JP5419145B2 - Aptamer classification device, aptamer classification method, program, and recording medium - Google Patents

Description

  The present invention relates to an aptamer classification device, an aptamer classification method, a program, and a recording medium.

  Aptamers are nucleic acid ligands that specifically bind to a target substance. The concept of aptamers was first reported in 1990 by GOLD et al. (Patent Document 1, Non-Patent Document 1). Aptamers are expected to be applied to pharmaceuticals and sensors. For example, pegaptanib sodium injection, which is a therapeutic agent for angiogenic age-related macular degeneration (AMD), has been developed and put to practical use as an aptamer drug. .

  In order to classify natural nucleic acid molecules that have acquired sequence diversity by accumulating mutations, a nucleic acid sequence classification method using a discriminating machine that has learned some data from known nucleic acid sequences is effective.

Japanese Patent No. 2763958

Science. (1990) 249, 505-510.

  However, as described below, in the case of an aptamer in which various sequences are artificially synthesized, the above-described general nucleic acid sequence classification method cannot be used. First, since aptamer functions are often realized by only a part of the sequence, depending on the nucleic acid sequence classification method based on the overall similarity of the base sequence, the binding mode with the target molecule, etc. This is because it is difficult to classify highly accurately according to the function. Second, when there are a large number of aptamers to be classified, the calculation cost at the time of classification such as memory capacity and calculation time is remarkably increased. Third, when the nucleic acid sequence classification method is applied to aptamer classification, regions important for aptamer function cannot be narrowed down, and important regions in aptamer sequences are not clarified after classification.

  Therefore, the present invention provides an aptamer classification device, an aptamer classification method, a program, and a recording medium that can classify aptamers with high accuracy, narrow down important features of aptamers, and classify a large number of aptamers at low cost. For the purpose.

In order to achieve the above object, the aptamer classification device of the present invention comprises:
An input means for inputting the aptamer sequence, a feature extraction means for extracting the feature of the aptamer sequence, a sequence classification means for classifying the aptamer sequence, and an output means for outputting the classification result of the aptamer sequence,
The feature extraction means executes the cluster formation step (A) below,
(A) Cluster formation step of extracting features of aptamer sequences inputted by the input means by a preset feature extraction method, and forming a plurality of clusters of aptamer sequences for each feature. ) Cluster integration step and (C) aptamer sequence classification step,
(B) calculating the similarity between the clusters with the highest similarity of the features;
Merge the clusters with the highest similarity into one cluster,
In one integrated cluster, a common feature is a feature of the cluster, and other features are deleted. Cluster integration step (C) The cluster integrated in the step (B) is regarded as one aptamer sequence. Then, when the step (B) is repeatedly executed and a preset termination condition is satisfied, the step (B) is terminated and the aptamer sequence of the obtained cluster and the aptamer sequence classification for determining the characteristics of the aptamer sequence are determined. Step The output means outputs the aptamer sequence determined in the step (B) and the characteristics of the aptamer sequence.
It is characterized by that.

Moreover, the aptamer classification method of the present invention includes:
It includes the following steps (A), (B) and (C).
(A) A cluster forming step of extracting features of the input aptamer sequence by a preset feature extraction method, and forming a plurality of clusters of aptamer sequences for each feature (B) The feature similarity is highest Calculate the similarity between clusters,
Merge the clusters with the highest similarity into one cluster,
In one integrated cluster, a common feature is a feature of the cluster, and other features are deleted. Cluster integration step (C) The cluster integrated in the step (B) is regarded as one aptamer sequence. Then, when the step (B) is repeatedly executed and a preset termination condition is satisfied, the step (B) is terminated and the aptamer sequence of the obtained cluster and the aptamer sequence classification for determining the characteristics of the aptamer sequence are determined. Process

The program of the present invention is
The aptamer classification method of the present invention is executed by an aptamer classification apparatus.

The recording medium of the present invention is
The program of the present invention is recorded.

  According to the present invention, an aptamer classification device, an aptamer classification method, a program, and a recording medium that can classify aptamers with high accuracy, narrow down important features of aptamers, and classify a large number of aptamers at low cost are provided. be able to.

FIG. 1 is a block diagram showing an embodiment of an aptamer classification device of the present invention. FIG. 2 is a flowchart showing an embodiment of the aptamer classification method and aptamer classification program of the present invention. FIG. 3 is a flowchart showing an embodiment of the aptamer classification method and aptamer classification program of the present invention. FIG. 4 is a block diagram showing an embodiment of the aptamer classification device of the present invention. FIG. 5 is a flowchart showing an embodiment of the aptamer classification method and aptamer classification program of the present invention. FIG. 6 is a diagram for explaining an embodiment of the aptamer classification method and the aptamer classification program of the present invention. FIG. 7 is a flowchart showing an embodiment of the aptamer classification method and aptamer classification program of the present invention. FIG. 8 is a graph showing the number of clusters formed in the process of aptamer sequence classification and the average hamming distance between the clusters in the aptamer classification method and aptamer classification program of the present invention. FIG. 9 shows the relationship between the number of clusters and the evaluation results of the ability to classify aptamer sequences, and the relationship between the number of clusters and the average of the Hamming distance between clusters in the examples of the aptamer classification method and the aptamer classification program of the present invention. It is a graph to show.

  Hereinafter, more specific embodiments of the present invention will be described with reference to the drawings. However, the present invention is not limited to the following embodiments.

[Embodiment 1]
FIG. 1A shows a configuration of an example of an aptamer classification device according to this embodiment. As shown in the figure, the aptamer classification device 10 includes an input unit 11, a feature extraction unit 12, a sequence classification unit 13, and an output unit 14. The feature extraction unit 12 and the array classification unit 13 may be incorporated in a data processing device 15 that is hardware as shown in FIG. The data processing device 15 may include a CPU, for example.

  The input means 11 is electrically connected to the feature extraction means 12. The feature extraction unit 12 is electrically connected to the array classification unit 13. The array classification unit 13 is electrically connected to the output unit 14.

  The input means 11 is a means for inputting an aptamer sequence to the feature extraction means 12. The input means 11 is not particularly limited in other respects, and for example, a normal input device, an input file, and another computer provided in a computer such as a keyboard and a mouse can be used. Further, the input means 11 may be means for reading an aptamer sequence stored in a database, for example, as shown in FIG. In this case, for example, as shown in the figure, the aptamer sequence stored in the server 16 in advance is called to the input means 11 through the network 17. The input means 11 may include a communication interface 111 as shown in FIG.

  The feature extraction unit 12 extracts information on aptamer sequences to be classified. The feature extraction means 12 extracts and associates the features of the aptamer sequence. The characteristics extracted at this time are not particularly limited, and examples thereof include the arrangement of base sequences, the appearance frequency, the estimated secondary structure, the content of each base, the appearance frequency of consecutive bases, and the like. The method for extracting the feature is not particularly limited, and can be performed by a conventionally known method. Examples thereof include a method using existing nucleic acid sequence analysis software and secondary structure prediction software.

  The cluster (initial cluster) of aptamer sequences formed by the feature extraction unit 12 based on the features is not particularly limited. For example, an aptamer group in which all the features of the aptamer sequences match can be used as a cluster. Good.

  The array classification unit 13 extracts information on the initial cluster from the feature extraction unit 12. Next, the similarity is determined between each cluster. The similarity may be determined using the extracted features, and is not particularly limited. For example, the number of features common to two clusters, the number of features common to the total number of features that each cluster originally had, The ratio of the total number is given. At this time, the degree of separation between clusters may also be determined. The degree of separation is an index indicating how well the cluster is classified, and the degree of separation is the highest when the classification is the best. The degree of separation is not particularly limited, and examples thereof include an average of feature distances between clusters and an average of feature correlations. The distance of the feature is not particularly limited, and examples thereof include a Hamming distance, a Manhattan distance, a Levenstein distance, and an Euclidean distance. The average is not particularly limited, and examples thereof include an average of weighted items, an average of uniformly treated items without applying weights, and the like. The average of the weights is not particularly limited, and examples thereof include an average of weights of the number of sequences included in the cluster, an average of weights of the strength of binding force for each sequence, and the like. When the degree of similarity and the degree of separation are determined, the group having the highest degree of similarity between clusters is collected as a new cluster in which only common features are inherited as new features. At this time, features that are not common between clusters are deleted. The term “deletion” means that the feature is not considered between the clusters. When a new cluster is formed, the procedure returns again to the similarity determination procedure between clusters, and the same procedure is repeated. This is repeated until there is no common feature between the clusters. When the repetition is completed, the array classification result is finally output to the output unit 14. The classification result output at this time follows the conditions input by the user. Conditions entered by the user include classification results when the degree of separation is maximum, classification results when the number of clusters falls below a certain number, classification results when the degree of cluster similarity falls below a certain level, etc. Is mentioned.

  The sequence classification means 13 determines the similarity between the clusters based on the features, determines a common feature among the clusters having the highest similarity, and reshapes the cluster based on the common features. Any means may be used. The array classification unit 13 is not particularly limited in other respects, but may include, for example, a central processing circuit (CPU) or may be operated by program control. In addition, when the aptamer classification apparatus of this embodiment is provided with an operation part, a display part, a memory, etc., the said arrangement | sequence classification means 13 may be electrically connected with these members, for example.

  The similarity between clusters determined by the sequence classification unit 13 based on the features is an index representing the degree of similarity between clusters, and may be determined using the extracted features. The arrangement classification means 13 is not particularly limited, and examples thereof include the number of features common to two clusters, the ratio of common features to the total number of features of each cluster, and the like.

  The sequence classification means 13 reforms a cluster of aptamer sequences based on the common feature. The cluster of the aptamer sequence is not particularly limited. For example, a cluster group in which all of the common features match among the clusters may be a new cluster. The sequence classification means 13 integrates the original cluster by this cluster reformation.

  Moreover, the said sequence classification | category means 13 may be provided with the function etc. which determine the isolation | separation degree between clusters, for example. The degree of separation is not particularly limited as long as it is an index representing the degree of cluster classification, and examples thereof include an average of feature distances between clusters and an average of feature correlations. The resolution is highest when the aptamers are best classified.

  In addition, the array classification unit 13 may have a function of discriminating the priority order in the array included in the cluster based on the distance from the center of gravity.

  The output unit 14 may be any unit that outputs the classification result of the aptamer sequence and is not particularly limited in other respects. For example, a normal output unit provided in a computer such as a display device or a printing device, an output file, and Other computers or the like can be used.

  Next, the aptamer classification method of the present embodiment will be described with reference to the flowcharts of FIGS. The aptamer classification method of this embodiment includes a cluster formation step (step A21), a cluster integration step (step A22), and an aptamer sequence classification step (step A23).

[Cluster formation process]
The feature extraction unit 12 extracts features of the input aptamer sequence by a preset feature extraction method, and forms a plurality of clusters of aptamer sequences for each feature (step A21).

[Cluster integration process]
The sequence classification means 13 calculates the similarity between the clusters with the highest similarity of the features, integrates the clusters with the highest similarity into one cluster, and is common in the integrated cluster. The feature to be made is the feature of the cluster, and other features are deleted (step A22). Details of step A22 will be described below with reference to FIG. The sequence classification means 13 takes out the sequence associated with the feature as an initial cluster (step A301). Next, a comparison source cluster is selected from the extracted initial clusters, and the similarity with the selected comparison destination cluster is determined (steps A302, A303, A304). If the similarity obtained as a result of the determination is greater than the maximum similarity of the comparison source cluster, the similarity obtained here is registered as the maximum similarity of the comparison source cluster, and the comparison destination cluster is the comparison destination cluster at that time. (Steps A305 and A306). Here, when there is a cluster whose similarity with the comparison source cluster is not determined, the cluster is selected as a new comparison destination cluster, and the procedure from step A303 is repeated (A307). When the similarity of all other clusters has been determined for the comparison source cluster, if there is a cluster that has not been selected as the comparison source cluster, that cluster is selected as the new comparison source cluster, and A302 or lower The above procedure is repeated (step A308). Next, for the combination that maximizes the similarity between the comparison source cluster and the comparison target cluster, if the similarity (maximum similarity) satisfies the condition, the features common to the two clusters are extracted, and the common features are extracted. Two clusters are collected as new feature clusters (steps A309, A310, A311, and A312). The condition is not particularly limited, and examples include that the maximum similarity is not less than a reference value and exceeds the reference value. In addition, this step may further include a step of determining the degree of separation between clusters.

[Aptamer sequence classification process]
The sequence classifying unit 13 regards the cluster integrated in the cluster integration step as one aptamer sequence, repeatedly executes the cluster integration step, and satisfies the preset termination condition, the cluster integration step is performed. The aptamer sequence of the cluster obtained and the characteristics of the aptamer sequence are determined (step A313). If the termination condition is not satisfied, the sequence classification unit 13 regards the cluster integrated in the cluster integration step as one aptamer sequence, and repeats the procedure from A301 onward. The preset termination condition is not particularly limited, but when the maximum similarity is less than the reference value or less than the reference value, when the total number of clusters is less than the reference value or less than the reference value, the number of sequences included in the cluster is For example, the number of cluster features is less than the reference value or less than the reference value. The above steps complete the classification of aptamer sequences.

  The aptamer sequences that have been classified may be output by the output means 14 in accordance with conditions specified by the user. The condition for the output means 14 to output the classification result is arbitrary and is not particularly limited. For example, the classification result when the degree of separation is maximum, the classification result when the number of clusters is less than or below a certain value, Examples include classification results when the cluster similarity is less than or below a certain level, classification results when the number of cluster features is below or below a reference value, and the like. Further, the output unit 14 may have a function of outputting the features, the cluster, the similarity, the priority, the separation, and the like, for example.

  According to the present embodiment, first, a plurality of aptamers can be classified with high accuracy based on the condition of sharing a specific feature. The reason is that not the aptamer sequence itself but features such as base frequency and secondary structure are extracted and clusters are formed between sequences having similar features. Aptamer properties are often influenced not by the sequence itself, but by the secondary structure formed by the sequence and the arrangement of some specific bases, so classification accuracy is higher than simple sequence similarity evaluation. Become.

According to the present embodiment, secondly, the cost for classification can be reduced. Reasons for reducing the cost include, for example, a reduction in memory capacity and a reduction in calculation time. The reason why the memory capacity can be reduced by the aptamer classification method of the present invention is as follows, for example. That is, in the case of a general nucleic acid sequence classification method, it is necessary to perform pair-wise alignment for all sequence pairs and hold the results. In this case, the required memory for the number n of arrays is _n C ₂ = n × (n + 1) / 2. On the other hand, in the aptamer classification method of the present invention, for each cluster, it is sufficient to hold the ID of the cluster having the highest similarity to the cluster and the similarity, and the required memory is n + n = 2n. Therefore, a general nucleic acid sequence classification method requires a memory in the order of the square of n with respect to the number of sequences n, whereas the aptamer classification method of the present invention is 1 for n with respect to the number of sequences n. A memory of order of power is sufficient. Therefore, according to the aptamer classification method of the present invention, the required memory capacity can be reduced, and this tendency is particularly remarkable when the number of sequences is large. The reason why the calculation time can be shortened by the aptamer classification method of the present invention is as follows, for example. That is, in the case of a general nucleic acid sequence classification method, it is necessary to directly compare the sequences. On the other hand, when determining the degree of similarity in the aptamer classification method of the present invention, only extracted features are to be compared, and simple processing that only determines the duplication of the features is sufficient. Moreover, as the classification progresses, the features are deleted as described above, and the number of features that need to be taken into consideration when determining the similarity is reduced. Therefore, according to the aptamer classification method of the present invention, the calculation time per sequence comparison can be shortened.

  Further, according to the present embodiment, third, important features are narrowed down among the features of the aptamer sequence. The reason is that, in the process of classifying aptamers, important features in each cluster are narrowed down by removing features that are not common between sequences with high similarity. As a result of narrowing down the important features, it becomes easy to process aptamers such as removing unnecessary regions in aptamer sequences and replacing them with modified nucleic acids.

  The program of the present invention is a program that causes the apparatus to execute the above-described method of using the aptamer classification device of the present invention. The program of the present invention may be recorded on a recording medium, for example. The recording medium is not particularly limited, and examples thereof include an HDD, a CD-ROM (CD-R, CD-RW), a DVD, and a memory card. The program of the present invention may be installed in advance in the above-described aptamer classification apparatus of the present invention, or may be installed via the recording medium or a network such as the Internet. Note that the program of the present invention does not necessarily have to be installed in the aptamer classification apparatus of the present invention described above. For example, the above-described usage method can be changed according to the program of the present invention stored in a server. May be executed.

[Embodiment 2]
FIG. 4 shows a configuration of an example of the aptamer classification device 40 of the present embodiment. In this figure, the same parts as those in FIG. As shown in FIG. 4, this aptamer classification device 40 has the same configuration as the aptamer classification device of Embodiment 1 except that it further includes sequence storage means 41 and feature storage means 42. That is, the aptamer classification device 40 includes an input unit 11, a feature extraction unit 12, a sequence classification unit 13, an output unit 14, a sequence storage unit 41, and a feature storage unit 42. The feature extraction unit 12 and the array classification unit 13 may be incorporated in a data processing device 15 that is hardware as shown in FIG. The feature extraction unit 12 and the arrangement classification unit 13 may be software, for example, or may be software in which software is incorporated. The data processing device 15 may include a CPU, for example. The array storage unit 41 and the feature storage unit 42 may be incorporated in a storage device 43 that is hardware as shown in FIG.

  As shown in the figure, the array storage means 41 is electrically connected to the input means 11 and the feature extraction means 12. The feature storage means 42 is electrically connected to the feature extraction means 12 and the arrangement classification means 13.

  The sequence storage unit 41 stores information on the sequence of aptamers to be classified, which is input from the input unit 11, and outputs the information to the feature extraction unit 12.

  The feature storage means 42 stores the features of the aptamer sequences extracted by the feature extraction means 12 in association with the respective sequences, and outputs the features to the sequence classification means 13.

  In this embodiment, for example, the aptamer sequences A, B, C, and D can be classified. Details will be described below.

  First, the input unit 11 outputs the data of the aptamer sequences A, B, C, and D to the feature extraction unit 12 (Step A1).

  The feature extraction unit 12 extracts features from the aptamer sequences A, B, C, and D (step A2). When the features a, b, c, d, and e are extracted as a result of the extraction, the feature extraction unit 12 sends the aptamer sequences A, B, C and D are output as clusters A, B, C, and D, respectively. In FIG. 6, the presence of a feature is represented by 1 and the absence of a feature is represented by 0.

  Hereinafter, the aptamer classification method of the present embodiment will be described with reference to FIGS. The array classification means 13 reads the array associated with the feature of FIG. 6A as cluster information (step A301, FIG. 5, hereinafter, the figure number is omitted). Next, cluster A is selected as the first comparison source, cluster B is selected as the first comparison destination, and the similarity between clusters A and B is determined (steps A302, A303, and A304). In the present embodiment, similarity is defined as the number of features common to two clusters. In this case, since the three features a, b, and e are common between the clusters A and B, the similarity is 3. This value is registered as the maximum similarity of the current comparison source cluster A, and cluster B is also registered as the comparison destination (steps A305 and A306). This completes the evaluation of cluster B with respect to cluster A.

  Since there are unevaluated clusters C and D as comparison destination clusters of the comparison source cluster A, the evaluation procedure of steps A303 to A306 with the comparison source cluster being A and the comparison destination cluster being C or D is similarly repeated (step A307). . At this time, since the similarities of the comparison destination clusters C and D with respect to the comparison source cluster A are 2 and 1, respectively, the maximum similarity is not updated. This completes the evaluation using the cluster A as the comparison source cluster.

  When the evaluation using the cluster A as the comparison source cluster is completed, the evaluation procedure of Steps A302 to A306 using the unevaluated clusters B, C, and D as comparison source clusters as the comparison source cluster is similarly repeated (Step A308). This completes the evaluation of all clusters.

After step A308, the degree of cluster separation was determined (step A501). In this embodiment, the degree of separation is defined as the average of the Hamming distance between clusters multiplied by the weight of the number of sequences included in the cluster. For example, in FIG. 6A, the hamming distances between clusters A-B, A-C, A-D, B-C, B-D, and C-D are 1, 3, 3 respectively. 2, 4 and 4, and each cluster contains one sequence, so the weight of the number of sequences in each cluster is 1/4. So in this case,
Separation degree = (1 × (1/4 + 1/4) + 3 × (1/4 + 1/4) + 3 × (1/4 + 1/4) + 2 × (1/4 + 1/4) + 4 × (1/4 + 1/4) +4 × (1/4 + 1/4)) / ₄ C ₂ = 1.42
Is determined.

  In FIG. 6 (a), the similarity between clusters A-B, A-C, A-D, B-C, B-D, and C-D is 4, 2, 2, 3 respectively. 1, 1. Therefore, in FIG. 9A, the comparison source cluster and the comparison destination cluster having the maximum maximum similarity are clusters A and B. In the present embodiment, it is assumed that the maximum similarity condition is 1 or more which is a reference value (step A502). Since the similarity between the clusters A and B is 3, which is 1 or more, which is the reference value, the features a, b, c, and e common to the clusters A and B are determined (steps A311 and A312). As a result of re-forming the cluster by the common features a, b, c, and e, the clusters A and B are merged into one as a cluster AB, resulting in the state shown in FIG. 6B. Note that the feature d not common to the clusters A and B is 0 in the cluster AB.

  In the present embodiment, the end condition is that the number of clusters is less than 2 which is a reference value (step A503). In FIG. 6B, the number of clusters is three, AB, C, and D, and is not less than 2, which is the reference value. Therefore, the procedure after step A301 is similarly repeated.

In FIG. 6B, the Hamming distances between clusters AB-C, AB-D, and CD are 2, 4, and 4, respectively, and the array weights of clusters AB, C, and D are 2 / 4, 1/4, and 1/4. So in this case,
Degree of separation = (2 × (2/4 + 1/4) + 4 × (2/4 + 1/4) + 4 × (1/4 + 1/4)) / ₃ C ₂ = 2.17
Is determined (step A501).

  Further, in FIG. 6B, the similarities between clusters AB-C, AB-D, and CD are 2, 1, and 1, respectively. Therefore, in FIG. 6B, the comparison source cluster and comparison destination cluster having the maximum maximum similarity are clusters AB and C (step A309). Since the similarity between the clusters AB and C is 2, which is 1 or more, which is the reference value of the maximum similarity in Step A311, the features a and b common to the clusters AB and C are determined (Step A311, A312). As a result of newly forming clusters based on the features a and b which are the logical products, the clusters AB and C are integrated into a single cluster ABC, resulting in the state shown in FIG. Note that the features c and e that are not common to the clusters AB and C are 0 in the cluster ABC.

  In FIG. 6C, the number of clusters is two, ABC and D, and is not less than 2, which is the reference value of the number of clusters in step A314, and thus the procedure from A301 is repeated in the same manner (step A503).

In FIG. 6C, the Hamming distance between the clusters ABC-D is 3, and the array weights of the clusters ABC and D are 3/4 and 1/4, respectively. So in this case,
Degree of separation = 3 × (3/4 + 1/4) / ₂ C ₂ = 3
Is determined (step A501).

  In FIG. 6C, since there is no common feature between the cluster ABC and the cluster D, the maximum similarity is 0, and is not 1 or more, which is the reference value of the maximum similarity in step A502. The process ends (step A502).

  When the classification is completed, the output means 14 outputs the classification result (step A4). The classification condition output condition specified by the user is arbitrary and is not particularly limited. For example, when the classification result when the cluster separation degree is maximum is the output condition, the separation degree is maximum during the classification of A3. The result shown in FIG. 6C is output as the classification result.

  As a result, aptamer sequences A, B, C, and D were classified according to this embodiment.

  Next, examples of the present invention will be described. However, the present invention is not limited by the following examples.

[Example 1]
In this example, 594 aptamer sequences for 17 targets were used as a correct data set, and the classification ability of aptamer sequences was evaluated. The aptamer sequence is registered in the aptamer database N30 having a random region length of 30 bases. The aptamer database N30 is available on the Internet (http://aptamer.icmb.utexas.edu/). Published by A specific classification method in this embodiment will be described below. First, in this example, the presence or absence of a combination of 3 bases allowing a 4 base mismatch to a 7 base window size was used as a feature extracted from the aptamer. In addition, as shown in FIG. 7, an end condition is added, and step A701 is provided after step A503. When the number of features of the cluster having the smallest number of features is less than the reference value (minimum inclusion number), the classification is finished. If the number of features is equal to or greater than the minimum inclusion number that is the reference value, the classification is continued. Other than that, classification was performed in the same manner as in the second embodiment. Details will be described below.

(Target judgment method)
According to the definition (A) described below, the target to which the classified cluster belongs was determined.
(A) Assuming that each target forms only one cluster, it is defined that the cluster to which each target sequence belongs most belongs to only that target.

(Evaluation method of classification results)
The classification results were evaluated by determining specificity, sensitivity, and accuracy according to the following rules. For example, when focusing on target A,
True Positive: Target A aptamer belongs to cluster of target A False False: Target A aptamer belongs to cluster other than target A (False Positive): target A Aptamers other than those belonging to the cluster of target A (False Negative): Assuming that the aptamer of target A does not belong to any cluster, the number of true positives is TP and the number of false negatives is FN. The number of false positives was defined as FP, and the specificity, sensitivity, and accuracy of the classification result were determined by the following calculation formula and definition.
Specificity = (TP / TP + FP)
Sensitivity = (TP / TP + FN)
Accuracy: geometric mean of specificity and sensitivity

  In Table 1, when the minimum inclusion number is changed, aptamer sequence classification is performed on a data set including all 594 sequences of 17 targets according to the above definition (A), and the sequence classification result is evaluated for each target. The evaluation result of aptamer sequence classification is shown.

  Table 1 shows that in this data set, the accuracy of aptamer sequence classification is almost the same as long as the minimum inclusion number is about 3 to 20.

  FIG. 8 is a graph showing the relationship between the number of clusters formed in the process of aptamer sequence classification and the average Hamming distance between the clusters at that time. The horizontal axis of the graph shows the number of clusters formed in the process of aptamer sequence classification, and the vertical axis of the graph shows the average of the Hamming distance between clusters. As shown in the graph, in the process of gradually narrowing down the number of clusters by integration by the neighborhood coupling method, it was confirmed that the average of the Hamming distances between the clusters was maximized around the number of clusters of 200 to 140. The minimum number of inclusions at this time is about 70 to 200. According to Table 1, the prediction accuracy at this time is high in specificity but low in sensitivity. However, since this evaluation is based on the above definition (A), in the actual analysis that each target forms a plurality of clusters, the prediction accuracy is the best when the number of clusters is 200 to 140. There is a possibility that

[Example 2]
Based on the assumption that each target forms a plurality of clusters, according to the following definition (B), the classification ability of aptamer sequences is evaluated in the same manner as in Example 2 except that the target to which the classified cluster belongs is determined. It was.
(B) Assuming that each target forms a plurality of clusters, when more than half of the aptamer sequences of each cluster are occupied by a single target sequence, the cluster is defined as belonging to the target.

  FIG. 9 is a graph showing the relationship between the number of clusters and the evaluation result of the aptamer sequence classification ability, and the relationship between the number of clusters and the average of the Hamming distance between clusters when the above definition (B) is followed. is there. The horizontal axis of the graph indicates the number of clusters formed in the process of aptamer sequence classification, and the vertical axis of the graph indicates the average Hamming distance between clusters or the specificity that is the evaluation result of the ability to classify aptamer sequences. Shows sensitivity and accuracy.

  As shown in FIG. 9, when the clusters are classified in the present example, the prediction accuracy is also maximized in the vicinity of the number of clusters of 200 where the average of the Hamming distance between the clusters is maximum. Therefore, even when classifying unknown aptamer sequences for which the correct cluster is not known using the method of the present invention, it is suggested that the best clustering result can be obtained by dividing by the number of clusters with the maximum Hamming distance. It was done. In addition, as a result of these analyses, it was found that each target belongs to multiple clusters. Therefore, the following analysis is based on the assumption that each target forms multiple clusters. The definition (B) is adopted.

  Further, if aptamer sequences are classified using the method of the present invention based on the definition (B), aptamers for the same target molecule may be classified into different clusters. In this case, even when aptamers for the same target molecule are different in the clusters, it is clear that the features are different, such as different binding modes, and an effect is obtained that new information on the properties of the aptamer sequences can be obtained. As a result, the efficiency of the experiment for investigating the properties of the acquired sequence increases, and it becomes easy to acquire aptamers with different functions.

[Comparative Example 1]
Hereinafter, as Comparative Example 1, analysis using ClustalW (Julie D. Thompson et al. Nucleic Acids Research, 1994, 22, 22 4673-4680) widely known as base sequence analysis software was performed. Since ClustalW is a tool for performing hierarchical clustering, in order to classify aptamers into individual clusters, it is necessary to divide the phylogenetic tree on some basis. This time, we focused on the genetic distance of each sequence, and estimated the parameter with the highest prediction accuracy by changing it.

  Table 2 below shows the specificity, sensitivity, and accuracy of aptamer classification for each target when Example 2 and Comparative Example 1 were the most accurate. In Table 2, the average means the average (weighted average) of the aptamer classification specificity, sensitivity, and accuracy for each target, weighted by the number of sequences. According to Table 2, it turned out that Example 2 shows the precision superior to the comparative example 1 about all the targets. Moreover, the average value of accuracy increased by 26 points (46% of Comparative Example 1) with respect to Comparative Example 1. From this result, it was shown that the method of this example is more accurate in aptamer classification than the method using sequence analysis software.

  The calculation costs in aptamer sequence classification in Example 2 and Comparative Example 1 are shown in Table 3 below.

  According to Table 3, in Comparative Example 1, the memory capacity required for aptamer sequence classification increases in proportion to the square of the number of sequences. On the other hand, the memory capacity required in the second embodiment increases in proportion to the first power of the number of arrays. Therefore, it was shown that Example 2 can analyze a larger amount of sequences. Moreover, in Example 2, the calculation time required for aptamer sequence classification was about 1/5 of the calculation time required in Comparative Example 1, indicating that the calculation time was also significantly reduced. From this result, it was shown that the calculation cost of the method of the present invention is significantly reduced as compared with the method using the sequence analysis software.

  According to the present invention, aptamers can be classified efficiently. In addition, according to the present invention, it is possible to select features that are important for the function of an aptamer, and information that is important for designing a new aptamer can be obtained.

10, 40 Aptamer classification device 11 Input means 12 Feature extraction means 13 Array classification means 14 Output means 15 Data processing device 16 Server 17 Line network 41 Array storage means 42 Feature storage means 43 Storage device 111 Communication interface

Claims (15)

An input means for inputting the aptamer sequence, a feature extraction means for extracting the feature of the aptamer sequence, a sequence classification means for classifying the aptamer sequence, and an output means for outputting the classification result of the aptamer sequence,
The feature extraction means executes the cluster formation step (A) below,
(A) Cluster formation step of extracting features of aptamer sequences inputted by the input means by a preset feature extraction method, and forming a plurality of clusters of aptamer sequences for each feature. ) Cluster integration step and (C) aptamer sequence classification step,
(B) calculating the similarity between the clusters with the highest similarity of the features;
Merge the clusters with the highest similarity into one cluster,
In one integrated cluster, a common feature is a feature of the cluster, and other features are deleted. Cluster integration step (C) The cluster integrated in the step (B) is regarded as one aptamer sequence. Then, when the step (B) is repeatedly executed and a preset termination condition is satisfied, the step (B) is terminated and the aptamer sequence of the obtained cluster and the aptamer sequence classification for determining the characteristics of the aptamer sequence are determined. Step The output means outputs the aptamer sequence determined in the step (B) and the characteristics of the aptamer sequence.
An aptamer classification device characterized by that.   The aptamer classification device according to claim 1, wherein the similarity is the number of features common to two clusters.   The aptamer classification apparatus according to claim 1, wherein the sequence classification unit further executes a step of determining a degree of separation between clusters.   The aptamer classification device according to claim 3, wherein the degree of separation is an average of feature distances between clusters.   The aptamer classification device according to claim 4, wherein the distance is a Hamming distance.   The aptamer classification apparatus according to claim 4 or 5, wherein the average is an average of weights of the number of sequences included in the cluster.   The apparatus according to any one of claims 1 to 6, further comprising: a sequence storage unit that stores the aptamer sequence; and a feature storage unit that stores the feature of the aptamer sequence in association with the aptamer sequence information. The aptamer classification device described. An aptamer classification method comprising the following steps (A), (B) and (C):
(A) A cluster forming step of extracting features of the input aptamer sequence by a preset feature extraction method, and forming a plurality of clusters of aptamer sequences for each feature (B) The feature similarity is highest Calculate the similarity between clusters,
Merge the clusters with the highest similarity into one cluster,
In one integrated cluster, a common feature is a feature of the cluster, and other features are deleted. Cluster integration step (C) The cluster integrated in the step (B) is regarded as one aptamer sequence. Then, when the step (B) is repeatedly executed and a preset termination condition is satisfied, the step (B) is terminated and the aptamer sequence of the obtained cluster and the aptamer sequence classification for determining the characteristics of the aptamer sequence are determined. Process   The aptamer classification method according to claim 8, wherein the similarity is the number of features common to two clusters.   The aptamer classification method according to claim 8 or 9, wherein the step (B) further includes a step of determining a degree of separation between clusters.   The aptamer classification method according to claim 10, wherein the degree of separation is an average of feature distances between clusters.   The aptamer classification method according to claim 11, wherein the distance is a Hamming distance.   The aptamer classification method according to claim 11 or 12, wherein the average is an average of weights of the number of sequences included in the cluster.   A program capable of executing the aptamer classification method according to any one of claims 8 to 13 on a computer.   15. A recording medium on which the program according to claim 14 is recorded.

Images