JP2003028855A - Method for evaluation and display of clustered result - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method in which how much each clustering method is reliable and what kind of feature each clustering method has are evaluated on the basis of a plurality of clustered results and to provide a method of evaluating and displaying a degree at which a gene has a designated biological function when a gene having an unknown function is estimated. SOLUTION: On the basis of a method in which a gene having a known biological function is sorted and on the basis of a rate at which the gene has e.g. a corresponding biological function is judged correctly, whether each clustering method is performed satisfactorily is judged. When to which cluster the gene having the unknown function belongs is found on the basis of the plurality of clustered results, the degree at which the gene has the designated biological function is evaluated.

Description

Detailed Description of the Invention

[0001]

TECHNICAL FIELD The present invention is intended to display gene expression data obtained by hybridizing a specific gene in a format that is visually easy to understand and that the function / role of the gene is easily guessed. Regarding the display method of.

[0002]

2. Description of the Related Art As the number of species whose genome sequences have been determined increases, we find genes that appear to correspond to evolution, search for a set of genes that are thought to have in common in all organisms, and vice versa. The so-called genome comparison method, which tries to find out something from the difference in genes between species, such as inferring individual characteristics to each other, has been actively performed.

However, in recent years, with the development of infrastructures such as DNA chips and DNA microarrays (hereinafter collectively referred to as biochips), the interest in molecular biology has changed from interspecies information to intraspecies information. We are moving to simultaneous expression analysis, and in addition to the comparison between species so far, the field of information extraction to association is beginning to expand greatly.

For example, if an unknown gene showing the same expression pattern as a known gene is found, it is assumed that it has the same function as the known gene. Functional implications of these genes and proteins themselves have been studied in the form of functional units and functional groups. In addition, the interaction between them may be caused by destroying or overreacting a gene by associating it with known enzyme reaction data or substance metabolism data, or more directly, to eliminate the expression of that gene, or to increase the amount in large quantities. It is expressed and the direct and indirect effects of that gene are analyzed by examining the expression pattern of all genes.

A successful case in this field is the expression analysis of yeast by P. Brown et al. At Stanford University (Michel B. Eisen et. A.
l.: Cluster analysis and display of genome-wide ex
pression patterns: Proc. Natl. Acad. Sci. (1998) D
ec 8; 95 (25): 14863-8). Using a biochip, they hybridized genes extracted from cells in time series and quantified the degree of gene expression (brightness of hybridized fluorescent signal). By correlating numerical values with colors, individual expression processes of genes are displayed in an easy-to-understand manner. At this time, genes that have similar expression pattern processes in a series of cell cycles (that is, genes that have similar expression levels at arbitrary points in time) are clustered.

FIG. 1 is an example in which the expression states of genes are displayed according to a method called hierarchical clustering, in which the experimental cases are arranged in the horizontal direction and the genes are arranged in the vertical direction. The dendrogram on the left side shows a situation in which two nearest clusters have been merged in the clustering process, and the length of each branch corresponds to the distance between the two clusters at the time of merging.

With the development of biology, the function of genes has been gradually clarified, and biological researchers are trying to combine gene expression data with known information to carry out gene analysis. In the dendrogram analysis, researchers look for biologically meaningful clusters (sets of genes). That is, if the expression patterns of the genes included in the cluster are similar and there are many known genes having the same function, they are extracted as a meaningful cluster. Such a cluster is called a functional cluster here. For example, if a gene having an unknown function is included in the genes included in the functional cluster, it can be inferred that the gene has a function similar to that of a known function in the same cluster. Moreover, by observing the expression pattern of functional clusters, the expression process peculiar to the function can be found.

[0008]

The method of clustering based on gene expression data obtained from a biochip is not limited to the hierarchical clustering described in FIG.
Various things such as self-organizing map, Fisher linear discriminant, support vector machine, k-Means method, Parzen estimation, decision tree, principal component analysis are studied. Moreover, these are classified differently depending on how the parameters are given to the clustering algorithm. For example, in hierarchical clustering, various classifications (dendrograms) can be obtained by a combination of a distance indicating the degree of similarity between expression patterns and an algorithm for merging clusters. Further, in the case of hierarchical clustering, as shown in FIG. 2, the size of the cluster differs depending on which level should be considered as one cluster.

However, which method is most suitable for classification of expression patterns among many clustering methods depends on the property of the data itself and is not known. Therefore, researchers apply various methods to expression data to explore functional clusters. However, if various clustering methods are tried, a plurality of classification results can be obtained, so it is necessary to compare them. That is, it is necessary to comprehensively judge which functional cluster a gene of unknown function belongs to by looking at various classification results. This is a very troublesome task for researchers.

[0010] Actually, depending on the clustering method, there are cases where genes are well classified according to gene function, and therefore, when they are compared equally, correct results may not be obtained. In addition, even if a functional cluster is found by clustering and the function of a gene whose function is unknown is inferred, in order to confirm this, usually, an additional test is conducted in an experiment other than biochips, or the same thing is done in other organisms. The public database will be used to check if is checked. These follow-up experiments
If you have a large amount of unknown genes, it will be a complicated task to examine them one by one, because you cannot perform comprehensive analysis like biochips.

Therefore, in the present invention, in view of the problems of the prior art, a method for evaluating how reliable each clustering method is and what features it has from a plurality of clustering results, and a gene whose function is unknown The purpose of the present invention is to provide a method and an indication for evaluating the degree to which a gene has a specified biological function when estimating the above.

[0012]

According to the present invention, it is possible to evaluate which clustering method is used for successful classification from a plurality of clustering results, and to see how genes with known biological functions are classified. To do.

FIG. 3 is a diagram showing an example of the display method according to the present invention. In this display example, genes are classified by various clustering methods, and genes whose functions are known are referenced to compare the proportions that are correctly determined to be the corresponding biological functions. This ratio is called the existence hit ratio. For example, the existence hit ratios circled in the figure are the biological functions “cytoplasmic r” after being classified by the decision tree (MOC1).
Of the genes with known functions classified into the functional cluster of "ibosomes", the ratio (78.5%) of those with a truly functional function of "cytoplasmic ribosomes" is shown.

By looking at the existence hit rate in each functional cluster, it is possible to evaluate how reliable each clustering method is and what kind of characteristics it has. For example, in the case of FIG. 3, focusing on the hierarchical cluster analysis, four out of five biological functions can be correctly discriminated at a rate of 80% or more. Looking at other algorithms, only three or less of the five biological functions could be correctly identified at a rate of 80% or more. However, SVM discriminates against more biological functions than cluster analysis at a rate of 90% or higher. Therefore, in this data, it can be confirmed that the hierarchical cluster analysis tends to be suitable for rough classification rather than precise classification.

Researchers want to focus on what works well for classification. Therefore, in order to emphasize it, the validity of each clustering method can be visually evaluated easily by inputting in the dialog box 301, for example, such that the existence hit ratio exceeds 80% is highlighted. be able to.

FIG. 4 is a diagram showing another example of the display method according to the present invention. In this display method, contrary to FIG. 3, various clustering methods are compared at the rate of correctly judging that the biological function is not the corresponding biological function. This ratio is called the non-existence hit ratio. From this result, it is possible to evaluate how reliable each clustering method is and what characteristics it has.

FIG. 5 is a diagram showing another example of the display method according to the present invention. In this display method, the number of known genes correctly determined to have the corresponding biological function and the number of known genes contained in the functional cluster are displayed. The presence hit rate is the number of known genes included in the functional cluster divided by the number of correctly determined known genes.
For example, even if the existence hit ratio is 50, the existence hit ratio is so reliable in a situation where one of two known genes has a corresponding function, as shown by the circled portion in FIG. You can see that it is not a numerical value.

FIG. 6 shows a hierarchical cluster analysis, self-organization map, SVM (3
FIG. 3 is a schematic diagram showing a collection of genes of unknown function belonging to the functional cluster of the biological function “TCA cycle”, which are classified by the following three clustering methods. Each point in FIG. 6 corresponds to a gene. These genes whose function is unknown will be inferred to have the function of "TCA cycle". Here, each gene can be distinguished by the difference in how it is classified. For example, the gene 601 was determined to be included in the functional cluster of the function “TCA cycle” by the three methods of hierarchical cluster analysis, self-organization map, and SVM (third order). (3
In the next), it was considered to be in the functional cluster of "TCA cycle", but in the other two clustering methods, "TCA cy"
cle ”was considered not to be in the functional cluster.

When a gene whose function is unknown is retested, many genes cannot be examined at once as described above. However, when the situation as shown in Fig. 6 is understood, the gene regarded as the function "TCA cycle" by the three methods of hierarchical cluster analysis, self-organization map, and SVM (third order) has the highest probability of function " It can be understood that it may have a "TCA cycle". Therefore, when performing a supplementary test, it is sufficient to first target genes in this range.

Further, by adding weight to each range based on the results obtained in FIGS. 3 and 4, it is possible to more flexibly evaluate the plausibility of the function estimation. For example, in FIG.
It can be seen from the hit ratio of the function clusters related to the function "TCA cycle" that the collection of the function "TCA cycle" is bad for the self-organizing map and good for the SVM (3rd order). Therefore, in FIG. 5, gene 602 has a higher probability of function “TCA cycl” than gene 603.
You can see that you have "e".

FIG. 7 is a diagram showing still another example of the display method according to the present invention, in which the degree to which each gene described above has a function is quantified. When a biological function name is entered in the box 701, the degree of having the entered function is calculated for each gene and arranged in order of degree (score). The score is the existence hit ratio shown in FIGS. 3 and 4,
Weighting is performed based on the non-existence hit ratio and other indexes, and then the gene is calculated depending on whether or not the gene belongs to the functional cluster obtained by each clustering method. From this display, it is possible to easily find the gene having the function shown in the box 701 with a high probability, and it can be useful for the additional test.

In summary, the gene clustering result evaluation method according to the present invention is a gene clustering result evaluation method for evaluating the result of clustering a plurality of genes by a predetermined clustering method based on the expression pattern thereof. As a result of clustering, regarding a gene group that is determined to have a predetermined biological function, among genes whose biological function belonging to the gene group is known, It is characterized in that the result of clustering is evaluated based on the ratio.

The gene clustering result evaluation method according to the present invention is also a gene clustering result evaluation method for evaluating the result of clustering a plurality of genes by a predetermined clustering method based on the expression pattern thereof. , For a group of genes that have been identified as not having a predetermined biological function,
It is characterized in that the result of clustering is evaluated based on the ratio of genes known to have no predetermined biological function among genes whose biological function belonging to the gene group is known.

The gene clustering result display method according to the present invention is a gene clustering result display method for displaying a result of clustering a plurality of genes according to a predetermined clustering method based on the expression patterns thereof. For a group of genes determined to have a biological function of, the ratio of genes known to have the predetermined biological function among the genes of known biological function belonging to the gene group is calculated. It is characterized by displaying.

The gene clustering result display method according to the present invention is also a gene clustering result display method for displaying a result of clustering a plurality of genes by a predetermined clustering method based on their expression patterns. , For a group of genes that have been identified as not having a predetermined biological function,
It is characterized by calculating and displaying a ratio of genes whose biological function is known to belong to the gene group and which are known not to have the predetermined biological function.

The gene clustering result display method according to the present invention is also a gene clustering result display method for displaying a result of clustering a plurality of genes by a predetermined clustering method based on their expression patterns. , For each gene group determined to have a predetermined biological function, the number of genes whose biological function is known belonging to the gene group and the number of genes known to have the predetermined biological function. It is characterized by displaying the number and.

The gene clustering result display method according to the present invention is also a gene clustering result display method for displaying a result of clustering a plurality of genes by a predetermined clustering method based on their expression patterns. , For each gene group that is determined not to have a predetermined biological function, the number of genes of known biological function belonging to the gene group and the number of genes that do not have the predetermined biological function are known. It is characterized by displaying the number of genes.

The method of extracting a gene candidate having a predetermined biological function from a plurality of genes according to the present invention is as follows:
The step of clustering a plurality of genes by a plurality of clustering methods based on the expression pattern, and the result of clustering by one clustering method, the genes determined to have the predetermined biological function, the clustering method The step of adding weights set for the step, and the step of adding the weights is executed for the clustering results by all the remaining clustering methods, the step of calculating a score for each gene, and the size of the score as a reference. And a step of extracting a candidate of a gene having the predetermined biological function.

The method of displaying a candidate of a gene having a predetermined biological function according to the present invention comprises
The step of clustering by a plurality of clustering methods based on the expression pattern, and the result of clustering by one clustering method, the genes determined to have the predetermined biological function are set for the clustering method. A step of adding weights, a step of adding the weights is executed for clustering results by all remaining clustering methods, and a step of calculating a score for each gene;
And displaying the genes in a descending order of the score together with the score of the gene.

[0030]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings. FIG. 8 is a system configuration diagram of the present invention. This system includes a biochip database 801, which records gene information and expression processes, a display device 802 for displaying analysis results, a keyboard 803 and a mouse 804 for inputting and selecting values into the system. Input means of, clustering processing,
A central processing unit 805 for calculating the existence hit ratio, non-existence hit ratio, score, and the like, and a program memory 80 for storing a program necessary for the processing in the central processing unit 805.
6 is provided. In the program memory 806,
Clustering processing program 807 for clustering expression pattern data, hit rate calculation processing program 808 for calculating presence hit rate and non-existence hit rate, score calculation processing program 809 for calculating degree of biological function for each gene, and these analyzes An analysis result display processing program 810 for displaying the calculation result is provided. These programs are CD-ROM, DV
It can be provided by being stored in a recording medium such as a D-ROM, MO, floppy (registered trademark) disk, or provided via a network. Further, the biochip database 801 may be stored in a storage device connected to the central processing unit 805, or may be configured to be managed by a server computer installed in a remote place, and the database can be used to store gene data via a network or the like. May be acquired.

FIG. 9 shows a biochip database 801.
3 shows a specific structure of data stored in. It is assumed that the gene information is stored in an array of length gNum called gene [i] (i = 1,2, ..., gNum). However, gNum is the number of genes included in the gene data. The gene data includes a gene ID (900) that uniquely determines a gene, attribute information (901) that represents the gene, and expression data (902) obtained from the biochip. Examples of attributes that represent genes include gene name (903) and ORF (90
4) and gene function (905). If the function of the gene is unknown, the function attribute 905 is "UNKOWN". In addition to these gene attributes, it is also possible to define them as members of the gene information structure. In addition, the expression data stores data in which the degree of gene expression (brightness of hybridized fluorescence signal) is quantified in each experiment. In the present invention, the number of experiments is n, and one gene is treated as an n-dimensional vector.

FIG. 10 shows a structure array for storing the clustering result.
It is assumed that they are stored in an array of length crNum called ult [i] (i = 1,2, ..., crNum). However, crNum is the number of times of clustering. If the parameters given to the clustering method are changed, they will be treated as separate classification results, and will also be taken separately as elements of the array. This structure array has a clustering method name (1000) for identifying what kind of clustering method the classification result is, a specific clustering method (1001) and parameter information (1002) given to the clustering algorithm. ), The total number of known genes contained in the genes to be clustered (1003), and the number of clusters indicating the number of classified clusters (1004)
And a pointer (1005) to the first cluster.
The actual contents of each cluster are stored in a list structure,
All clusters can be traced from the first cluster 1005. Each list corresponds to each cluster,
As a member of the list, the number of belonging genes (1006) indicating the number of genes included in the cluster, the belonging gene ID (1007) indicating the included genes, the function cluster name (1008) of the cluster, and the function at the time of score calculation The weight to apply to the cluster (1009) and the pointer to the next cluster (1010) that points to the list to the next cluster
Consists of. 1 if the cluster is not a functional cluster
The functional cluster name is “UNKOWN” as in 011 and 10
The weight is set to 0 like 12. When the end of the list is reached, the pointer to the next cluster is set to NULL as in 1013.

FIG. 11 shows a specific structure of sequence data used for score calculation. This is score [i]
It is assumed that they are stored in an array of length sNum called (i = 1,2, ..., sNum). However, sNum is the total number of genes targeted for the score. This structure sequence has the gene ID (1
100) and a score (1101). 12
3 is a flowchart showing an outline of processing according to the present invention. First, gene expression pattern data gene [i] (i = 1,2, ..., gNum) is read from the biochip database 801 (step 1200). Next, the clustering method to be applied and the parameters to be given to the clustering algorithm are selected using the keyboard 803 or the mouse 804 (step 1201). After that, cluster analysis is performed. The result is registered in clusterResult [i] (i = 1, ..., CrNum) (step 1202). That is, step 120
The name that uniquely identifies the clustering method determined in 1 at the time of display is the clustering method name of clusterResult [i] 10
00, the actual clustering method and parameters in the clustering method 1001 and parameter information 1002,
The number of known genes in the target gene data is 1003, and the number of clusters is 10.
04, the number of genes included in the cluster is entered in the number of belonging genes 1006, the included genes are entered in the belonging gene ID 1007, and the functional cluster name is entered in 1008, and these are pointed to the next cluster. 1
A list structure is connected by 010, and the head pointer is registered in the pointer 1005 to the head cluster.

The clustering method includes a method of clustering using known information (with teacher) and a method of clustering without using known information (without teacher). If there is a teacher, the function cluster is uniquely determined.
It is not decided without a teacher. So without teachers,
Which cluster is a functional cluster may be determined by, for example, selecting a known gene having the most relevant biological functions as a functional cluster or a known gene having the relevant biological function. It is necessary to determine in advance, for example, that the gene to which a representative gene belongs is a functional cluster.

This clusterResult [i] is obtained in step 120.
In addition to storing the result output in 2, the previous clustering result may be stored in a biochip database or the like and used as it is. Then clus
Based on the result of terResult [i], the existence hit ratio, the non-existence hit ratio, and the score are calculated. These detailed processes will be described later (steps 1203 and 1204). With that, the process ends.

FIG. 13 shows the existence hit ratio in FIG.
It is a detailed flow of a process (step 1203) of calculating a non-existence hit ratio. Using the keyboard 803 and the mouse 804, a biological function to be investigated is selected and, when highlighted, the threshold value is input (step 13).
01). The following processing is performed until i becomes crNum with i = 1 (steps 1301 and 1302). First, clusterRes
From ult [i], a function cluster C corresponding to the biological function to be investigated selected in step 1300 is searched. I.e. c
Pointer 1005 to the first cluster of lusterResult [i]
The list is traced from, and a function cluster name 1008 that matches the biological function is searched for and is set as C (step 1303).

Next, the existence hit rate is calculated. Function cluster
Among the gene IDs belonging to C, the number R of known genes having a biological function selected in step 1300 and the number S of known genes having no function are obtained. That is, when R is obtained, the genes included in the belonging gene ID 1007 are counted, and those whose function 905 is the corresponding biological function are counted. When S is obtained, the function 905 is not the corresponding biological function. , And count those whose functions are not "UNKNOWN". After that, existence hit rate 100 ×
Calculate R / (R + S) (steps 1304, 130)
5).

Next, the non-existence hit ratio is calculated. The gene IDs included in the clusters other than the functional cluster C are examined to obtain the number T of known genes having the biological function selected in step 1300. This is also the gene ID 100
Among the genes included in 7, the functions whose function 905 is the corresponding biological function are enumerated. Then clusterRes
Using the total known gene number 1003 of ut [i], the non-existence hit rate 100 × {total known gene number- (R + S + T)} / {total known gene number- (R + S)} is calculated ( Steps 1306 and 1307).

Steps 1303 to 1307 are performed for all the biological functions selected in step 1300, and when they are completed, i is incremented by 1 and the process returns to step 1302 (step 1308). If i reaches crNum in step 1302, it means that all the existing hit ratios and non-existent hit ratios have been calculated. Once all are calculated, the hit rate, the non-existent hit rate,
The values of R and S are displayed on the display device 802, and the process ends (step 1309).

FIG. 14 is a detailed flow of the process (step 1204) for calculating the score in FIG. A keyboard 803 and a mouse 804 are used to select a biological function to be scored. Weights are set for each clustering method. The weight may use the value of the existence hit rate or the non-existence hit rate, or may be set based on another index. This is the weight of each cluster 10
09 (step 1400).

The following process is performed until i becomes crNum with i = 1 (steps 1401 and 1402). First cluste
From rResult [i], the function cluster C corresponding to the biological function to be investigated selected in step 1400 is searched. That is, the pointer 10 to the first cluster of clusterResult [i]
The functional cluster name 1008 is traced from 05.
Finds one that matches the biological function and sets it as C (step 1403). The gene ID belonging to the functional cluster C is investigated, and it is the gene I of score [i] (i = 1,…, sNum)
If not registered in D1100, the gene ID
Is registered in the gene ID 1101 and the score 1101 is initialized to 0. Gene IDs already belonging to functional cluster C are s
If it is registered in core [i], the value of weight 1009 is added to score 1101 (step 1404).

Perform steps 1403 and 1404 for all biological functions selected in step 1400,
After that, i is incremented by 1 and the process returns to step 1402 (step 1406). Step 1
If i reaches crNum in 402, it means that all the scores have been calculated. When all are calculated, they are sorted in order of the score 1101 value of score [i], and the display device 802
Is displayed on the screen and the process is completed (step 1406). With the above processing, it is possible to evaluate and analyze the cluster analysis results as shown in FIGS. 3, 4, 5, and 7.

[0043]

As described above, according to the present invention,
From the results of a plurality of clustering, it is possible to evaluate how reliable each clustering method is and what kind of characteristics it has, by observing how genes with known biological functions are classified. In addition, it is possible to evaluate the degree of having a designated biological function by observing to which cluster a gene whose function is unknown belongs to a plurality of clustering results.
Such evaluation results help in confirming the function of a gene whose function is unknown.

[Brief description of drawings]

FIG. 1 is a diagram showing a display example of a hierarchical clustering result.

FIG. 2 is an explanatory diagram showing a difference in cluster size in hierarchical clustering.

FIG. 3 is a view showing a ratio of correctly discriminating the biological function.

FIG. 4 is a view showing a ratio of correctly discriminating that the biological function is not applicable.

FIG. 5 is a diagram showing the number of known genes correctly discriminated and the number of known genes contained in a cluster.

FIG. 6 is an explanatory diagram showing the relationship of genes estimated by a plurality of clustering methods.

FIG. 7 is a diagram showing scoring based on the plausibility of estimated genes.

FIG. 8 is a block diagram showing the configuration of a cluster result analysis system according to the present invention.

FIG. 9 is a diagram showing an example of a data structure of gene data.

FIG. 10 is a diagram showing a data structure of clustering result data.

FIG. 11 is a diagram showing a data structure of score calculation sequence data.

FIG. 12 is a flowchart showing an outline of processing of the cluster result analysis system according to the present invention.

FIG. 13 is a flowchart showing the details of the calculation processing of the existence hit ratio and the non-existence hit ratio.

FIG. 14 is a flowchart showing details of score calculation processing.

[Explanation of symbols]

801 ... Biochip database, 802 ... Display device, 803 ... Keyboard, 804 ... Mouse, 805 ... Central processing unit, 806 ... Program memory, 807 ... Clustering processing unit, 808 ... Hit ratio calculation processing unit, 80
9 ... Score calculation processing unit, 810 ... Analysis result display processing unit

   ─────────────────────────────────────────────────── ─── Continued front page    (72) Inventor Ryo Nakashige             6-81 Onoe-cho, Naka-ku, Yokohama-shi, Kanagawa             Hitachi Software Engineering Stock Association             In-house (72) Inventor Toshiko Matsumoto             6-81 Onoe-cho, Naka-ku, Yokohama-shi, Kanagawa             Hitachi Software Engineering Stock Association             In-house (72) Inventor Shingo Ueno             6-81 Onoe-cho, Naka-ku, Yokohama-shi, Kanagawa             Hitachi Software Engineering Stock Association             In-house (72) Inventor Takuro Tamura             6-81 Onoe-cho, Naka-ku, Yokohama-shi, Kanagawa             Hitachi Software Engineering Stock Association             In-house F-term (reference) 2G045 DA12 DA13                 4B063 QA08 QA18 QQ42 QQ52 QR08                       QR42 QR55 QR62 QS25 QS34                       QS39 QX01

Claims (8)

[Claims] 1. A gene clustering result evaluation method for evaluating a result of clustering a plurality of genes by a predetermined clustering method based on expression patterns thereof, wherein the clustering result is determined to have a predetermined biological function. Characterized in that the result of clustering is evaluated based on the ratio of the genes known to have the predetermined biological function among the genes whose biological functions belonging to the gene group are known. Method for evaluating clustering result of genes to be used. 2. A gene clustering result evaluation method for evaluating the result of clustering a plurality of genes by a predetermined clustering method based on the expression pattern thereof, wherein the clustering result does not have a predetermined biological function. For the determined gene group, it is possible to evaluate the clustering result based on the ratio of the genes known to have no predetermined biological function among the genes having known biological functions belonging to the gene group. A method for evaluating clustering results of characteristic genes. 3. A method for displaying a gene clustering result, which displays a result of clustering a plurality of genes by a predetermined clustering method based on their expression patterns, wherein the result of clustering determines that the gene has a predetermined biological function. The gene group characterized by calculating the ratio of genes known to have the predetermined biological function among genes whose biological function is known to belong to the gene group Clustering result display method. 4. A gene clustering result display method for displaying the result of clustering a plurality of genes by a predetermined clustering method based on the expression pattern thereof, wherein the clustering result does not have a predetermined biological function. Regarding the discriminated gene group, it is characterized by calculating and displaying a ratio of genes whose biological function is known to belong to the gene group and which are known not to have the predetermined biological function. Gene clustering result display method. 5. A method for displaying a gene clustering result, which displays a result of clustering a plurality of genes by a predetermined clustering method based on their expression patterns, wherein the clustering result is determined to have a predetermined biological function. Genes characterized by displaying, for each gene group, the number of genes having a known biological function belonging to the gene group and the number of genes known to have the predetermined biological function Clustering result display method. 6. A gene clustering result display method for displaying a result of clustering a plurality of genes by a predetermined clustering method based on the expression pattern thereof, wherein the clustering result has no predetermined biological function. For each of the determined gene groups, the number of genes having a known biological function belonging to the gene group and the number of genes known not to have the predetermined biological function are displayed. Gene clustering result display method to be used. 7. A method of extracting a candidate of a gene having a predetermined biological function from a plurality of genes, the method comprising clustering a plurality of genes by a plurality of clustering methods based on the expression pattern thereof. As a result of clustering by one clustering method, a gene determined to have the predetermined biological function is added with a weight set for the clustering method, and a step of adding the weight is used for all the remaining steps. The clustering result by the clustering method of
A method comprising: a step of calculating a score for each gene; and a step of extracting a gene candidate having the predetermined biological function based on the magnitude of the score. 8. A method of displaying a candidate of a gene having a predetermined biological function, which comprises: clustering a plurality of genes by a plurality of clustering methods based on their expression patterns; and As a result of the clustering, the step of adding the weight set for the clustering method to the gene determined to have the predetermined biological function, and the step of adding the weight according to all the remaining clustering methods. Run on the clustering results,
A method comprising: a step of calculating a score for each gene; and a step of arranging and displaying the genes together with the score of the gene in descending order of score.