JP4280831B2 - Compound group display device, compound group display method, program, and computer-readable recording medium - Google Patents

Description

  The present invention relates to a compound group display device and a compound group display method for classifying and displaying a plurality of compounds, a program for realizing the device, and a computer-readable recording medium on which the program is recorded. is there.

  In the fields of pharmaceuticals, agricultural chemicals, cosmetics, and other chemical industries, research is constantly being conducted to search for compounds having excellent functional properties such as activity and physical properties. Recently, for example, a combination of combinatorial chemistry technology and high-throughput screening methods is frequently used. As a result, a large number of compounds can be synthesized and evaluated simultaneously and rapidly, and large-scale screening data relating to the functional properties of the compounds can be accumulated.

  In order to synthesize and develop a new compound having further excellent functional characteristics based on the screening data accumulated in this way, it is essential to arrange the obtained screening data in association with the structure of the compound.

  In order to sort out the relationship between the structure of the compound and the screening data, first, a method for expressing the structure of the compound quantitatively is necessary. A number of molecular descriptors have been devised so far. Typical examples include structural keys (Daylight Chemical Infomation Systems, MDL Information Systems), Molconn-Z descriptors (Hall Associate Consulting), ClogP (Biobyte), ACD / LogD (ACD Labs), etc. Can be mentioned.

  On the other hand, many methods have been proposed for managing screening data using these molecular descriptors. The method is roughly divided into the following two.

  One is a method of performing multivariate analysis using screening data as an objective variable and molecular descriptor as an explanatory variable, and managing them as a model expression that quantitatively represents the relationship (for example, Non-Patent Document 1). This makes it possible to predict the functional characteristics of unknown compounds by calculating molecular descriptors from the chemical structure and applying them to model equations.

  Another method is a method of constructing a database in which molecular descriptors and screening data (functional properties) of each compound are associated (for example, Non-Patent Document 2). If structurally similar, functional characteristics are expected to be similar, so if you search a database for compounds that are structurally similar to unknown compounds, you can predict the functional characteristics of unknown compounds based on the screening data of the searched compounds It becomes possible to do. In this case, the similarity / dissimilarity evaluation is performed based on the molecular descriptor.

  Each of the above methods is based on a statistical or stochastic approach, and empirically predicts the relationship between molecular descriptors and functional properties in a compound. Therefore, in these methods, the validity of the calculation result is often asked as compared with the theoretical prediction based on the mechanism theory, and an expression method that can objectively evaluate the validity is desired. Data visualization is a very effective method for solving this problem. By visualizing the data, it is possible to make appropriate judgments by utilizing the advanced analytical capabilities cultivated by humans through experience. In addition, since visualization technology makes it easy to share large-scale information resources, it is possible to unify intentions in project research and promote projects efficiently.

Conventionally, as a method of visualizing the functional properties of a compound in association with structural information, (a) a single molecular descriptor or (b) an overall characteristic value based on a plurality of molecular descriptors is used as a three-dimensional or lower axis. A method of mapping a compound in a characteristic space has been employed (for example, Non-Patent Document 3 and Patent Document 1). Thereby, it is possible to display the functional properties and the structural information in association with each other for the compound group listed in the database. If there is an unknown compound whose characteristics are to be evaluated, the unknown compounds are visually evaluated from the functional characteristics of the compounds stored in the database by plotting the unknown compounds on the characteristic space based on the structural information. can do.
JP2002-53894 (Released on September 24, 2002) C. Hansch et al., Chem-bioinformatics: comparative QSAR at the interface between chemistry and biology. Chem. Rev., 102: 783-812, 2002 RP Sheridan and SK Kearsley, Why do we need so many chemical similarity search methods? Drug Discovery Today, 7: 903-911, 2002 Y. Takahashi et al., MolSpace: a computer desktop tool for visualization of massive molecular data.J. Mol. Graph Model., 21: 333-339, 2003

  However, the conventional method has the following problems.

  First, in the conventional method, since each coordinate axis in the characteristic space represents an interval scale, the arrangement of the plotted compounds becomes discrete or unevenly distributed. As a result, there is a possibility that a blank portion where the compound does not exist is formed in a part of the display area of the display, or conversely, the compound is concentrated in a part and an overlapping portion is formed. Thus, the conventional method cannot effectively use the display area of the display when plotting the compound in the characteristic space.

  Further, in the display area of the display, it is the limit to display a three-dimensional characteristic space, and a four-dimensional or higher characteristic space cannot be displayed. This means that four or more pieces of structure information cannot be handled simultaneously as the structure information of the compound. Therefore, it is necessary to reduce the structural information to 3 or less in some form, and at this time, a part of the structural information of the compound is lost. As described above, the conventional method cannot comprehensively evaluate the relationship between the structural characteristics and the functional characteristics composed of a large amount of structural information for a compound.

  The present invention has been made in view of the above problems, and an object of the present invention is to efficiently display a plurality of compounds by associating a first characteristic and a second characteristic such as a structural characteristic and a functional characteristic. An object of the present invention is to provide a compound group display device capable of simultaneously handling a large amount of characteristic information as a first characteristic.

  In order to solve the above-described problem, the compound group display device according to the present invention provides, for a plurality of compounds, first characteristic information representing the first characteristic of each compound and second characteristic information representing the second characteristic. The compound group display device that displays the relationship between the first characteristic and the second characteristic based on the above, wherein the compound is classified into a plurality of hierarchical clusters based on the first characteristic information A classifying unit, a compound arranging unit that determines the arrangement of the icon of the compound in a two-dimensional plane so as to have a recursive nested structure based on the hierarchical cluster classified by the hierarchical classifying unit, and the compound Image data generating means for generating image data in which the icon of the compound is drawn with the color, pattern, and / or shape based on the second characteristic information in the arrangement determined by the arrangement means; Characterized in that it comprises a display unit for displaying an image based on image data generated by the serial image data generating means.

  In addition, as a 1st characteristic, the structural characteristic etc. of a compound are mentioned, for example, As a 2nd characteristic, the functional characteristic etc. of a compound are mentioned, for example.

  According to the above configuration, the hierarchical classifying unit classifies the compounds into hierarchical clusters based on the first characteristic information. Then, the compound placement unit and the image data generation unit display the compound in a two-dimensional plane so as to have a nested structure based on the classified clusters. Thereby, each compound expresses the first characteristic by the nested structure including the icon, and expresses the second characteristic by the color, pattern, and / or shape of the icon. As described above, the compound group display device of the present invention can display a compound group in association with the first characteristic and the second characteristic. Note that the above colors include monochrome and gray scale.

  Here, when classifying into hierarchical clusters, the number of variables included in the first characteristic information may be plural. That is, even if the first characteristic information is composed of multidimensional vectors, the compounds can be classified into hierarchical clusters. Therefore, the compound group display device can simultaneously handle a large amount of characteristic information as the first characteristic.

  In addition, the compound is not represented by a point in the characteristic space with the first characteristic information as a coordinate axis, but is represented by an icon in a recursive nested structure based on the first characteristic information. Here, the position of each icon is not determined at a single point based on the numerical value of the first characteristic information, but the relative positional relationship is roughly such that icons of similar compounds exist in the vicinity. To be determined. Therefore, it is possible to arrange the icons so as to be as uniform as possible so that the icons are not discretely arranged or unevenly distributed. Thus, the display area of the display can be used effectively.

  The first characteristic information is composed of n (where n is a natural number) quantitative variables, and the hierarchical classifying means uses the compound as a coordinate axis in an n-dimensional space with each quantitative variable as a coordinate axis. You may classify the said compound based on the distance between each compound or each cluster, treating as a point.

  According to the above configuration, the compounds can be classified using the distance calculated from the coordinates of the compound and the center of gravity of the cluster as a measure of dissimilarity.

  The hierarchy classification means may classify the compound by creating a decision tree using the first characteristic information as an explanatory variable and the second characteristic information as an objective variable.

  According to the above configuration, the second characteristic information tends to be uniform among the compounds in the cluster classified by creating the decision tree based on the first characteristic information. Accordingly, the correspondence relationship between the first characteristic and the second characteristic is clarified, and what kind of second characteristic the unknown compound has is easily estimated based on the first characteristic information of the compound. become able to. Note that the decision tree includes a regression tree.

  The compound group display device further includes non-hierarchical classification means for classifying the compound into a non-hierarchical cluster based on the second characteristic information, and the hierarchical classification means uses the first characteristic information as an explanatory variable. The compound may be classified by creating a decision tree using the categorical variable corresponding to the cluster formed by the non-hierarchical classification means as an objective variable.

  According to the above configuration, even if the second characteristic information is composed of a plurality of variables, the non-hierarchical classification means classifies the compounds into non-hierarchical clusters based on the second characteristic information. The second characteristic information possessed can be handled as a categorical variable that can take the same number of values as the number of clusters. This makes it possible to clarify the second property of the compound in each cluster of the decision tree.

  The compound group display device further includes non-hierarchical classification means for classifying the compounds into non-hierarchical clusters based on second characteristic information, and the image data generating means is formed by the non-hierarchical classification means. Image data in which an icon of the compound is drawn by a color, pattern, and / or shape corresponding to each cluster may be generated.

  According to the above configuration, even if the second characteristic information includes a plurality of variables, the non-hierarchical classification unit classifies the compounds into non-hierarchical clusters based on the second characteristic information. The second characteristic possessed by can be expressed by a color, pattern, and / or shape based on the classified cluster.

  Further, the second characteristic information is composed of a plurality of quantitative variables, and the non-hierarchical classification means converts the compound into a non-hierarchical structure by a self-organizing map method using the second characteristic information as a pattern vector. May be classified into different clusters.

  According to the self-organizing map method, data represented by a multi-dimensional vector can be mapped to a two-dimensional map while maintaining its correlation with other data while leaving its features. On this two-dimensional map, nodes having similar multidimensional data are arranged close to each other, and the relationship between the data can be easily understood visually.

  The first characteristic information includes (a) one or more molecular descriptors representing a structural fragment or structural topology of the molecular structure of the compound; (b) 1 corresponding to a physicochemical property obtained by calculation or experiment. It may contain at least one of two or more molecular descriptors.

  The second property information may be a parameter of at least one of the biological activity, toxicity, physicochemical property, and pharmacokinetic property of the compound.

  The second characteristic information may be a categorical variable.

  By the way, each means in the said compound group display apparatus may be implement | achieved by hardware, and may be implement | achieved by making a computer run a program. Specifically, a program according to the present invention is a program that causes a computer to operate as each of the above-described means, and a recording medium according to the present invention records the program.

When these programs are executed by a computer, the computer operates as each unit of the compound group display device. Therefore, the same effect as the above compound group display device can be obtained. In order to solve the above problem, the compound group display method according to the present invention provides a first property that represents the first characteristic of each compound for a plurality of compounds. A compound group display method in a compound group display device that displays a relationship between a first characteristic and a second characteristic based on characteristic information and second characteristic information representing a second characteristic, the compound group A display device includes a hierarchy classification unit, a compound arrangement unit, an image data generation unit, and a display unit, and the hierarchy classification unit classifies into a plurality of hierarchical clusters based on the first characteristic information; The arrangement of the icons of the compounds in the two-dimensional plane is determined so that the compound arranging means has a recursive nested structure based on the hierarchical cluster classified in the hierarchical classification step. Image data in which the icon of the compound is drawn with the color, pattern, and / or shape based on the second characteristic information in the arrangement determined by the compound arrangement step and the image data generating means determined in the compound arrangement step The image data generating step for generating the image data and the display unit for displaying an image based on the image data generated in the image data generating step.

  In addition, as a 1st characteristic, the structural characteristic etc. of a compound are mentioned, for example, As a 2nd characteristic, the functional characteristic etc. of a compound are mentioned, for example.

  According to the above configuration, the compounds are classified into hierarchical clusters based on the first characteristic information in the hierarchical classification step. In the compound placement step and the image data generation step, the compound is displayed in a two-dimensional plane so as to have a nested structure based on the classified clusters. Thereby, each compound expresses the first characteristic by the nested structure including the icon, and expresses the second characteristic by the color, pattern, and / or shape of the icon. As described above, the compound group display method of the present invention can display a compound group in association with the first characteristic and the second characteristic. Note that the above colors include monochrome and gray scale.

  Here, when classifying into hierarchical clusters, the number of variables included in the first characteristic information may be plural. That is, even if the first characteristic information is composed of multidimensional vectors, the compounds can be classified into hierarchical clusters. Therefore, this method can simultaneously handle four or more characteristics as the first characteristics.

  In addition, the compound is not represented by a point in the characteristic space with the first characteristic information as a coordinate axis, but is represented by an icon in a recursive nested structure based on the first characteristic information. Here, the position of each icon is not determined at a single point based on the numerical value of the first characteristic information, but the relative positional relationship is roughly such that icons of similar compounds exist in the vicinity. To be determined. Therefore, it is possible to arrange the icons so as to be as uniform as possible so that the icons are not discretely arranged or unevenly distributed. Thus, the display area of the display can be used effectively.

  As described above, the compound group display device according to the present invention includes a hierarchical classification unit that classifies a compound into a plurality of hierarchical clusters based on the first characteristic information, and a hierarchical cluster classified by the hierarchical classification unit. A compound arrangement means for determining the arrangement of the icon of the compound in the two-dimensional plane so as to have a recursive nested structure based on the color, and the color based on the second characteristic information in the arrangement determined by the compound arrangement means A configuration comprising: image data generating means for generating image data in which an icon of a compound is drawn according to a pattern, and / or shape; and a display unit for displaying an image based on the image data generated by the image data generating means It has become.

  Further, the compound group display method according to the present invention includes a hierarchical classification step in which the hierarchical classification means classifies into a plurality of hierarchical clusters based on the first characteristic information, and a hierarchy in which the compound placement means is classified in the hierarchical classification step. A compound arrangement step for determining the arrangement of the icon of the compound in the two-dimensional plane so as to form a recursive nested structure based on a typical cluster, and an arrangement in which the image data generating means is determined in the compound arrangement step. An image data generation step for generating image data in which the icon of the compound is drawn with a color, pattern, and / or shape based on the characteristic information of 2, and a display unit based on the image data generated in the image data generation step And a display process for displaying the image.

  Therefore, as described above, it is possible to efficiently display the first characteristic and the second characteristic in association with each other for a plurality of compounds, and it is also possible to simultaneously handle a large amount of characteristic information as the first characteristic. Play.

Embodiment 1
An embodiment of the present invention will be described below with reference to FIGS. The compound group display device 1 of the present embodiment displays the first characteristic and the second characteristic in association with each other for a plurality of compounds. Although the type of the first characteristic is not particularly limited, in the present embodiment, a structural characteristic is used as an example, and a molecular descriptor is used as structural information representing this. The structural information may be a single molecular descriptor or may be expressed as an n-dimensional vector having n molecular descriptors, that is, values of each molecular descriptor as components. Further, the molecular descriptor may represent a structural fragment or structural topology of the molecular structure of the compound, or may correspond to a physicochemical property obtained by calculation or experiment, A combination of these may also be used.

  The second property is not particularly limited as in the case of the first property information. For example, functional properties such as biological activity, toxicity, physicochemical properties, or pharmacokinetic properties of the compound, etc. And various parameters can be used as the functional characteristic information representing this. Further, these parameters may be appropriately combined. In this specification, these parameters are hereinafter referred to as “functional characteristic information”.

  In the present embodiment, the functional characteristic information may be a quantitative variable or a categorical variable represented by a one-dimensional vector. The quantitative variable is a variable measured on an interval scale such as an activity value of a compound, for example, and may be a discrete variable, and the categorical variable is, for example, a type of activity (A activity, A variable that can take a plurality of values, such as B activity or C activity.

  With such a configuration, the compound group display device of the present embodiment can display the structure and functional characteristics of the compound in association with each other. Therefore, when the user wants to know the functional characteristics of an unknown compound, the functional characteristics of the unknown compound can be determined from the functional characteristic information of the compound having a similar structure by knowing where the unknown compound is displayed from the structural information. Can be guessed. It is also possible to visually determine how the structure and functional properties are related from the displayed distribution of known compounds.

  FIG. 1 is a functional block diagram of the compound group display device 1 of the present embodiment. As shown in FIG. 1, the compound group display device 1 includes a descriptor input unit 11, a functional property input unit 12, a hierarchy classification unit (hierarchy classification unit) 13, a compound arrangement unit (compound arrangement unit) 14, and an image data generation unit. (Image data generating means) 15 and a display unit 16 are provided.

  The descriptor input unit (first input unit) 11 is for acquiring a molecular descriptor (first characteristic information) of a compound to be displayed from a user or another device. The hardware configuration of the descriptor input unit 11 may be one that accepts input from the user, such as a keyboard, mouse, touch panel, or tablet, or an external storage device such as various input / output interfaces. Or an input from another computer.

  The functional property input unit (second input unit) 12 is for acquiring functional property information (second property information) of a compound to be displayed from a user or another device. The hardware configuration of the functional characteristic input unit 12 is the same as that of the descriptor input unit 11.

  The hierarchical classification unit 13 is for classifying compounds into a plurality of hierarchical clusters based on molecular descriptors (first characteristic information). More specifically, the compounds are hierarchically classified by various algorithms based on the molecular descriptor.

  The compound arrangement unit 14 determines the arrangement of the compounds in the two-dimensional plane so as to have a recursive nested structure based on the hierarchical cluster classified by the hierarchy classification unit 13. More specifically, the compound placement unit 14 places each compound in a two-dimensional plane so that the hierarchically classified compounds are leaf nodes and the child nodes are sequentially included in the parent node. To do. As an algorithm for this arrangement, for example, Heiankyo View ("Heiankyo View-Visualization Method for Arranging Hierarchical Data on a Board", Takayuki Ito, Koji Koyamada, Visualization Society of Japan 9th Visualization Conference Abstract, published in 2003). In Heiankyo view, leaf nodes in the hierarchical data are first arranged in a grid. Subsequently, a large amount of information is expressed in a limited screen space by efficiently arranging branch nodes corresponding to the leaf nodes on the screen.

  The image data generation unit 15 generates image data in which the icon of the compound is drawn with the color based on the second characteristic information in the arrangement determined by the compound arrangement unit 14. More specifically, the image data generation unit 15 generates image data in which the compound is arranged at the position determined by the compound arrangement unit 14. At this time, each compound corresponding to the leaf node may be expressed by a simple figure such as a circle, a square, or a point, or may be expressed by a symbol that indicates the type of the compound. Hereinafter, a graphic or symbol representing a leaf node is referred to as an icon. At this time, the icons corresponding to the respective compounds (leaf nodes) are color-coded based on the second characteristic information. Further, the image data generation unit 15 sequentially encloses the nodes in a frame such as surrounding a parent node to which each compound that is a leaf node belongs, enclosing a further parent node to which each node enclosed in the frame belongs, and so on. Enclose with. Thereby, it is expressed to which node of what hierarchical structure each compound which is a leaf node belongs.

  Note that the color used for the above color classification does not necessarily have to be a color, and may be a monotone shade (gray scale). Further, instead of expressing the functional characteristics of the compound by color, it may be expressed by a pattern such as shading or the shape of an icon. Of course, these may be used in appropriate combination.

  The display unit 16 is for displaying an image based on the image data generated by the image data generation unit 15. The hardware configuration of the display unit 16 is not particularly limited as long as an image can be displayed in color. For example, the display unit 16 includes various displays such as a liquid crystal display capable of color display and a CRT (Cathode Ray Tube) display. Alternatively, a color printer may be provided. Thereby, the display unit 16 can display an image represented by the image data generated by the image data generation unit 15.

  The hierarchy classification unit 13, the compound arrangement unit 14, and the image data generation unit 15 are functions realized by an arithmetic device such as a CPU executing a program code stored in a storage device such as a ROM or a RAM. It is a block.

  Next, the operation of the compound group display device of this embodiment will be described. FIG. 2 is a process diagram of the compound group display method of the present embodiment.

  First, in step S10, the functional characteristic input unit 12 acquires functional characteristic information. In the present embodiment, the function characteristic information to be acquired may be either a quantitative variable or a categorical variable represented by a one-dimensional vector. Table 1 is a table showing an example of the acquired functional characteristic information.

  The acquired function characteristic information is transmitted to the image data generation unit 15.

  Next, in step S11, the descriptor input unit 11 acquires a molecular descriptor. Although only one type of molecular descriptor may be acquired, n types are used in the present embodiment. These molecular descriptors are associated with compounds. Table 2 is a table showing an example of the acquired molecular descriptor.

  In Table 2, there are three types of molecular descriptors associated with each compound, but the present invention is not limited to this, and an arbitrary number of molecular descriptors can be handled.

  Then, the acquired molecular descriptor is transmitted to the hierarchy classification unit 13.

  Next, in step S12, the hierarchical classification unit 13 classifies the compounds into hierarchical clusters based on the received molecular descriptor. In this step, unless otherwise specified, it is assumed that the hierarchy classification unit 13 performs each process.

  When the explanatory variable is a quantitative variable such as the molecular descriptor described above, the hierarchical classification unit 13 treats each compound as a point in the n-dimensional space. That is, in the case of Table 1, since there are three types of molecular descriptors, Compound 1 is a point (0.5, 2, 0) in the three-dimensional space. Similarly, the compound 2 becomes a point (0.9, 4, 1). At this time, standardization processing may be performed on the molecular descriptor. Specifically, for example, standardization is performed such that the average value between compounds of each molecular descriptor is 0 and the standard deviation value is 1.

  In addition, when the structure information is composed of a multidimensional vector having a plurality of molecular descriptors as elements, a principal component analysis may be performed on the structure information to reduce the dimension (number of variables) of the structure information.

  And the dissimilarity between compounds is computed from the distance between compounds in the space where each compound is arranged in this way. That is, as the distance between the compounds is longer, the degree of similarity between the compounds is considered to be smaller, so that they are not classified into the same cluster. As the distance for calculating the dissimilarity, for example, a suitable distance among the Euclidean distance, the Manhattan distance, the power distance, and the like can be used.

  As a modification, the similarity may be calculated instead of the dissimilarity based on the distance. As a method for calculating the similarity, for example, a Pearson correlation coefficient, a Tanimoto coefficient, a cosine function, or the like can be used.

  Moreover, as an algorithm for classifying into hierarchical clusters based on the distance, a known algorithm such as nearest neighbor method, group average method, or Ward method can be used.

  Through the above processing, for example, information on the dissimilarity (distance) and hierarchical structure between clusters that can represent a dendrogram as shown in FIG. 3 is obtained. Note that the hierarchical structure and distance in FIG. 3 are not accurate. Of the obtained information, hierarchical information is transmitted to the compound placement unit 14. Table 3 is an example of data indicating a hierarchical structure (hierarchical structure table).

  As shown in Table 3, in the hierarchical structure table, branch node IDs and IDs of child nodes and leaf nodes included in the branch nodes are stored in association with each other.

  Next, in step S13, the compound placement unit 14 places each compound (leaf node) in the two-dimensional plane based on the hierarchical structure expressed by the hierarchical structure table. At this time, each leaf node is set so that the leaf nodes included in the same branch node are close to each other and the interval between the adjacent leaf nodes is substantially constant regardless of the distance between the leaf nodes. Deploy.

  FIG. 4 is a diagram showing an example of the arrangement of compounds in a two-dimensional plane. In the figure, each leaf node is represented by a black square. The frame surrounding each leaf node is added so that the hierarchical structure can be visually recognized, and is not included in the arrangement data created by the compound arrangement unit 14. That is, the arrangement data generated by the compound arrangement unit 14 includes information on the compound (leaf node) and the display position of the compound (for example, the upper left coordinates in the square in the figure) in association with each other. The generated arrangement data is transmitted to the image data generation unit 15.

  In step S14, the image data generation unit 15 generates image data based on the received arrangement data. In the generated image data, each compound (leaf node) is surrounded by a frame indicating a branch node, and the branch node surrounded by the frame is further surrounded by a frame indicating its parent node. It has become. Further, the icon of each compound (leaf node) is color-coded based on the functional characteristic information acquired in step S10.

  For example, when the input functional characteristic information is a categorical variable, the image data generation unit 15 determines the color of the compound based on a table or the like in which each categorical variable is associated with a color. That is, the image data generation unit 15 determines the color of each compound so that compounds having the same categorical variable have the same color. FIG. 5 is an example of color coding when compound 1 and compound 2, compound 3 and compound 4 each have functional characteristic information composed of the same categorical variable.

  When the input functional characteristic information is a quantitative variable, the image data generation unit 15 converts the value of each functional characteristic information into color information (for example, any one color of R, G, and B by a conversion formula). The color of the compound may be determined so that the color of the icon changes continuously according to the value of the functional property information.

  The image data generated in this way is transmitted to the display unit 16. In step S14, the image data generation unit 15 may give a height to an icon indicating each compound (leaf node) arranged in the plane. For example, when each compound has the third characteristic information represented by a one-dimensional vector, the value of the third characteristic information is converted into a height, and the icon of each compound has a height. Drawn image data may be generated. Thereby, the compound group display device 1 can display the first to third characteristic information in association with each other.

  Finally, in step S15, the display unit 16 displays an image corresponding to the received image data. FIG. 5 is an example of display by the display unit 16. In the compound group display device 1, when each compound in the image displayed by the display unit 16 is designated with a pointer, it is preferable to display the molecular descriptor and functional property information of the compound.

[Embodiment 2]
Another embodiment of the present invention will be described with reference to FIGS. 6 and 7 as follows. In addition, about the member which has the same function as Embodiment 1 mentioned above, the same number as Embodiment 1 is attached, and description is abbreviate | omitted.

  The compound group display device 2 of the present embodiment displays the first characteristic information and the second characteristic information in association with each other for a plurality of compounds.

  The type of the first characteristic information is not particularly limited, and may be either a quantitative variable or a categorical variable, and the number of these variables may be single or plural. Also good. In the present embodiment, a plurality of molecular descriptors are used as in the first embodiment.

  Further, the second characteristic information is not particularly limited as in the case of the first characteristic information. For example, parameters such as the biological activity, toxicity, physicochemical properties, or pharmacokinetic properties of the compound are set. Can be used. Further, these parameters may be appropriately combined. In this specification, these parameters are hereinafter referred to as “functional characteristic information”.

  Note that the compound group display device 2 of the present embodiment creates a decision tree by using molecular descriptors as explanatory variables and functional property information as objective variables when classifying compounds into hierarchical clusters. Is different. As a result, the hierarchically classified compounds show uniform functional properties within the cluster. Therefore, the relationship between the structure and the functional characteristics is clarified, and the functional characteristics of the unknown compound can be easily estimated based on the molecular descriptor.

  Further, the feature characteristic information may be a quantitative variable represented by an n-dimensional vector (n ≧ 2) in addition to the quantitative variable and categorical variable represented by a one-dimensional vector, as in the first embodiment. Is different. That is, the compound group display device 2 of the present embodiment can display the compound structure and the functional characteristics in association with each other even for compounds whose functional characteristic information is expressed in n dimensions.

  FIG. 6 is a functional block diagram of the compound group display device 2 of the present embodiment. As shown in FIG. 6, the compound group display device 2 includes a descriptor input unit 11, a functional property input unit 12, a hierarchy classification unit (hierarchy classification unit) 23, a non-hierarchy classification unit (non-hierarchy classification unit) 27, a compound arrangement A unit (compound arrangement unit) 14, an image data generation unit (image data generation unit) 15, and a display unit 16.

  The functional characteristic input unit 12 is basically the same as that of the first embodiment, but in this embodiment, n characteristic values (quantitative variables) are acquired.

  The non-hierarchical classification unit 27 is for classifying compounds into non-hierarchical clusters based on n characteristic values (quantitative variables) input to the functional characteristic input unit 12.

  The hierarchical classification unit 23 is for classifying compounds into a plurality of hierarchical clusters based on molecular descriptors (first characteristic information). More specifically, a decision tree is created with the molecular descriptor as an explanatory variable and the cluster formed by the non-hierarchical classification unit 27 as a category.

  When the second characteristic information is composed of a single quantitative variable, the hierarchy classification unit 23 creates a decision tree (regression tree) using the molecular descriptor as an explanatory variable and the second characteristic information as an objective variable. Thus, the compounds may be classified hierarchically.

  The image data generation unit 25 is basically the same as that of the first embodiment, but image data in which the icon of the compound is drawn with a color based on the cluster formed by the non-hierarchical classification unit 27 when performing color coding. Is different from the first embodiment.

  The descriptor input unit 11, the compound arrangement unit 14, and the display unit 16 are the same as those in the first embodiment.

  The non-hierarchical classification unit 27, the hierarchical classification unit 23, the compound arrangement unit 14, and the image data generation unit 25 execute program codes stored in a storage device such as a ROM or RAM by an arithmetic device such as a CPU. It is a functional block realized by this.

  Next, the operation of the compound group display device of this embodiment will be described. FIG. 7 is a process diagram of the compound group display method of the present embodiment.

  First, in step S20, the functional characteristic input unit 12 acquires functional characteristic information. In the present embodiment, the acquired function characteristic information is assumed to be a quantitative variable represented by an n-dimensional vector as an example. Table 4 is an example of the acquired functional characteristic information.

  In Table 4, there are three types of functional property information (activity values) associated with each compound, but the present invention is not limited to this, and any number of functional property information is handled. be able to.

  Then, the acquired function characteristic information is transmitted to the non-hierarchical classification unit 27.

  Next, in step S21, the descriptor input unit 11 acquires a molecular descriptor. This step is the same as step S11 of the first embodiment. The acquired molecular descriptor is transmitted to the hierarchical classification unit 23.

  Next, in step S22, the non-hierarchical classification unit 27 classifies the compounds into non-hierarchical clusters based on the received functional characteristic information. Hereinafter, clusters classified based on functional characteristic information are also referred to as functional clusters. Here, as a method for classifying into a non-hierarchical cluster, a known method can be used. As a general thing, k-means method etc. are mentioned. Further, a neural network model, for example, Kohonen's self-organizing map method may be used. Further, a plurality of methods may be combined.

  Thereby, the functional cluster to which each compound belongs and the average functional characteristics of each functional cluster can be obtained. In other words, it is determined to which cluster each compound has what functional characteristics. The functional characteristics of each cluster are obtained, for example, as coordinates of the center of gravity of the cluster in the case of the k-means method, and as the pattern vector of the node in the case of the self-organizing map method.

  Table 5 is an example of a table showing functional clusters to which compounds belong, obtained by classification.

  As shown in Table 5, in the above table, the compound and the ID of the functional cluster are associated with each other.

  Information on the functional cluster to which the compound belongs obtained by classification is transmitted to the hierarchical classification unit 23 and the image data generation unit 25.

  Next, in step S <b> 23, the hierarchical classification unit 23 classifies the compounds into hierarchical clusters based on the molecular descriptor received from the descriptor input unit 11. Here, in this embodiment, when classifying into clusters, the functional properties of the compounds in each cluster are classified as uniform as possible. Specifically, a decision tree is created using the molecular descriptor of the compound as an explanatory variable and the function cluster to which the compound received from the non-hierarchical classification unit 27 belongs as an objective variable. As a branching rule at this time, a known rule can be used, for example, a method using C5.0, C & RT, QUEST, CHAID by Quinlan or an information gain G by Mitchell et al. It is done.

  Thereby, information having a hierarchical structure as shown in Table 3 of the first embodiment is obtained. Information on the obtained hierarchical structure is transmitted to the compound placement unit 14.

  Next, in step S24, the compound placement unit 14 places each compound (leaf node) in the two-dimensional plane based on the hierarchical structure information formed by the hierarchy classification unit 23. This step is the same as step S13 of the first embodiment.

  Next, in step S25, the image data generation unit 25 generates image data based on the received arrangement data. This step is the same as Step S14 of Embodiment 1 except that each compound (leaf node) is color-coded based on the functional cluster determined in Step S23.

  Finally, in step S26, the display unit 16 displays an image corresponding to the received image data. This step is the same as step S15 of the first embodiment.

  As described above, the compound group display device 2 according to the present embodiment creates a decision tree using molecular descriptors as explanatory variables and functional characteristic information as objective variables when classifying compounds into hierarchical clusters. As a result, the hierarchically classified compounds show uniform functional properties within the cluster. Therefore, the relationship between the structure and the functional characteristics is clarified, and the functional characteristics of the unknown compound can be easily estimated based on the molecular descriptor.

  In addition, the compound group display device 2 of the present embodiment allows the compounds to be classified into non-hierarchical clusters (functions) based on the functional property information even when the functional property information about each compound is composed of a plurality of property values. Cluster), and the functional cluster to which it belongs is treated as the functional property information of the compound. When the image data generation unit 25 performs color coding of the icon, the functional characteristics of the compound are expressed by color coding according to the functional cluster.

  In each of the above-described embodiments, the hierarchical classification unit, the non-hierarchical classification unit, the compound arrangement unit, and the image data generation unit that constitute the compound group display device indicate that “an arithmetic device such as a CPU is in a storage device such as a ROM or a RAM. In the above description, the function block realized by executing the stored program code is described as an example. However, the function block may be realized by hardware that performs similar processing. Further, it can also be realized by combining hardware that performs a part of the processing and the above-described calculation means that executes the program code for controlling the hardware and the remaining processing. Further, even among the members described above as hardware, the hardware for performing a part of the processing and the arithmetic means for executing the program code for performing the control of the hardware and the remaining processing It can also be realized by combining them. The arithmetic means may be a single unit, or a plurality of arithmetic means connected via a bus inside the apparatus or various communication paths may execute the program code jointly.

  The program code itself that can be directly executed by the computing means, or a program as data that can be generated by a process such as decompression described later, stores the program (program code or the data) in a recording medium, A recording medium is distributed, or the program is distributed by being transmitted by a communication means for transmitting via a wired or wireless communication path, and is executed by the arithmetic means.

  In addition, when transmitting via a communication path, each transmission medium which comprises a communication path propagates the signal sequence which shows a program, and the said program is transmitted via the said communication path. Further, when transmitting the signal sequence, the transmission device may superimpose the signal sequence on the carrier by modulating the carrier with the signal sequence indicating the program. In this case, the signal sequence is restored by the receiving apparatus demodulating the carrier wave. On the other hand, when transmitting the signal sequence, the transmission device may divide and transmit the signal sequence as a digital data sequence. In this case, the receiving apparatus concatenates the received packet groups and restores the signal sequence. Further, when the transmission apparatus transmits a signal sequence, the signal sequence may be multiplexed with another signal sequence and transmitted by a method such as time division / frequency division / code division. In this case, the receiving apparatus extracts and restores individual signal sequences from the multiplexed signal sequence. In any case, the same effect can be obtained if the program can be transmitted via the communication path.

  Here, it is preferable that the recording medium for distributing the program is removable, but it does not matter whether the recording medium after distributing the program is removable. In addition, the recording medium may be rewritten (writeable), volatile, or the recording method and shape as long as a program is stored. Examples of recording media include tapes such as magnetic tapes and cassette tapes, magnetic disks such as floppy (registered trademark) disks and hard disks, CD-ROMs, magneto-optical disks (MO), mini-discs (MD) and digital A disk such as a video disk (DVD) may be mentioned. The recording medium may be a card such as an IC card or an optical card, or a semiconductor memory such as a mask ROM, EPROM, EEPROM, or flash ROM. Or the memory formed in calculating means, such as CPU, may be sufficient.

  The program code may be a code for instructing the arithmetic means of all the procedures of the processes, or a basic program capable of executing a part or all of the processes by calling according to a predetermined procedure. If (for example, an operating system or a library) already exists, a part or all of the entire procedure may be replaced with a code or a pointer that instructs the arithmetic means to call the basic program.

  The format for storing the program in the recording medium may be a storage format that can be accessed and executed by the arithmetic means, for example, as in a state where the program is stored in the real memory, or is stored in the real memory. Installed in the local recording medium from the storage format after being installed in a local recording medium (for example, real memory or hard disk) that is always accessible by the computing means, or from a network or a transportable recording medium The previous storage format may be used. Further, the program is not limited to the compiled object code, but may be stored as source code or intermediate code generated during interpretation or compilation. In any case, the above calculation is performed by a process such as decompression of compressed information, decoding of encoded information, interpretation, compilation, linking, allocation to real memory, or a combination of processes. If the means can be converted into an executable format, the same effect can be obtained regardless of the format in which the program is stored in the recording medium.

[Example 1]
One embodiment of the present invention will be described below. Note that the present invention is not limited to this. This example relates to the display of the measurement results of the solubility of a compound in water. Specifically, the structure and solubility of the compound were related and displayed.

  In this example, as the second characteristic information (functional characteristic information), a document (Y. Ran, et al., Prediction of aqueous solubility of organic compounds by the general solubility equation (GSE). J. Chem. Inf. Comput Sci., 41: 1208-1217, 2001; G. Yan, et al., Prediction of the aqueous solubility: comparison of the general solubility equation and the method using an amended solvation energy relationship.J. Pharm. Sci., 91 : 517-533, 2002) was used. The total number of solubility data collected, that is, the total number of compounds handled in this example was 908. The solubility data was obtained as a table associated with the compound.

  Moreover, as the first characteristic information (structure information), Lipinski et al. (CA Lipinski, et al., Experimental and computational approaches to estimate solubility and permeability in drug discovery and development settings. Adv. Drug Deliv. Rev. 23: 3 -25, 1997) used four molecular descriptors presented to identify drug-like compounds, specifically ClogP, molecular weight, number of hydrogen bond donating groups, and number of hydrogen bond accepting groups. These molecular descriptor data were acquired as a table associated with the compound. A part of this correspondence table is shown in Table 6.

  For each of these molecular descriptors, standardization was performed such that the average between compounds was 0 and the standard deviation was 1. Table 7 shows the correspondence table of each molecular descriptor after standardization.

Next, 908 compounds were classified into hierarchical clusters based on the normalized molecular descriptor. As the classification algorithm, the Ward method (JH Ward, Hierarchical Grouping to Optimize an Objective Function. J. Am. Stat. Assoc., 58: 236-244, 1963) was used. The Ward method is a method of selecting a set of clusters that sets the information loss in the initial state (unconnected state) to 0 and minimizes the increase in information loss ΔS _pq at each connection step in cluster merger. ΔS _pq is given by:

Where x _pm and x _qm with bars are the values of the m-th molecular descriptor at the center of gravity of cluster p and cluster q before binding, respectively, and n _p and n _q are the clusters belonging to clusters p and q or The number of compounds. In this embodiment, L is the number of types of molecular descriptors, that is, four.

Based on this method, cluster merging was repeated until the number of clusters was 1. Then, the threshold of information deficiency ΔS _pq was set to 1.0, and those below it were regarded as the same hierarchy, and 908 compounds were classified.

  Next, the arrangement in a two-dimensional plane was determined so that the compound classified hierarchically might become a recursive nested structure based on a hierarchical cluster. Heiankyo View ("Heiankyo View-Visualization Method for Arranging Hierarchical Data on the Board", Takayuki Ito, Koji Koyamada, Visualization Society of Japan 9th Visualization Conference Abstract, 2003) Was used.

  And about 908 compounds, the image data which drew the icon by the color which made the logarithm value of solubility the hue by the arrangement | positioning determined was produced. An image based on this image data is shown in FIG. In the figure, the icon indicating the compound is displayed in color.

  As shown in FIG. 8, clusters in which only those having relatively high solubility or only low solubility were observed, indicating that compounds having similar chemical structures have relatively similar solubility.

  Moreover, what gave height to the icon which shows each compound is shown to Fig.9 (a) and FIG.9 (b). FIG. 9A shows ClogP of each compound by the height of the bar. On the other hand, FIG.9 (b) represents molecular weight with the height of the bar instead of ClogP. In this way, it is possible to easily grasp visually that the smaller the ClogP value and the smaller the molecular weight, the easier it is to dissolve in water, even though the data is as large as 908 compounds. did it.

[Example 2]
One embodiment of the present invention will be described below. This example relates to the classification of the metabolic pattern of a compound (drug) by the drug-metabolizing enzyme cytochrome P450 (CYP), and the structure of the compound and the metabolic pattern are displayed in association with each other.

  CYP is an important group of drug metabolizing enzymes related to biological substance metabolism and foreign body detoxification, and about 20 molecular species have been identified in humans. Many of the drugs that are foreign to the living body are lost due to metabolism by CYP, so it can be said that metabolism by CYP is deeply related to the effectiveness and safety of the drug. Among the CYP superfamily, five of CYP1A2, CYP2C9, CYP2C19, CYP2D6, and CYP3A4 are major molecular species involved in drug metabolism, and it is said that about 90% of drug metabolism can be explained by these. Therefore, evaluation and prediction of drug metabolism by these five CYPs is a very important issue in drug discovery research.

  In this example, as the second characteristic information (functional characteristic information), the data on metabolic sensitivity by the above five CYPs for 161 drugs investigated by Bonaburi et al. (P. Bonnabry, et al., Quantitative) drug interactions prediction system (Q-DIPS): a dynamic computer-based method to assist in the choice of clinically relevant in vivo studies. Clin. Pharmacokinet., 40: 631-640, 2001). Incidentally, metabolic sensitivity data by CYP is evaluated in three stages. Metabolic susceptibility data was obtained as a table associated with the compounds. Table 8 shows a part of metabolic sensitivity data after standardization in the same manner as in Example 1.

  In addition, as the first characteristic information (structure information), among 220 molecular descriptors (Molconn-Z descriptors) calculated by Molconn-Z, those having a path length of 7 or more and dispersion in the data set A total of 115 molecular descriptors were used, excluding molecular descriptors with a zero. The Molconn-Z descriptor of each drug was calculated from the chemical structure using Molconn-Z (trademark, manufactured by Hall Associate Consulting). Table 9 shows a part of the Molconn-Z descriptor of each compound.

  Next, in this example, since the functional property information (metabolic pattern) of each compound is composed of five quantitative variables (5-dimensional vectors), the compounds are classified into non-hierarchical functional clusters based on the functional property information. . As an algorithm for classifying into functional clusters, a combination of Kohonen's self-organizing map method (T. Kohonen, “Self-Organizing Maps”, Springer, Berlin, Heidelberg, 1995) and the k-means method was used. These will be described in detail below.

  First, a five-dimensional pattern vector having an attribute of metabolic strength by each CYP was considered, and a self-organizing map (7 × 9 = 63 nodes) was learned using the obtained metabolic sensitivity data by CYP. Table 10 shows a part of a table indicating pattern vectors of each node obtained by learning.

  Moreover, what visualized said Table 10 is shown in FIG. In FIG. 10, each component of the pattern vector possessed by each node of the self-organizing map is displayed by hue. In FIG. 10, since a wide range of nodes showed high values for elements corresponding to CYP3A4, it was visually perceived that many drugs are metabolized by CYP3A4. In addition, since there was almost no overlap between CYP1A2 and CYP2C9 or CYP2C19, it was revealed that CYP1A2 exhibits a substrate selectivity significantly different from CYP2C9 and CYP2C19.

  Next, since the number of nodes in the self-organizing map is 63, and there are a large number to use as a functional cluster as it is, the nodes are further classified into the optimum number of clusters by the k-means method. Here, the Davies-Bouldin Index (DB) represented by the following equation was used as an index for obtaining the optimum number of clusters.

However, S _n (Q _i) represents the average value of the Euclidean distance between the cluster center and the cluster element in a cluster Q _i, S (Q _i, Q _j) is the Euclidean distance between the centers of clusters Q _i and Q _j Represents.

  The optimal number of clusters was determined to be 6 by DB.

  Next, 63 nodes in the self-organizing map were classified into 6 clusters by the k-means method, and these were defined as functional clusters. As a result, a table indicating which of the six clusters each node in the self-organizing map belongs to was obtained. A part of the table is shown in Table 11.

  Moreover, what visualized Table 11 is shown in FIG. In FIG. 11, the cluster to which each node belongs is represented by a hue.

  Next, the node to which the drug belongs (best match node) is determined from the distance between the pattern vector of metabolism of each drug and the pattern vector of each node of the self-organizing map, and the node is assigned to any of the above six clusters. Drugs were classified into functional clusters according to whether they belonged. Table 12 shows the average score of metabolic sensitivity by CYP of drugs classified into each functional cluster.

  From Table 12, the functional cluster 1 is significantly metabolized by CYP3A4 and CYP2C9, the functional cluster 2 is significantly metabolized by CYP2C19, the functional cluster 3 is only CYP2C9, the functional cluster 4 is only CYP3A4, and the functional cluster 5 is CYP2D6. It can be seen that functional cluster 6 is a collection of drugs that are metabolized by CYP1A2.

  Next, binary trees were created to classify compounds hierarchically based on structural information. When creating a binary tree, a Molconn-Z descriptor is used as the explanatory variable, a functional cluster is used as the objective variable, and the following information gain G is used as the branch rule (TM Mitchell, “Machine Learning”, McGraw-Hill, Singapore, 1997) Was used.

However, N is the indicated number of clusters when the classification of drugs function clusters (= 6), s is the total number of drug in the set before the division, c _i represents the drug number belonging to functional cluster i, s ₁ and s ₂ indicates the total number of drugs in each of the two divided sets, and c _{1, i} and c _{2, i} indicate the number of drugs belonging to the functional cluster i in each of the two divided sets.

  Here, in each division step of the binary tree, a condition (type and type of Molconn-Z descriptor used for determination of division) and a condition for obtaining the maximum information gain G are determined, and the value of the maximum information gain G obtained is 0. The division was repeated recursively until it was 25 or less. A part of the data obtained by creating the binary tree is shown in Table 13.

  As shown in Table 13, in the root node (node 1), the drug is divided into two child nodes (node 2 and node 3) depending on whether or not the SHCsats of the Molconn-Z descriptor is less than 4.80369. Similarly, in node 2, the drug is divided into two child nodes (node 30 and node 31) depending on whether or not the SHHBA of the Molconn-Z descriptor is less than 31.98136.

  Next, the arrangement | positioning in a two-dimensional plane was determined so that it might become a recursive nested structure based on a hierarchical cluster about the drug classified hierarchically. Heiankyo View was used as the algorithm for determining this arrangement.

  Then, for 161 drugs, image data was created in which the icons were drawn with the determined arrangement and the color of the function cluster type as a hue. An image based on this image data is shown in FIG. In the image, an icon indicating a drug is displayed in color, and the color of the icon is a color corresponding to the functional cluster to which the icon belongs when classified based on the metabolic pattern.

  As shown in FIG. 12, in the displayed image, drugs having common structural features are classified so as to show the same metabolic pattern. Thus, it was shown that the relationship between structure and metabolic pattern can be effectively visualized for a large number of 161 drugs.

  FIG. 13 shows the metabolic sensitivity of cytochrome P450 by each molecular species CYP1A2, CYP2C9, CYP2C19, CYP2D6, and CYP3A4 as individual images. In the image, metabolic sensitivity (3 levels) is represented by the color of the icon. As described above, the sensitivity of each drug-metabolizing enzyme can be comprehensively understood only by looking at the displayed image, and the usefulness of the present invention was shown.

  The present invention is not limited only to the above-described embodiments and examples, and various modifications are possible within the scope of the claims, and technical means disclosed in different embodiments and examples, respectively. Embodiments obtained by appropriately combining the above are also included in the technical scope of the present invention.

  Since the present invention can effectively visualize the entire screen of screening data, it can assist in the selection of candidate compounds, and can show the direction of synthesizing and developing functionally superior compounds. Therefore, it can be used not only for pharmaceuticals and agricultural chemicals but also in the chemical industry in general for exploratory research of other functional chemical substances, and is considered to be very useful.

1 is a functional block diagram of a compound group display device, showing an embodiment of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1, showing an embodiment of the present invention, is a process diagram showing processing steps of a compound group display device. It is a figure which shows an example of the dendrogram obtained by hierarchical classification. It is a figure which shows an example of arrangement | positioning of the compound by a compound arrangement | positioning part. It is a figure which shows an example of the image data produced | generated by the image data production | generation part. Another embodiment of this invention is shown and it is a functional block diagram of a compound group display apparatus. FIG. 5 is a process diagram showing another embodiment of the present invention and showing a treatment process of a compound group display device. BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1, showing an example of the present invention, is a diagram showing an image displayed by a compound group display device. BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1, showing an example of the present invention, is a diagram showing an image displayed by a compound group display device. FIG. 10 is a diagram showing another embodiment of the present invention, and showing each component of the pattern vector of each node in the self-organizing map in hue. It is a figure which shows another Example of this invention and showed the functional cluster to which each node in a self-organization map belongs by the hue. It is a figure which shows another Example of this invention and shows the image displayed by the compound group display apparatus. It is a figure which shows another Example of this invention and shows the image displayed by the compound group display apparatus.

Explanation of symbols

1, 2 Compound group display device 13, 23 Hierarchy classification unit (hierarchy classification means)
14 Compound arrangement part (compound arrangement means)
15, 25 Image data generation unit (image data generation means)
16 Display section

Claims (7)

For each of the plurality of compounds, acquisition means for acquiring structural information indicating the structure of each compound, and functional property information indicating the functional properties of each compound,
Hierarchical classification means for classifying each of the above compounds into a plurality of hierarchical clusters;
Compound placement means for determining placement in a two-dimensional plane so that icons representing each of the compounds do not overlap so as to have a recursive nested structure based on a hierarchical cluster classified by the hierarchy classification means ,
Image data generating means for generating image data in which an icon of the compound is drawn with a color, pattern, and / or shape based on the functional property information in an arrangement determined by the compound arranging means;
A display unit for displaying an image based on the image data generated by the image data generation means,
The hierarchical classification means is
Each compound is classified into a hierarchical cluster based on its structure, and predetermined functionalities are determined based on the acquired structural information and functional property information so that the functional properties of the compounds in each cluster are uniform. Perform the above classification using an algorithm ,
further,
Non-hierarchical classification means for classifying the compound into a non-hierarchical functional cluster based on the functional property information,
The hierarchical classification means classifies the compounds by creating a decision tree using the structural information as explanatory variables and the categorical variables corresponding to the functional clusters formed by the non-hierarchical classification means as objective variables. A compound group display device. The functional characteristic information is composed of a plurality of quantitative variables,
2. The compound group display device according to claim 1, wherein the non-hierarchical classification means classifies the compounds into non-hierarchical functional clusters by a self-organizing map method using the functional property information as a pattern vector. . The above structural information
(a) one or more molecular descriptors representing structural fragments or structural topologies of the molecular structure of the compound;
(b) one or more molecular descriptors corresponding to the physicochemical properties determined by calculation or experiment;
3. The compound group display device according to claim 1, comprising at least one of the following. The functional property information is at least one parameter of biological activity, toxicity, physicochemical property, and pharmacokinetic property of the compound, according to any one of claims 1 to 3. Compound group display device. The compound group display device according to any one of claims 1 to 4, wherein the functional property information includes categorical variables. The program for operating a computer as each means of any one of Claim 1 to 5. A computer-readable recording medium on which the program according to claim 6 is recorded.