KR100463840B1 - An automatic template generation method for constructing protein interaction networks - Google Patents

Abstract

PURPOSE: A method for automatically generating a template for constructing a protein interaction network is provided to reduce a network construction expense by automatically generating the basic template for the interaction network of an object protein based on the interaction network of the proteins, which are present in other species, similar to the object protein. CONSTITUTION: The proteins of a function similar to the object protein are searched from many species(S202). The interaction network of the searched proteins is generated based on a previously defined interaction relation database(S204). A similarity relation between the proteins existing in the generated interaction networks is set(S206). The network template of the object protein is generated by integrating the protein nodes between the different networks setting the similarity relation(S207).

Description

Automatic template generation for building protein interaction networks {AN AUTOMATIC TEMPLATE GENERATION METHOD FOR CONSTRUCTING PROTEIN INTERACTION NETWORKS}

FIELD OF THE INVENTION The present invention relates to techniques for building protein interaction networks in bioinformatics, and in particular, interaction networks for particular proteins in one species to interaction with proteins of similar function in other species. The present invention relates to a method for automatically generating a template for building a protein interaction network suitable for building using a network.

In general, protein interaction networks are used as important information for revealing the function of the protein from an overall perspective. That is, the function of a particular protein that is not revealed in the protein interaction network can be inferred from other proteins that interact with the protein. In addition, the effects on the living body can be predicted by inhibiting or activating the function of a specific protein.

Therefore, this protein interaction network is used as important information for selecting a target protein in drug development. In other words, it is possible to search for proteins that interact with the protein to activate or inhibit the function of a specific protein, and to select a protein that has already been identified in detail, and to develop a substance that modulates the function. These substances can be developed into very valuable products. Therefore, in order to develop high value-added new drugs, it is necessary to systematically construct a large network of interaction proteins.

In the previous study, a network for expressing an interaction relationship for a specific protein was constructed by the following method.

First, the "east to hybrid" method uses the characteristics of a particular gene (Gal4 gene) to reveal the interaction between the two proteins. In other words, the DNA binding domain region of the Gal4 gene and the gene expressing the target protein are expressed in one vector. In other vectors, the expression domain region of the Gal4 gene and the gene of the posterior protein are inserted and expressed. In this case, when the target protein and the candidate protein have an interaction relationship, the target protein binds to the DNA and the expression site of the candidate protein expresses the report gene of the DNA. However, this method has a problem that a lot of cost is required because the experiment must be performed on all of the vast proteins.

Second, microarray data analysis is a technique using expression patterns of genes under specific conditions. In other words, if the expression patterns of the two genes are dependent on each other, it is assumed that the two genes interact. For example, when the expression level of gene G1 increases, it may be said that G1 and G2 have an interaction relationship that activates or inhibits each other when the expression level of another gene G2 increases or decreases. However, this method does not interact under certain conditions, but when the expression patterns of the two genes are related to each other, there is a problem that the two proteins cannot be defined as precisely interacting.

Third, the protein structure analysis method is a technique of extracting a motif from the protein sequence and expressing this site in three dimensions. In addition, by analyzing the binding relationship and three-dimensional structure between the existing motifs to determine whether the two proteins interact. Binding of motifs is highly dependent on the three-dimensional structure of the protein. However, there is a problem that it is very difficult to technically predict the three-dimensional structure of the protein.

Fourth, the text mining method uses a variety of linguistic processing techniques to construct a database by extracting the parts describing the interaction relationship between proteins in the bio literature. However, this method is used to build a database by extracting interactions between previously discovered proteins, rather than revealing new protein interactions.

The present invention has been made to solve the problems of the prior art, the network construction by automatically generating a basic template for the interaction network of the target protein based on the interaction network of proteins that exist in different species and similar to the target protein The aim is to provide a method for automatically generating templates for constructing protein interaction networks to reduce costs.

In order to achieve this object, the present invention provides a method for automatically generating a template for constructing a protein interaction network, the method comprising: searching for proteins of a function similar to a target protein in different species; Generating an interaction network of the retrieved proteins based on the previously known interaction relationship database; Establishing a similar relationship between the proteins present between the generated interaction networks; The present invention provides a method for automatically generating a template for building a protein interaction network, which includes generating protein templates for a target protein by integrating protein nodes between different networks having similar relationships.

1 is a hardware diagram of an automatic template generation system for building protein interaction networks according to the present invention;

2 is an overall flow chart for template automatic generation according to a preferred embodiment of the present invention;

FIG. 3 is a detailed process of FIG. 2, which is a flowchart of a process of searching for a similar protein existing in another species with similar function to a target protein

FIG. 4 is a detailed process of FIG. 2, which is a flowchart of a process of generating an interaction network for the retrieved proteins from an already established interaction relationship database;

FIG. 5 is a detailed process of FIG. 2, which is a flowchart of a process of establishing similar relationships among similar protein nodes existing between generated networks;

FIG. 6 is a detailed process of FIG. 2 and illustrates a process of integrating networks having similar relationships into one network template. FIG.

<Explanation of symbols for the main parts of the drawings>

100: main memory 101: central processing unit

102: input / output unit 103: protein DB

104: interaction relationship DB 105: template generation unit

106: system bus

Hereinafter, with reference to the accompanying drawings will be described in detail a preferred embodiment of the present invention.

The invention automatically generates a basic template of an undisclosed interaction network for a particular species of protein from a database of protein interaction relationships already known, which is characterized by the interaction networks for large proteins. By presenting a template that can be incrementally deployed from an overall perspective, we propose a way to significantly reduce the cost of building an interactive network. To this end, a method for automatically generating a basic template of an interaction network for an unknown target protein is provided by using a database of known protein interaction relationships.

According to the present invention, using the interaction relationship database of a large number of proteins present in several species, it is characterized by generating a basic template of the interaction network for target proteins of a specific species.

To this end, the present invention provides a method for searching for similar proteins of other species that perform functions similar to the target protein.

It also provides a method for generating an interaction network for heterologous similar proteins retrieved from an already established interaction relationship database.

It also provides a method of establishing similar relationships between protein nodes in different networks.

It also provides a method for generating a template for a target protein by integrating similarly established networks into one network.

1 is a hardware configuration diagram of a template automatic generation system for building a protein interaction network according to the present invention. The main memory 100, the central processing unit 101, the input / output unit 102, the protein DB 103, It consists of the interaction relationship DB 104, the template generation part 105, and the system bus 106. FIG.

As shown in FIG. 1, the main memory 100 is loaded with template generation system information according to the present invention and information of the protein DB 103 and the interaction relationship DB 104 required in each step. The protein DB 103 information may be, for example, "SWISS-PROT", and the interaction relationship DB 104 information may be, for example, "KEGG or INTERACT".

The central processing unit 101 executes the template generation system information mounted in the main memory 100 step by step, and the input / output unit 102 receives the necessary information from the system from the user and automatically generates the network template generated by the system. Prints the relevant information on the screen. At this time, messages or information between the devices are transmitted / received through the system bus 106.

2 is a flowchart of a process of automatically generating a basic template of an interaction network for a target protein present in one species as a preferred embodiment of the present invention.

First, in step S201, a target protein and a name of a species for which a network is required to be input are input, and the process proceeds to step S202, in which a similar protein existing in another species and having a similar function to the input target protein DB 103 Search from). The species name, protein name and protein sequence are then used.

In step S204, a network is generated from the interaction relationship DB 104 for similar proteins of the different species searched. Here, interaction relationship information for the proteins verified through various bioinformatics techniques and bio experiments are recorded in the interaction relationship DB 104.

In step S206, a similar relationship is established between protein nodes existing in one network and protein nodes in another network.

In operation S207, an interaction network template for the target protein is generated by integrating into a virtual network node through a similar relationship between protein nodes existing between networks.

Hereinafter, a detailed step-by-step process of the method for automatically generating a template for constructing a protein interaction network according to the present invention will be described in more detail with reference to FIGS. 3 to 6.

(A) Method of searching for similar protein of different species

There are many ways to search for proteins of another species that perform functions very similar to that of one species of protein. The heterologous similar protein search method in the present invention makes it possible to search from a database for proteins of other species that perform functions very similar to the target protein using the protein's name and sequence. Figure 3 illustrates the step-by-step process of searching for such heterologous proteins.

As shown in FIG. 3, in step S301, a target protein name such as P.name and a species name such as P.organism are input. Here, it is assumed that the interaction network of the target protein is not known. In general, the same protein names are often the same even though they are different species. Typically, the protein AGRN (agrin) is expressed by the same name in mice and humans.

Therefore, in step S302, a protein P _i of the same name but a different species from the target protein P is searched from the protein DB 103. That is, P _i .name = P.name and P _i .organism Search for all proteins P _{i = 1 .., n} that are P. organism.

However, since there is no standardized method for naming proteins, the names may be labeled differently if proteins with functions very similar to the target protein exist in different species. For example, human precursor cell expressed, developmentally down-regulated 5 of human (Homo sapiens) and protein SEPT2 (septin 5) of musculus, although different in name, are found to have exactly the same function. Can be. By the way, two proteins with very similar functions generally have almost identical protein sequences. These two proteins also have the same sequence because they have the same function in vivo.

Therefore, in step S303, proteins of other species having a different name from the target protein P but having a very identical sequence are searched for in the protein DB 103. That is, P _j .seq P.seq and P _j .organism Search for all proteins P _{j = 1 .., m} that are P. organism. The similarity test of the two sequences uses a local alignment algorithm using a similar matrix such as BLAST. At this time, the sequence similarity of the two proteins is calculated to a value between 0 and 1, and when the similarity is 0.9 or more, it is generally found that the two proteins perform very similar functions.

In step S304, when several proteins similar to the target protein are detected in one species, similar proteins are selected according to the following rules. First, if several proteins for a species are searched by name and by sequence, only those proteins searched by name are selected. Second, if a protein is not searched by name for one species but several proteins are searched by sequence, a protein having the sequence that most closely matches the target protein is selected as a similar protein. Third, since the next step is to create an interaction network for similar proteins, only those proteins whose interaction networks have already been identified are selected.

(B) how to create a network from interaction relationships;

Heterologous proteins similar to the target protein are already identified, in part or in whole, with the interaction network and stored in the interaction relationship DB 104. In the present invention, the relations related to heterologous similar proteins with respect to the target protein retrieved in the previous process are extracted, and an interaction network for the similar proteins is generated through a wide-first search method. At this time, one queue in charge of the first-in, first-out function is used for the breadth-first search. 4 illustrates a method of generating an interaction network formed from one protein.

As shown in FIG. 4, in step S401, one protein p of heterologous similar proteins detected in step S202 of FIG. 2 is input. A network for this protein is generated from the interaction relationship DB 103.

Therefore, in step S402, a visit flag is set to P in order not to visit P repeatedly.

In step S403, proteins P _{i = 1, .., n} having a direct interaction with P are extracted from the interaction relationship DB 103.

In step S404, a network is formed by gradually connecting the interaction relationship iLink between P and the retrieved protein P _i .

In step S405, all proteins P _{k, 1} not visited among proteins P _{i = 1, .., n k} Insert all _n into the queue. By inserting these proteins into the queue, they are used for breadth-first search.

In step S406, the visit flag is set so as not to visit the protein P _k inserted into the queue again.

In step S407, if the queue is empty, the process ends. If not, the protein P of the queue is extracted in step S408, and step S407 is repeated from step S403. When this process is finished, it is possible to generate an interaction network for the protein input in step S401.

Therefore, all proteins retrieved in step S202 of FIG. 2 generate an interaction network through this process.

(C) establishing interactive network affinity

The template for the interaction network of the target protein is expressed in an integrated form that can encompass the interaction networks for heterologous proteins similar to the target protein. Therefore, in FIG. 5, a method of establishing a similar relationship "hLink" of protein nodes existing between interaction networks of similar proteins with respect to the target protein generated in FIG. 4 will be described.

As shown in FIG. 5, in step S501, ⁿ generated from similar proteins P ⁽¹⁾ , P ⁽²⁾ , ..., P ⁽ⁿ⁾ for the target protein in step S204 of FIG. N interaction networks N ⁽¹⁾ , N ⁽²⁾ , ...., N ⁽ⁿ⁾ are input.

In step S502, two networks N ^(s) and N ^{(k) of these} are selected to establish a similar relationship between proteins present in the two networks. Therefore, a total of n (n-1) / 2 combinations is generated to establish a similar relationship between all n networks. Step S503 and step S504 are performed for each of these combinations.

First, in step S503, a similar relationship hLink between nodes P ^(s) _i and P ^(k) _j having the same name is set for all the protein nodes of the two heterogeneous networks. At this time, the similarity of hLink is set to 1.0.

In step S504, hLink is set on a sequence basis for all protein nodes that are not set on a name basis. At this time, the similarity relationship hLink of two protein nodes is established only when the sequence similarity is 0.9 or more, and the similarity is determined by the similarity of the sequence.

(D) Create interactive network templates

In this process, we generate networks of target proteins by inputting networks of heterologous proteins similar to target proteins and integrating similar protein nodes with similarity hLink into one virtual node. Thus, in Figure 6, we describe how to create a template network for the target protein by navigating the network interaction relationship iLink in a breadth-first search method and integrating similar protein nodes with the similarity hLink into one virtual protein node. Shall be.

As shown in FIG. 6, in step S601, networks N ⁽¹⁾ , N ⁽²⁾ , ..., N ^{(n) in} which affinity hLink is set between heterologous protein nodes are input.

In step S602, the network template NT is initially initialized to the network N ⁽¹⁾ .

Step (S603) and the network N ⁽²⁾ repeat the process from step to (S610) in the step (S604) for the N ⁽ⁿ⁾ incorporating a network NT as a template.

First, in S604, the initial visited protein node p is set at N ⁽ⁱ⁾ for the breadth-first search.

In step S605, all protein nodes P _j = 1, .., s having an interaction relationship iLink with P in N ⁽ⁱ⁾ are extracted.

In step S606, a queue is inserted to search only the first visited node among the protein nodes extracted in step S605, and a visit is displayed in step S607.

In step S608, it is determined whether the current protein node P has a similarity hLink with one protein node P _NT of the network template. If not, otherwise proceed to step S610.

In step (S609) and integrating information of the two nodes P and P proteins as a protein _{NT NT} P node. That is, all the interaction relationship information associated with P is changed into the relationship with P _NT . Also, the node name and sequence information of P are added to the node name and sequence information of P _NT , and then the P node is deleted.

In step S610, it is determined whether the queue is empty. If the queue is not empty, the process proceeds to step S611 and one protein node of the queue is extracted until the queue is empty. That is, step S605 is repeated from step S605.

If it is determined in step S610 that the queue is empty, the process proceeds to step S612 and the generated network template NT is output.

According to the present invention, a basic template of an undisclosed interaction network for a specific protein can be automatically generated using an interaction relationship database for previously known heterologous similar proteins. Therefore, it is possible to significantly reduce the network construction cost by allowing the user to gradually build an interactive network, which is important information of the target protein, from an overall perspective.

As mentioned above, although this invention was demonstrated concretely based on the Example, this invention is not limited to this Example, Of course, various changes are possible within the range which does not deviate from the summary of the claim mentioned later.

Claims (5)

In a method for automatically generating a template for building a protein interaction network using an interaction relationship between a plurality of proteins, automatically generating an unknown interaction network template for a specific target protein, Searching for a protein having a function similar to the target protein in a plurality of species; A second step of generating an interaction network of the proteins retrieved in the first step based on the previously known interaction relationship database; Establishing a similar relationship between the proteins present between the interaction networks created in the second step; And a fourth step of generating a network template for the target protein by integrating protein nodes between different networks established by the third step in a similar relationship. The method of claim 1, The first step is, Searching for any protein of the same name and different species as the target protein when the name of the target protein and the species name are input; Retrieving from the protein DB other proteins of different species with the same name and sequence as the target protein; As a result of the search, a method for automatically generating a template for building a protein interaction network, characterized in that the step of selecting a protein similar to the target protein. The method of claim 1, The second step, When a protein of one of the heterologous similar proteins found in the first step is input, generating a network for the input protein from the interaction relationship DB; Extracting a protein having a direct interaction with the input protein from the interaction relationship DB; Gradually connecting the interaction relationship between the input protein and the protein having the interaction relationship to generate a network; Inserting all non-visited proteins among the proteins with the interaction relationship into a queue and setting a visit flag; Repeating the protein extraction process until the cue is empty, thereby generating an interaction network for the input protein. The method of claim 1, The third step, When the n interaction network information generated from the analogous protein for the target protein is input, selecting an arbitrary number of network information among the input network information to establish a similar relationship between the proteins existing in the selected network. Wow; Establishing a similar relationship between nodes of the same name for all protein nodes of the selected number of heterogeneous networks; A method of automatically generating a template for constructing a protein interaction network, comprising: establishing sequence-based similarity for all protein nodes not set based on name. The method of claim 1, The fourth step, Initializing a network template when network information in which similar relationships are established between heterologous protein nodes is input; Establishing an initial visited protein node for breadth first search; Extracting all protein nodes having an interaction relationship with the protein nodes in the network; Inserting a first visited node among the extracted protein nodes into a queue and displaying a visit; Determining whether the current protein node has a similar relationship with one protein node of the network template, and integrating the two protein nodes if determined to have a similar relationship; If it is determined that the current protein node does not have a similar relationship with one protein node of the network template, it is determined whether the queue is empty, and if the queue is not empty, extracting one protein node of the queue until the queue is empty; ; And, if it is determined that the queue is empty, outputting the generated network template.