein and the auxiliary protein is obtained respectively;

According to the traditional Chinese medicine database, the candidate traditional Chinese medicines in the traditional Chinese medicine clustering combination are screened, and the recommended traditional Chinese medicines are obtained.

In the second aspect, the present invention provides a traditional Chinese medicine prescription recommendation system based on novel coronavirus protein heterogeneous network clustering, including:

The network building block is configured to construct the edge connection between the candidate traditional Chinese medicine and its compound, as well as the edge connection between the compound and the new crown protein target, and construct the traditional Chinese medicine-compound-protein heterogeneous network according to the above-mentioned edge connection;

The similarity calculation module is configured to determine the meta-path of the traditional Chinese medicine-compound-protein heterogeneous network, and calculate the similarity between the heterogeneous nodes under the meta-path, so as to obtain the transition probability matrix of the traditional Chinese medicine-protein, and obtain the heterogeneity according to the transition probability matrix. Embedding vector for constructing nodes;

The clustering module is configured to cluster the proteins according to the Euclidean distance of the traditional Chinese medicine and the protein calculated by the embedding vector, and obtain the traditional Chinese medicine clustering combination centered on the structural protein, the non-structural protein and the auxiliary protein respectively;

The recommendation module is configured to screen the candidate traditional Chinese medicines in the traditional Chinese medicine clustering combination according to the traditional Chinese medicine database to obtain the recommended traditional Chinese medicines.

In a third aspect, the present invention provides an electronic device, comprising a memory, a processor, and computer instructions stored in the memory and executed on the processor, and when the computer instructions are executed by the processor, the method described in the first aspect is completed .

In a fourth aspect, the present invention provides a computer-readable storage medium for storing computer instructions, and when the computer instructions are executed by a processor, the method described in the first aspect is completed.

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

The present invention recommends potential Chinese medicines for treating COVID-19 by establishing a Chinese medicine-compound-protein heterogeneous network and clustering based on the novel coronavirus protein. First, the candidate traditional Chinese medicines and the compounds contained in them are established to establish edge connections, and molecular docking is used to establish the edge connections between the compounds and the new crown protein target, so as to obtain the traditional Chinese medicine-compound-protein heterogeneous network; secondly, the structured network embedding model SDNE is used to obtain all the The embedding vector of the node; finally, by calculating the Euclidean distance between the protein and the traditional Chinese medicine, the clustering combination of the traditional Chinese medicine centered on the structural protein, the non-structural protein and the auxiliary protein is obtained, and the corresponding new crown protein of these traditional Chinese medicine may have potential curative effect. In addition, the present invention also calculates the cosine similarity between the traditional Chinese medicine in the cluster combination and the traditional Chinese medicine in the diagnosis and treatment guide, and obtains the potential traditional Chinese medicine with the highest similarity with the recommended traditional Chinese medicine in the guide traditional Chinese medicine and used for the recommended treatment, and these results can replace the traditional Chinese medicine in the diagnosis and treatment guide. possibility and provide a potential candidate traditional Chinese medicine for the treatment of COVID-19.

Because there is no direct connection between some heterogeneous nodes, the first-order similarity is not enough to preserve the network structure. In order to solve this problem, the present invention introduces the second-order similarity, retains the global information through the second-order similarity, and realizes the first-order similarity through the first-order similarity. degree and second-order similarity preserve both global and local structure.

The heterogeneous network established by the invention includes a plurality of heterogeneous nodes and relationships, and the interactive information of traditional Chinese medicine, compounds and proteins is deeply excavated to generate better clustering results.

Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will become apparent from the description which follows, or may be learned by practice of the invention.

Description of drawings

The accompanying drawings forming a part of the present invention are used to provide further understanding of the present invention, and the exemplary embodiments of the present invention and their descriptions are used to explain the present invention, and do not constitute an improper limitation of the present invention.

1 is a flowchart of a method for recommending traditional Chinese medicine prescriptions based on novel coronavirus protein heterogeneous network clustering provided in Example 1 of the present invention;

Figures 2(a)-2(f) are schematic diagrams of the recommended flow of Chinese medicine prescriptions that can replace the new crown treatment provided in Example 1 of the present invention.

Detailed ways:

The present invention will be further described below with reference to the accompanying drawings and embodiments.

It should be noted that the following detailed description is exemplary and intended to provide further explanation of the invention. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

It should be noted that the terminology used herein is for the purpose of describing specific embodiments only, and is not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly dictates otherwise, the singular is intended to include the plural as well, furthermore, it is to be understood that the terms "including" and "having" and any conjugations thereof are intended to cover the non-exclusive A process, method, system, product or device comprising, for example, a series of steps or units is not necessarily limited to those steps or units expressly listed, but may include those steps or units not expressly listed or for such processes, methods, Other steps or units inherent to the product or equipment.

Embodiments of the invention and features of the embodiments may be combined with each other without conflict.

Example 1

As shown in Figure 1 and Figure 2(a)-(f), this embodiment provides a method for recommending traditional Chinese medicine prescriptions based on novel coronavirus protein heterogeneous network clustering, including:

S1: Construct the edge connection between the candidate traditional Chinese medicine and its compound, as well as the edge connection between the compound and the new crown protein target, and construct the traditional Chinese medicine-compound-protein heterogeneous network according to the above edge connection;

S2: Determine the meta-path of the traditional Chinese medicine-compound-protein heterogeneous network, calculate the similarity between the heterogeneous nodes under the meta-path, so as to obtain the transition probability matrix of the traditional Chinese medicine-protein, and obtain the embedding vector of the heterogeneous nodes according to the transition probability matrix;

S3: Cluster the proteins according to the Euclidean distance between the Chinese medicine and the protein calculated by the embedding vector, and obtain the Chinese medicine clustering combination centered on the structural protein, the non-structural protein and the auxiliary protein respectively;

S4: Screen the candidate traditional Chinese medicines in the traditional Chinese medicine clustering combination according to the traditional Chinese medicine database, and obtain the recommended traditional Chinese medicines.

In this example, 482 candidate traditional Chinese medicines, 13,448 compound structures and corresponding ADME information were obtained from the traditional Chinese medicine system pharmacology database and analysis platform TCMS; then they were converted into PDBQT format through OpenBabel.

Download the existing SARS-CoV-2 structure from PROTEIN DATA BAND (RSCB), including S, nsp5, nsp7, nsp8, nsp9, nsp10, nsp12, nsp15, nsp16; for SARS-CoV with unclear structure -2, in this example, the predicted protein structure is downloaded from the laboratory, and the predicted protein structure is E, N, M, orf3a, orf6, orf7a, orf8, orf10, nsp1, nsp2, nsp3, nsp4, nsp6, nsp13, nsp14; a total of For 24 protein structures, the downloaded pdb files were dehydrated and hydrogenated using OpenBabel and converted to PDBQT format.

The target of traditional Chinese medicine refers to the biological macromolecules that have pharmacodynamic functions in the body and can be acted on by traditional Chinese medicine, such as some biological macromolecules such as proteins and nucleic acids. The interaction between traditional Chinese medicines and target proteins is the basis for many traditional Chinese medicines to exert their biological functions. Since there is currently no specific drug for the new crown virus, this example searches for the traditional Chinese medicine that hits the target by studying the protein encoded by the new coronavirus.

SARS-CoV-2 belongs to the family Coronaviridae and is a non-segmented single-stranded positive-stranded RNA virus. It encodes 29 proteins, including 4 structural proteins (S, E, M and N proteins) and 16 non-structural proteins. Proteins (nsp1-nsp16) and 9 accessory proteins; among them, 4 structural proteins S, E, M and N form the outer layer of the coronavirus and protect the internal RNA; the S protein is an important determinant for the virus to enter the host cell, and ACE2 It has a high binding strength; the N protein binds to the viral genome RNA and packages the RNA into a ribonucleoprotein (RNP) complex; the E protein is an envelope protein that wraps the genetic material inside the virus. 16 non-structural proteins (nsp1-nsp16) are used to form viral replicase transcriptase complexes. These proteins are generally the targets of traditional small-molecule antiviral Chinese medicines. Among the non-structural proteins, NSP7, NSP8 and NSP12 form ribonucleic acid ( RNA)-dependent RNA polymerase complex, which plays an important role in the replication and transcription cycle of the virus; NSP5 is a 3C proteolytic enzyme, which can liberate most other NSP proteins from the protein chain, and is considered to be anti-COVID-19 virus An important target for Chinese medicine research.

表示边；w表示边权重的集合，如果v_i和v_j中间没有边相连，w_ij＝0，否则w_ij>0；每一个节点v_i∈V都属于一个特定的节点类型

每一个边 e_ij∈E都属于关系类型集合R：φ(e)∈R中的一个特定的关系类型，当节点类型的总数目|A|>1或者边的类型的总数目|R|>1时，得到的图是异构图。For heterogeneous graphs; the graph is defined as G=(V,E,w), where V={v ₁ ,v ₂ ,...,v _n } denotes n vertices;

Represents an edge; w represents the set of edge weights, if there is no edge connected between _{vi and v j, w ij = 0} , otherwise w _ij >0; each node vi _∈V belongs to a specific node type

Each edge e _ij ∈ E belongs to the set of relation types R: φ(e)∈R for a particular relation type, when the total number of node types |A|>1 or the total number of edge types |R|> 1, the resulting graph is a heterogeneous graph.

Then, after obtaining the candidate traditional Chinese medicine, the compound contained in the traditional Chinese medicine, and the new crown protein target in this example, according to the edge connection between the candidate traditional Chinese medicine and its compound, and the edge connection between the compound and the new crown protein target, the traditional Chinese medicine-compound-protein heterogeneous network G= (V,E); specifically:

First, establish edge connections between 482 traditional Chinese medicines and the 13448 compounds they contain. If the traditional Chinese medicines contain certain compounds, there are edge connections between the traditional Chinese medicines and these compounds;

Then, establish the edge connection between the compound and the new crown protein in 24; in this example, the molecular docking method is used to judge the binding affinity of the compound-protein, and the scoring function c is used to calculate the binding affinity, namely:

是空间相互作用的加权和，r_ij是原子间的距离。in,

is the weighted sum of spatial interactions, and r _ij is the distance between atoms.

From the point of view of thermodynamics, the ligand-receptor interaction is a comprehensive equilibrium process. The stable conformation of biomolecules is the conformation with the lowest free energy. In this example, the compound and the protein are docked 10 times respectively, and the average of 10 times is taken. The value is the result of the docking. When the binding free energy, that is, the binding affinity is less than -7kcal/mol, the compound and the protein are considered to be able to bind effectively, otherwise, it is invalid, and the compound with the binding free energy less than -7kcal/mol is connected with the protein.

表示一个节点类型

的综合关系，

表示关系的合成器。元路径描述节点之间的语义关系，不同的元路径描述节点之间不同的语义关系，这个语义关系的开发是多种后续任务的基石。In the traditional Chinese medicine-compound-protein heterogeneous network G, heterogeneous nodes are connected by a variety of meta-paths. The meta-path is the core of the heterogeneous network and is a sequence of node types with edge types when modeling a specific relationship. A meta-path ρ is defined as a sequence of paths of node types A ₁ ,...A _l+1 connected by edge types R ₁ ,...R _l , such as

Represents a node type

comprehensive relationship,

A compositor representing a relationship. Meta-paths describe the semantic relationship between nodes, and different meta-paths describe different semantic relationships between nodes. The development of this semantic relationship is the cornerstone of various subsequent tasks.

In this example, four meta-paths are constructed, including Chinese medicine-compound-Chinese medicine, Chinese medicine-compound-protein, protein-compound-Chinese medicine, protein-compound-protein; wherein, the meta-path of Chinese medicine->compound->Chinese medicine represents two Chinese medicine contains the same compound, Chinese medicine->compound->protein means that Chinese medicine hits a protein through a certain compound, protein->compound->traditional medicine means that protein is combined with a compound in a Chinese medicine, protein->compound->protein Indicates that a compound can hit two proteins.

After the meta-path is determined, the random walk method is used to calculate the proximity between nodes v _i and v _j . For example, given a meta-path p: H->C->H, a traditional Chinese medicine node h ₁ connects three compounds, of which only There is a compound connected to traditional Chinese medicine h ₂ , and this compound is also connected to traditional Chinese medicine h ₃ , then the transition probability of traditional Chinese medicine h ₁ to traditional Chinese medicine h ₂ is

Since there is no direct connection between some nodes, only the first-order similarity is not enough to preserve the network structure. The first-order similarity represents the local proximity between two nodes. If there is a connection between the two nodes, the edge The weight represents the first-order similarity between nodes v _i and v _j ; otherwise, the first-order similarity between nodes v _i and v _j is 0; while the second-order similarity measures the similarity between two nodes without edge connections , the second-order similarity of nodes v _i and v _j is the similarity between adjacent network structures.

Therefore, in this embodiment, the second-order similarity is considered, and the global information is preserved through the second-order similarity. Among them, in this embodiment, p _u =(w _u ,1,...,w _u ,|V|) is used to represent the first-order similarity between u and all other vertices, then the second-order similarity between u and v is represented by p _u and p _v ; if the nodes v _i and v _j have more neighbors in common, the two nodes are more similar; if there is no vertex connected to both u and v, the second-order similarity between u and v is 0.

Based on the determined meta-paths HCH, HCP, PCH, and PCP, this embodiment uses a random walk method to calculate the proximity between two nodes, and constructs a traditional Chinese medicine-protein relationship matrix A, where each element of the matrix A is between two nodes. The transition probability of , the row vector is the embedding vector of each node, and the matrix A provides the neighborhood structure information of each node. Among them, the random walk method includes: given a graph G=(V, E), select an initial node, randomly select a neighbor node, move the initial node to the neighbor node; then use the current node as the starting point, repeat the above process, The randomly selected node sequence constitutes a random walk process.

The TCM-protein transition probability is defined as follows:

Among them, s(v _i , v _j |ρ) represents the similarity between node v _i and node v _j under the meta-path ρ, and l is the threshold of the meta-path length.

For similarity-based random walk events, there are the following properties:

是通过关系类型ψ(v_i,v′)从v_i到v′的转移概率，如果从v_i出发有m条边属于关系类型ψ(v_i,v′)，那么

in,

is the transition probability from v _i to v' through the relation type ψ(vi , v' _{), if there are m edges from v i belonging} to the relation type ψ(vi _, v'), then

This embodiment defines the similarity calculation based on the meta-path of traditional Chinese medicine as follows:

Among them, ρ[i:j] is from node v _i to node v _j , and the similarity calculation based on meta-path is a dynamic design method.

代表第i个节点的向量，根据中药-蛋白质的转移概率矩阵得到异构节点的嵌入向量，构建基于相似度的元路径中药图嵌入；具体地：给定网络G＝(V，E)，寻映射f，使得f：v_i→y_i，其中y_i的维度是d，且y_i∈R_d，d<<V。因此，映射f对每个节点向量进行降维，尽管节点类型不一样，最终得到的向量维度相同。这个过程是中药网络的嵌入，得到的嵌入向量仍然保留原来网络的结构及语义关系。In this embodiment, a random walk is used to calculate the similarity between two nodes, and then the transition probability matrix T of traditional Chinese medicine-protein is obtained, and the i-th row of the transition probability matrix is obtained.

The vector representing the ith node, the embedding vector of heterogeneous nodes is obtained according to the transition probability matrix of traditional Chinese medicine-protein, and the similarity-based meta-path traditional Chinese medicine graph embedding is constructed; specifically: given the network G=(V, E), find Map f such that f: v _i →y _i , where the dimension of y _i is d, and y _i ∈ R _d , d<<V. Therefore, the mapping f reduces the dimension of each node vector, although the node types are different, the final vector dimension is the same. This process is the embedding of the traditional Chinese medicine network, and the obtained embedding vector still retains the structure and semantic relationship of the original network.

Since only a few compounds can hit the protein target, there are only a few edges between TCM nodes and protein nodes, so in the TCM-protein transition probability matrix T, most elements are 0, that is, the transition probability matrix is A sparse matrix; in addition, the heterogeneous network constructed in this embodiment also has a high degree of nonlinearity. In order to capture the highly nonlinear structure and solve the sparse problem at the same time, this embodiment adopts structured deep network embedding (SDNE), SDNE not only It can capture the nonlinear structure, and can also preserve the local and global network structure. SDNE is a semi-supervised model that implements network embeddings to explore first- and second-order similarities, SDNE consists of a supervised part and an unsupervised part. The unsupervised component is a deep autoencoder, which can capture the second-order similarity after reconstructing the input data t _i , and the second-order similarity preserves the global structure; the supervised part can preserve the first-order through the Laplacian feature map similarity.

Thus, the present embodiment adopts SDNE to construct the objective function of traditional Chinese medicine-compound-protein heterogeneous network, specifically:

A deep autoencoder consists of an encoder and a decoder, the encoder compresses the input into a latent space representation, and the decoder reconstructs the input from the latent space representation. Given an input x _i , the hidden representation of each layer is as follows:

where σ is the sigmoid function; W ^(k) is the weight matrix of the kth layer; b ^(k) is the bias of the kth layer.

假设第k层是节点v_i的表示向量(仅考虑全局信息)，那么从第k层开始解码，最终得到解码后的向量

那么，自编码器的误差就是输入节点v_i的邻接向量的重构误差。因此，自编码器的目标是尽量减小输入和输出的重构误差。损失函数如下：This embodiment obtains the output by reversing the calculation process of the encoder

Assuming that the kth layer is the representation vector of the node v _i (only global information is considered), then start decoding from the kth layer, and finally get the decoded vector

Then, the error of the autoencoder is the reconstruction error of the adjacency vector of the input node v _i . Therefore, the goal of an autoencoder is to minimize the reconstruction error of the input and output. The loss function is as follows:

Due to the sparseness of the network constructed in this embodiment, that is, the number of zero elements in the transition probability matrix T is far more than the number of non-zero elements. If T is used directly as input, it is easier to reconstruct the zero elements in T; in order to solve the problem caused by sparsity, a larger penalty is imposed on the reconstruction error of non-zero elements than zero elements; the modified objective function as follows:

邻接矩阵中的0对应b＝1，非0元素的b>1；这样的目的是对有边连接的节点赋予更高权重。Among them, ⊙ is Hadamard product, which means that the corresponding elements are multiplied;

0 in the adjacency matrix corresponds to b=1, and b>1 for non-zero elements; the purpose of this is to give higher weights to nodes connected by edges.

The above loss function to obtain the second-order similarity, the unsupervised part of the model can maintain the global network structure by reconstructing the second-order proximity between vertices; in addition, this embodiment also uses the first-order similarity to capture the local structure . The first-order similarity loss function is defined as follows:

和

是自编码器对节点v_i和v_j的编码结果，LE可以保持一条边的两个节点在嵌入空间中的表示相对接近；在上式中s(v_i,v_j)是基于元路径利用随机游走得到的节点v_i到节点v_j的转移概率。The first-order similarity loss function draws on the idea of Laplacian Eigenmap. LE reconstructs the local feature structure by building a similarity graph. If nodes v _i and v _j are connected by edges, the distances in the embedding space are also relatively close.

and

is the encoding result of nodes v _i and v _j by the _{autoencoder, LE can keep the representation of two nodes of an edge relatively close in the embedding space; in the above formula s(vi , v j )} is based on the meta-path utilization The transition probability from node v _i to node v _j obtained by random walk.

In order to preserve the global structure and local structure at the same time, this embodiment optimizes L1 and L2 at the same time:

Among them, adding Lreg prevents overfitting:

也就是需要求偏导

和

The objective function is optimized using stochastic gradient descent (ASGD) with the goal of minimizing

That is to say, the need for partial guidance

and

After obtaining the embedding vector of each node through the SDNE model, calculate the Euclidean distance between the vectors, and cluster the traditional Chinese medicine nodes and protein nodes according to the Euclidean distance. The combination of three proteins, that is, the traditional Chinese medicine with the smallest output distance of structural protein, non-structural protein and auxiliary protein, is deduplicated, that is, the clustering combination of traditional Chinese medicine centered on structural protein, non-structural protein and auxiliary protein is obtained. The new coronavirus protein may have potential therapeutic effects, and the clustering results are shown in Table 1.

The present embodiment crawls the characteristics of all traditional Chinese medicines from the Herbnet database, including medicinal properties, meridians, efficacy, indications, compatibility of traditional Chinese medicines, etc., involving a total of 689 characteristics. Matching, it is found that 21 kinds of traditional Chinese medicines except Hypericum perforatum and Phellodendron phellodendri can find corresponding traditional Chinese medicine information in Herbnet, then these 21 kinds of traditional Chinese medicines are regarded as recommended traditional Chinese medicines.

The 21 recommended Chinese medicines have a certain possibility to replace the similar Chinese medicines in the diagnosis and treatment guidelines. A 97*689-dimensional 0-1 matrix was established for these 21 Chinese medicines and the 76 Chinese medicines involved in the Chinese eighth edition diagnosis and treatment guidelines. If the traditional Chinese medicine has corresponding characteristics, the corresponding element is recorded as 1, otherwise it is 0; then calculate the cosine similarity of the 21 recommended traditional Chinese medicines and the traditional Chinese medicine in the diagnosis and treatment guide, and obtain the top 10 similarity between the guide traditional Chinese medicine and the recommended traditional Chinese medicine and the potential for the recommended treatment. traditional Chinese medicine. Potential traditional Chinese medicines that can be used as an alternative for the treatment of COVID-19 are shown in Table 2.

Table 1: Clustering results of traditional Chinese medicine based on SARS-CoV-2

Table 2: Potential Chinese Medicines for Alternative Treatment of COVID-19

Example 2

This embodiment provides a traditional Chinese medicine prescription recommendation system based on novel coronavirus protein heterogeneous network clustering, including:

The network building block is configured to construct the edge connection between the candidate traditional Chinese medicine and its compound, as well as the edge connection between the compound and the new crown protein target, and construct the traditional Chinese medicine-compound-protein heterogeneous network according to the above-mentioned edge connection;

The similarity calculation module is configured to determine the meta-path of the traditional Chinese medicine-compound-protein heterogeneous network, and calculate the similarity between the heterogeneous nodes under the meta-path, so as to obtain the transition probability matrix of the traditional Chinese medicine-protein, and obtain the heterogeneity according to the transition probability matrix. Embedding vector for constructing nodes;

The clustering module is configured to cluster the proteins according to the Euclidean distance of the traditional Chinese medicine and the protein calculated by the embedding vector, and obtain the traditional Chinese medicine clustering combination centered on the structural protein, the non-structural protein and the auxiliary protein respectively;

The recommendation module is configured to screen the candidate traditional Chinese medicines in the traditional Chinese medicine clustering combination according to the traditional Chinese medicine database to obtain the recommended traditional Chinese medicines.

It should be noted here that the foregoing modules correspond to the steps described in Embodiment 1, and the examples and application scenarios implemented by the foregoing modules and corresponding steps are the same, but are not limited to the content disclosed in Embodiment 1 above. It should be noted that the above modules may be executed in a computer system such as a set of computer-executable instructions as part of the system.

In further embodiments, there is also provided:

An electronic device includes a memory, a processor, and computer instructions stored on the memory and executed on the processor, and when the computer instructions are executed by the processor, the method described in Embodiment 1 is completed. For brevity, details are not repeated here.

It should be understood that, in this embodiment, the processor may be a central processing unit CPU, and the processor may also be other general-purpose processors, digital signal processors DSP, application-specific integrated circuits ASIC, off-the-shelf programmable gate array FPGA or other programmable logic devices , discrete gate or transistor logic devices, discrete hardware components, etc. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.

The memory may include read-only memory and random access memory and provide instructions and data to the processor, and a portion of the memory may also include non-volatile random access memory. For example, the memory may also store device type information.

A computer-readable storage medium for storing computer instructions, when the computer instructions are executed by a processor, the method described in Embodiment 1 is completed.

The method in Embodiment 1 may be directly embodied as being executed by a hardware processor, or executed by a combination of hardware and software modules in the processor. The software modules may be located in random access memory, flash memory, read-only memory, programmable read-only memory or electrically erasable programmable memory, registers and other storage media mature in the art. The storage medium is located in the memory, and the processor reads the information in the memory, and completes the steps of the above method in combination with its hardware. To avoid repetition, detailed description is omitted here.

Those of ordinary skill in the art can realize that the unit, that is, the algorithm step of each example described in conjunction with this embodiment, can be implemented by electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Skilled artisans may implement the described functionality using different methods for each particular application, but such implementations should not be considered beyond the scope of this application.

The above are only preferred embodiments of the present invention, and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Although the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, they do not limit the scope of protection of the present invention. Those skilled in the art should understand that on the basis of the technical solutions of the present invention, those skilled in the art do not need to pay creative work. Various modifications or deformations that can be made are still within the protection scope of the present invention.

Claims (10)

1. A traditional Chinese medicine prescription recommendation method based on neocoronal protein heterogeneous network clustering is characterized by comprising the following steps:

constructing the connection of the candidate traditional Chinese medicine and the compound thereof, and the connection of the compound and the new crown protein target, and constructing the traditional Chinese medicine-compound-protein heterogeneous network according to the connection of the sides;

determining a meta-path of the traditional Chinese medicine-compound-protein heterogeneous network, calculating the similarity between heterogeneous nodes under the meta-path to obtain a traditional Chinese medicine-protein transition probability matrix, and obtaining an embedded vector of the heterogeneous nodes according to the transition probability matrix;

clustering proteins according to Euclidean distances of the traditional Chinese medicines and the proteins calculated by the embedded vectors to obtain traditional Chinese medicine clustering combinations respectively taking the structural proteins, the non-structural proteins and the auxiliary proteins as centers;

and screening candidate traditional Chinese medicines in the traditional Chinese medicine cluster combination according to a traditional Chinese medicine database to obtain a recommended traditional Chinese medicine.

2. The method of claim 1, further comprising obtaining potential chinese traditional medicines with highest similarity to the recommended chinese traditional medicines and used for recommended treatment in the guideline based on cosine similarity between the recommended chinese traditional medicines and the guideline chinese traditional medicines.

3. The method of claim 1, wherein the binding affinity of the compound to the neo-coronin is calculated using a molecular docking method, and the edge connection between the compound and the neo-coronin target is established according to the determination of the binding affinity.

4. The method of claim 1, wherein the meta pathway comprises TCM-compound-TCM, TCM-compound-protein, protein-compound-TCM, protein-compound-protein.

5. The method for recommending a prescription of a chinese medicine based on neo-crown protein heterogeneous network clustering of claim 1, wherein after determining the meta path, a random walk method is employed to calculate the first order similarity and the second order similarity between heterogeneous nodes.

6. The method for recommending a prescription of chinese medicine based on neo-crown protein heterogeneous network clustering according to claim 1, wherein said chinese medicine-protein transition probability is:

wherein, s (v)_i,v_j| ρ) represents the node v under the meta-path ρ_iAnd node v_jIs the threshold value of the meta-path length.

7. The method for recommending traditional Chinese medicine prescriptions based on neocoronal protein heterogeneous network clustering of claim 1, wherein a structured network embedding model SDNE is used to obtain the embedded vectors of heterogeneous nodes.

8. A traditional Chinese medicine prescription recommendation system based on new crown protein heterogeneous network clustering is characterized by comprising:

the network construction module is configured to construct the connection of the candidate traditional Chinese medicine and the compound of the candidate traditional Chinese medicine, the connection of the compound and the new coronary protein target, and construct a traditional Chinese medicine-compound-protein heterogeneous network according to the connection of the compound and the new coronary protein target;

the similarity calculation module is configured to determine a meta-path of the traditional Chinese medicine-compound-protein heterogeneous network, calculate the similarity between heterogeneous nodes under the meta-path to obtain a traditional Chinese medicine-protein transition probability matrix, and obtain an embedded vector of the heterogeneous nodes according to the transition probability matrix;

a clustering module configured to cluster the proteins according to Euclidean distances of the Chinese medicines and the proteins calculated by the embedded vectors to obtain Chinese medicine clustering combinations respectively centered on the structural proteins, the non-structural proteins and the auxiliary proteins;

and the recommending module is configured to screen the candidate traditional Chinese medicines in the traditional Chinese medicine cluster combination according to the traditional Chinese medicine database to obtain the recommended traditional Chinese medicines.

9. An electronic device comprising a memory and a processor and computer instructions stored on the memory and executed on the processor, the computer instructions when executed by the processor performing the method of any of claims 1-7.

10. A computer-readable storage medium storing computer instructions which, when executed by a processor, perform the method of any one of claims 1 to 7.

Images