Compound, or salt of the same, and guanine quadruplex binding ligand

Abstract

To provide a compound used for detection of a guanine quadruplex (G4) forming sequence, or a salt thereof, and a G4 binding ligand.SOLUTION: A G4 forming sequence prediction device includes: a determination part for determining a feature vector exhibiting a feature of a base sequence expressed by 5'-Ny0Gx1 Ny1Gx2 Ny2Gx3 Ny3Gx4 Ny4-3' (in the base sequence, Ny0-Ny4 expresses any sequence fragment of a base number of 0 or more, and Gx1-Gx4 expresses a sequence fragment including only G of a base number of 2 or more) from sequence data of the base sequence; and a prediction part for predicting a formation state of a guanine quadruplex in the base sequence from the feature vector using a prediction model. The prediction model is built by performing machine learning using teacher data that is a set of a feature vector exhibiting a feature of the base sequence, and G4 formation state information exhibiting a formation state of the guanine quadruplex in the base sequence.SELECTED DRAWING: None

Description

[Technical field]

[0001] The present invention relates to a method for determining whether a nucleotide sequence forms a G-quadruplex (whether a nucleotide sequence forms a G-quadruplex). G-quadruplexes at base sequences, such as whether they form a G-quadruplex or not, and the stability of G-quadruplexes at base sequences A device for predicting the formation state of a G-quadruplex and a prediction model for use in the device A machine learning method for constructing a database, and a compound or The present invention relates to a compound comprising the above-mentioned G-quadruplex-binding ligand and a salt thereof. [Background technology]

[0002] Conventionally, a G-quadruplex (hereinafter sometimes abbreviated as “G4”) is shown in FIG. As shown in b), the π-π interaction between guanine bases and hydrogen bonds are the driving forces behind the formation of More than 20 types of proteins that bind to G4 have been identified. It is a dynamic structure that is repeatedly formed and dissociated in the body. As shown in Figure 9a), A base sequence that forms a G4 in the coding region (hereinafter, abbreviated as "G4 forming sequence") When a low molecular weight compound that stabilizes G4 (G4 ligand) is applied to the ribosomal The ribosome stalls at the G4 formation site, dissociates from the mRNA, and translation is inhibited. G4 formed by NA inhibits transcription and replication by blocking polymerase elongation. Therefore, G4 has attracted attention as a characteristic nucleic acid higher-order structure and is involved in protein expression and replication. It is expected to become a new drug discovery target.

[0003] Clinical development of anticancer drugs based on the cancer gene suppression activity targeting G4 is progressing overseas. Meanwhile, in the field of antiviral drug discovery, several types of RNAs, including Corona, Retro, and Flavi, are being developed. G4-forming sequences were found in the A virus genome, and the virus was induced by treatment with a G4 ligand. Inhibition of viral proliferation and viral protein synthesis has been reported. The mechanism by which G4 is expressed is being proven. However, in (+)ssRNA viruses, the genome and mRNA have the same sequence, so the G4 tag is A targeted drug discovery strategy could block all stages of the viral life cycle, including transcription, translation, and replication. This allows for high drug selectivity against viruses.

[0004] G4 is in equilibrium with single-stranded sequences and is involved in the timing of biological events such as replication and transcription. G4s are thought to form dynamically and play important roles in biological processes. We are developing molecular tools to detect G4 formation in vivo and to predict the state of DNA / There is a need for technology to predict the higher-order structure of RNA. Currently, there are molecular tools to detect G4 and For example, small molecular weight compounds that interact with G4 have been reported. A G4-binding ligand that emits fluorescence only when it acts has been reported (Patent Document 1).

[0005] In response to this, there are prediction techniques for the higher-order structure of DNA / RNA, particularly for the higher-order structure of RNA. Many G4 formation sequence prediction systems that use secondary structure prediction models have been reported. The prediction model proposed by these authors is based on typical base pairing (AT,GU). A prediction model was applied to predict the formation state of G4s consisting of non-canonical base pairs. Sequence prediction devices have also been reported in recent years. For example, the G-quartet There have been reports of methods to extract only G4, which is a set of three or more overlapping planes (Non-Patent Documents 1 and 2 In addition, G4s with two or more G-quartet planes were extracted and the length and evenness of the loop bases were evaluated. It has been reported that a score is given as the probability of G4 formation by performing the above-mentioned procedure (Non-Patent Document 3). ) and also shows the probability of G4 formation based on the G (guanine) and C (cytosine) content. It has been reported that a method for scoring is used (Non-Patent Document 4). It has also been reported that the incomplete G4 gene can be extracted (Non-Patent Document 5). [Prior art documents] [Patent documents]

[0006] [Patent Document 1] JP 2020-176108 A [Non-patent literature]

[0007] [Non-Patent Document 1] Nucleic Acids Res. 2005, 33, 2908-2916. [Non-Patent Document 2] Nucleic Acids Res., 2005, 33, 2901-2907. [Non-Patent Document 3] Nucleic Acids Res., 2006, 34, 676-682. [Non-Patent Document 4] Nucleic Acids Res. 2016, 44, 1746-1759. [Non-Patent Document 5] Biochimie., 2017, 135, 54-62. Summary of the Invention [Problem to be solved by the invention]

[0008] However, the G4 formation sequence prediction system was applied to predict the G4 formation state. There have been few reports on G4 formation, and it is unclear what kind of base sequence forms G4 or what base sequence is involved. It has been difficult to accurately predict the stability of G4s. The prediction device uses the number of layers of the G-quartet plane and the loop arrangement composition based on the experimental researcher's empirical rules. A prediction model based on the G (guanine) content and Tether sequence combination was used. By integrating the number of occurrences of G-quartet planes, the number of G4s in the base sequence can be statistically and objectively determined. There is no predictive model that evaluates the presence or absence of G4 formation and the stability of G4 and predicts the state of G4 formation. This is because there are not enough G4-forming sequences to perform sufficiently reliable statistical procedures. This is due to the lack of experimental data sets collected under a single set of conditions. In addition, the G4 forming sequence (G3N 1-7 )3G3 and other arrangement rules However, several G4-forming sequences that deviate from these sequence rules have been reported. Furthermore, previous models for predicting the formation of G4 have been limited to the G (guanine) position. The prediction accuracy was low for sequences with low content of guanine. In recent years, it has been reported that G4 can be formed even in base sequences where the base pair is less than 50% of the original length. Therefore, even for base sequences with a low G (guanine) content, the presence or absence of G4 formation and the There is a need for a G4 formation sequence prediction device that can predict the G4 formation state, such as the stability, with high accuracy. There are.

[0009] The present invention has been made in view of the above points, and its object is to provide a method for producing a base A device for predicting G4 formation in a sequence, capable of predicting the state of G4 formation in the sequence with high accuracy, and the device A machine learning method for constructing a predictive model for the positioning of G4-forming sequences and a compound for detecting G4-forming sequences The object of the present invention is to provide a compound or a salt thereof, and a G4-binding ligand. [Means for solving the problem]

[0010] Therefore, in the present invention, the inventors have used machine learning to identify G4-forming sequences experimentally identified by the inventors, and have developed a method to identify G4-forming We created a sequence prediction system to predict G4 formation sequences. We verified the results of the prediction device by conducting a verification experiment of G4 formation using a binding ligand. By testing the results and relearning them, the prediction accuracy of the G4 formation sequence prediction device was improved.

[0011] In order to solve the above problems, the present invention provides a device for predicting G-quadruplex formation sequences, y0 G x1 N y1 G x2 N y2 G x3 N y3 G x4 N y4 -3' (nucleotide sequence Medium, N y0 ~N y4 represents any sequence fragment with 0 or more bases, and G x1 ~G x4 is the number of bases A feature vector showing the characteristics of the above base sequence is A determination unit that determines from sequence data and a prediction model are used to determine from the feature vector. a prediction unit for predicting a state of G-quadruplex formation in the base sequence from the prediction model; The delta is a feature vector that indicates the characteristics of the base sequence and the G-quadruplex in the base sequence. Machine learning is performed using training data that is a set of G4 formation state information that indicates the chain formation state. The present invention is characterized in that the prediction model is constructed by: Effect of the Invention

[0012] According to the present invention, the state of G4 formation in a base sequence can be predicted with high accuracy. [Brief description of the drawings]

[0013] [Figure 1] FIG. 1 is a diagram for explaining an outline of a G4 formation sequence prediction device according to a first embodiment. [Diagram 2] FIG. 1 is a diagram illustrating a schematic configuration of a computer that realizes a G4 formation sequence prediction device according to the first and second embodiments. [Diagram 3] FIG. 13 is a diagram for illustrating an outline of a G4 formation sequence prediction device according to a second embodiment. [Figure 4] FIG. 1 is a diagram showing an outline of a process flow for predicting whether or not a DNA sequence is a G4 forming sequence when evaluating the accuracy of the prediction device constructed in Example 1. [Diagram 5] 1 is a table showing the results of determining the prediction results of the prediction device constructed in the first embodiment through a verification experiment. [Figure 6] This is a table showing the dissociation constant Kd values ​​of the training data and the predicted stability values ​​of G4 when the stability of G4 in a DNA sequence is predicted using the prediction device of Example 2, together with the predicted value of the presence or absence of G4 formation obtained when the presence or absence of G4 formation in a DNA sequence is predicted using the prediction device of Example 1. [Figure 7] 13A is a graph showing the predicted value of G4 stability versus log(Kd[M]), which is the logarithm of the dissociation constant Kd value of the teacher data, when the stability of G4 in a DNA sequence is predicted using the prediction device of Example 2; and FIG. 13B is a graph showing the predicted value of the presence or absence of G4 formation versus log(Kd[M]), which is the logarithm of the dissociation constant Kd value of the teacher data, when the prediction device of Example 1 is used to predict the presence or absence of G4 formation in a DNA sequence. [Figure 8]FIG. 1 shows graphs in which the solid lines represent the UV absorption difference spectra of six types of G4-forming sequences (NCOR2, telo24, BIN1, OTP, WT1, and ZN318) before the addition of compound L2H2-6OTD, and the dashed lines represent the UV absorption difference spectra of three types of sequences in the high Kd value group (OTP, WT1, and ZN318) after the addition of compound L2H2-6OTD. [Figure 9] FIG. 1 is a diagram that illustrates the structure of a guanine quadruplex (G4), and a schematic diagram of the phenomenon that occurs when a G4-binding ligand, which is a low-molecular-weight compound that stabilizes G4, is applied to a G4-forming sequence. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0014] The following describes the G4 formation sequence prediction device and machine learning method of the present invention, as well as the compound or a salt thereof, and an embodiment relating to a G4 binding ligand. " refers to a G-quadruplex.

[0015] A. First embodiment Regarding the embodiment of the present invention, the G4 formation sequence prediction device and the machine according to the first embodiment The learning method and G4 binding ligands are illustrated.

[0016] <G4 formation sequence prediction device> FIG. 1 is a diagram for explaining the outline of the G4 formation sequence prediction device according to the first embodiment. The configuration of a computer that realizes a G4 formation sequence prediction device according to the first embodiment is shown in FIG. FIG.

[0017] As shown in FIG. 1, the G4 formation sequence prediction device 10 according to the first embodiment includes an input unit 2 and a The apparatus includes a memory unit 4, a processing unit 6 including a determination unit 6a and a G4 stability prediction unit 6b, and an output unit 8. The G4 formation sequence prediction device 10 is implemented, for example, by a computer 100 shown in FIG. The computer 100 includes a processing unit (computer) 110, a storage unit 120, and an input The device 130 includes an output device 140, an input / output interface (I / F) 150, and the like.

[0018] The input part of the G4 formation sequence prediction system is 5'-N y0 G x1 N y1 G x2 N y2 G x3 N y 3G x4 N y4 -3' DNA sequence (base sequence) (in the DNA sequence, N y0 ~N y 4 represents any sequence fragment with 0 or more bases, and G x1 ~G x4 Only G with 2 or more bases The sequence data of the sequence fragment containing the sequence is input. The memory unit stores prediction models and other The determination unit of the processing unit determines a feature vector x that indicates the characteristics of the DNA sequence. The G4 stability prediction part of the processing section uses a prediction model to determine the G4 stability from the sequence data of the DNA sequence. By doing this, we predict the stability of G4 in a DNA sequence from the feature vector x. The output part is Outputs the prediction results from the G4 stability prediction unit.

[0019] The prediction model stored in the memory is a feature vector x i and D.N. Dissociation constant K of compound 6OTD-Py of sequence A d Using the training data, which is a pair of The compound 6OTD-Py was developed by performing machine learning on the The prediction model is a G4-binding ligand according to the first embodiment described below. It was built using machine learning methods that

[0020] By using the G4 formation sequence prediction device according to the first embodiment, it is possible to predict the stability of G4 in a DNA sequence. In the method for predicting the stability of G4, first, the 5'-N y0 G x 1N y1 G x2 N y2 G x3 N y3 G x4 N y4 -3' DNA sequence (DNA sequence In row, N y0 ~N y4 represents any sequence fragment with 0 or more bases, and G x1 ~G x4 is a base A sequence fragment containing only Gs of two or more is prepared.

[0021] Next, sequence data of the DNA sequence is input via the input unit of the G4 formation sequence prediction device. The input sequence data is stored in the storage unit by the processing unit.

[0022] Next, a feature vector showing the characteristics of the DNA sequence is generated by the determination unit of the processing unit of the G4 formation sequence prediction device. The feature vector x is determined from the sequence data. The determined feature vector x is stored in the memory by the processing unit. The method of determining the feature vector x from the array data is described in the first embodiment. The feature vector x in the machine learning method according to the embodiment i The method is similar to the method for determining

[0023] Next, the G4 stability prediction unit of the processing unit of the G4 formation sequence prediction device predicts the Using a predictive model, predict the stability of G4 in a DNA sequence from a feature vector x. Specifically, the regression equations of the prediction model represented by the equations (3-1) and (3-2) described below are The value calculated by inputting the feature vector x into f(x) is the compound 6OTD of the DNA sequence. -Py dissociation constant K d The predicted value of [M] is obtained.

[0024] Next, the output section of the G4 formation sequence prediction device outputs the predicted result by the G4 stability prediction section. , the dissociation constant K of the compound 6OTD-Py in the DNA sequence d Output the predicted value of the dissociation constant K d According to the predicted value of, the lower the predicted value, the higher the possibility of G4 formation in the DNA sequence, and the higher the probability of D The higher the predicted value, the higher the stability of the G4 formed in the NA sequence. The possibility of G4 formation in the A sequence is low, and even if G4 is formed in the DNA sequence, The stability of the G4 formed in the sequence is predicted to be low.

[0025] <Machine learning method> A machine learning method according to the first embodiment will be described. The method includes a teacher data acquisition process and a machine learning process.

[0026] [Teacher data acquisition process] In the teacher data acquisition process according to the first embodiment, first, 5'-N y0 G x1 N y1 G x 2N y2 G x3 N y3 G x4 N y4 -3' DNA sequence (base sequence) In row, N y0 ~N y4 represents any sequence fragment with 0 or more bases, and G x1 ~G x4 is a base A feature vector that represents a sequence fragment containing only Gs with a number of 2 or more, and a DNA sequence The dissociation constant K of compound 6OTD-Py dThe pair of values ​​[M] is obtained as training data.

[0027] [Selection of G4-forming and non-G4-forming sequences for training data] In the training data acquisition process, specifically, first, a DNA microarray obtained in advance is Regarding the various types of DNA sequences arranged, the G4 bonds according to the first embodiment described below are A method is used to detect the presence of G4-forming base sequences in a sample using a cyclic ligand. By this, 2000 types of sequences to which G4-binding ligands bind were identified as G4-forming sequences (G 4) and 2000 types of sequences to which G4-binding ligands do not bind. These are detected as non-G4 forming sequences (sequences that do not form G4). For the sequence, we will analyze the DMS footprint, EMSA, CD spectrum, etc. Using the method described above, we experimentally verify whether G4 formation occurs. Of the 2,000 sequences that were generated, 130 were experimentally verified to form G4. We selected sequences and experimentally identified G4-forming sequences from 2,000 sequences that were detected as non-G4 forming sequences. Select 130 types that have been verified not to form G4. Among the sequences that are detected, 5'-N y0 G x1 N y1 G x2 N y2 G x3 N y3 G x4 N y4 Sequences not represented at -3' cannot be included, but sequences that are experimentally verified not to form G4 Among them, 5'-N y0 G x1 N y1 G x2 N y2 G x3 N y3 Gx4 N y4 -3' Therefore, among the sequences that have been experimentally verified not to form G4, In the 130 selections, 5'-N was selected from the sequences that were verified not to form G4. y0 G x1 N y1 G x2 N y2 G x3 N y3 G x4 N y4 - Exclude sequences not represented in 3', 5 '-N y0 G x1 N y1 G x2 N y2 G x3 N y3 G x4 N y4 -3' of the sequence Next, 130 sequences were experimentally verified to form G4 and selected. Among them, the dissociation constant K of compound 6OTD-Py d The value of K d ≦1×10 -5.2 Become M 83 types of sequences are selected as G4 forming sequences for training data. Then, G4 is experimentally Among 130 sequences that were verified to not form schizophrenia, the compound 6OTD-Py was selected. separation constant K d The value of K d >1×10 -5.2 The 89 sequences that will become M are used as training data for G4 Select as non-forming sequence.

[0028] [Acquisition of teacher data] Next, we used 83 types of training data for the 83 types of G4-forming sequences. We obtained 89 types of training data for 89 types of G4 non-forming sequences. By acquiring each of these, a total of 172 types of training data will be obtained. The training data for the sequence is a feature vector x i and, The dissociation constant K of the compound 6OTD-Py of the G4-forming sequence d We obtain a pair of values ​​of various G As training data for the G4 non-forming sequence, a feature vector showing the characteristics of the G4 non-forming sequence is used. x i and the dissociation constant K of the compound 6OTD-Py of the G4 non-forming sequence d Get the pair with the value of In this case, the G4-forming sequence in the training data for various G4-forming sequences is Dissociation constant K of compound 6OTD-Py d The value of, for example, the oligonucleotide of the G4-forming sequence The actual value can be set by a general method for the thiol. The compound 6OTD-Py of the G4 non-forming sequence in the training data for the G4 non-forming sequence Dissociation constant K d The value of is, for example, The dissociation constant K d Considering the possibility that the value of may not be accurately obtained, we set the The dissociation constant K of the compound 6OTD-Py with the G4 non-forming sequence in the training data d of value, the dissociation constant K d The value is 1×10 -5.2 One larger than M A fixed value may be set.

[0029] Here, various DNA sequences of 83 types of G4 forming sequences and 89 types of G4 non-forming sequences were Feature vector x iA method for determining the above will be described.

[0030] A feature vector x that indicates the characteristics of various DNA sequences i The method of determining the Based on this, we extract the feature vector x from the DNA sequence data. i Ingredient x in Get the specifics. Specifically, the stability of G4 in DNA sequences is affected by any of the five base sequences. N for ka y0 ~N y4 Along with the homology of the sequence fragments, x1 ~G x4 Base G in Since the number of copies is expected to be an important factor, N for any 5-base sequence y0 ~N y4 The homology score of the sequence fragment and G x1 ~G x4 The length of the sequence fragment and the feature vector x i Ingredient x in Obtain as.

[0031] N for any 5-base sequence y0 ~N y4 The homology score of the sequence fragment and The method of obtaining the sequence and the sequence of all five bases are described below. How to arrange five letters selected from the four letters A, T, C, and G. Because it is a type 4 5 = 1024 types. Therefore, for any arbitrary 5-base sequence, N y0 ~N y4 When obtaining the homology score for each sequence fragment, 1024 types of This will give a score for the homology of each sequence fragment to any 5-base sequence. , 1024 scores are obtained. Therefore, N y0 ~N y4 The homology score for each sequence fragment is 1024 × 5 = 5120 scores. These 5120 scores are then calculated as the feature vector x i 5120 pieces Ingredient x in It becomes.

[0032] In addition, the N y0 ~N y4 Homology score for each sequence fragment If you want to obtain any 5 salts by using the Smith-Waterman method, After aligning each sequence fragment with respect to the base sequence, the sequence is If the characters at the same position in each sequence fragment match (match(m)), a score of +2 is given. If the characters in the same position of the string fragment do not match (mismatch(i)), the score is -1. When a character is deleted in the base sequence or each sequence fragment (gapopen(o)), a gap of -1 is added. Core, any 5-base sequence or sequence fragment with consecutive character deletions (gapex on(x)) is given a score of -0.1, and the total score is calculated for all adjacent putts. The system finds the arrangement pattern that maximizes the total score among all the patterns. The maximum value of the total is the N y0 ~N y4 The homology of each sequence fragment Get it as a score.

[0033] Here is an example of how to obtain a homology score. For example, a base sequence of 5 bases, "ATGCA " for N y0 ~N y4An example of each sequence fragment, "ATCCGCA", is shown in Table 1 below. When juxtaposed, the sum of the relationship scores for characters in the same position is 2.9.

[0034] [Table 1]

[0035] Then, for an arbitrary 5-base sequence, such as "ATGCA", y0 ~N y4 Each distribution In all alignment patterns of the sequence fragment "ATCCGCA", The alignment pattern that maximizes the total score of the character relationships is shown in Table 2 below. The total score of the relationships of the letters in the same position in the juxtaposition pattern is 8.9. Therefore, for an arbitrary 5-base sequence such as "ATGCA", y0 ~N y4 Each array of An example fragment "ATCCGCA" receives a homology score of 8.9.

[0036] [Table 2]

[0037] On the other hand, G x1 ~G x4 The length of each sequence fragment is the number of G bases contained in the fragment. Yes. x1 ~G x4 To obtain the length of a sequence fragment, use G x1 ~G x4 Four types: The base lengths of the sequence fragments are obtained, and the base lengths of these four types of sequence fragments are Ktor x i 4 components of x in It becomes.

[0038] Feature vector xi Ingredient x in In the acquisition of the above, all the five arbitrary bases acquired as above are N for the base sequence y0 ~N y4 The homology score of the sequence fragment (5120 components x i n ) and G x1 ~G x4 The length of the sequence fragment (four components x in ) by integrating , feature vector x i 5124 components x i1 ~x i5124 This will give you A feature vector x that indicates the characteristics of the DNA sequence i 5124 components x i1 ~x i5124 A 5124-dimensional feature vector x i Determine.

[0039] [Machine learning process] Next, in the machine learning process, the above 5124-dimensional feature vector x i A total of 1 By using machine learning with 72 types of training data, 5'-N y0 G x1 N y1 G x 2N y2 G x3 N y3 G x4 N y4 -3' DNA sequence (in the DNA sequence, N y0 ~N y4 represents any sequence fragment with 0 or more bases, and G x1 ~G x4 is a sequence of 2 or more bases A prediction model was constructed to predict the stability of G4 in In this case, specifically, SVR (supported virtual reality) with RBF kernel is used. By using a vector regression method, machine learning was performed using a total of 172 types of training data. By doing so, we constructed a predictive model that constructs a regression equation to predict the stability of G4 in DNA sequences. The following is the process flow for constructing a predictive model using SVR with RBF kernel. This will now be explained in detail.

[0040] In the process of constructing a predictive model, first, a 5124-dimensional feature vector x i A total of 172 types of training data were used. i (i=1~N, N=172) Lange multiplier α i and α i * The object is represented by the following formulas (1-1) and (1-2) in which Function L D (α) is the maximum within the range that satisfies the constraints expressed by the following formulas (2-1) to (2-3). The largest Lagrange multiplier α i and α i * Machine learning is used to determine the combination of In machine learning, the RBF kernel parameter γ is set in the range of 0.001 to 1000. Set the cost parameter C to the optimal value and set it to the optimal value within the range of 0.001 to 1000. value and the objective function L D The Lagrangian for which (α) is maximized within the range where the constraints are satisfied. Multiplier α i and α i * The combination of 172 types of training data Tr i of The corresponding difference of Lagrange multipliers α i -α i * is α i -α i * Especially those that are ≠ 0 These are called support vectors.

[0041]

number

[0042]

number

[0043] N: Number of training data (172) α i , α i * :Lagrange multiplier for each training data x i : 5124-dimensional feature vector for each training data y i : Dissociation constant K of each training data d Value of ε: Error deadband parameter γ: RBF kernel hyperparameter C: Cost parameters

[0044] Next, the Lagrange multiplier α determined by the above machine learning is i and α i * Set of By using matching, G4 in the predicted DNA sequence other than the training data sequence can be predicted. The regression equation for predicting the stability of f(x) is expressed by the following equations (3-1) and (3-2). Enter the feature vector x, which indicates the characteristics of the DNA sequence to be predicted, into the regression equation f(x). The value calculated by this is the dissociation constant K of the compound 6OTD-Py for the predicted DNA sequence. d This is the predicted value of the dissociation constant K d The lower the value of , the more stable the G4 formed in the DNA sequence. Therefore, the dissociation constant K d According to the predicted value of, the lower the predicted value, The probability of G4 formation in the DNA sequence is high, and the stability of the G4 formed in the DNA sequence is high. The higher the predicted value, the lower the probability of G4 formation in the DNA sequence. Even if G4 is formed in the DNA sequence, the stability of the G4 formed in the DNA sequence is low. Predicted.

[0045]

number

[0046] x: A feature vector that indicates the characteristics of the DNA sequence to be predicted N: Number of training data (172) α i , α i * :Lagrange multiplier for each training data x i : 5124-dimensional feature vector for each training data γ: RBF kernel hyperparameter b: constant term C: Cost parameters

[0047] In addition, the optimal value of the error deadband parameter ε and the RBF kernel hyperparameter γ The optimal value of the cost parameter C is examined using the grid search method. In this case, the following error dead zone parameter ε and RBF kernel hyperparameter The candidates for the data γ and the cost parameter C are as follows:

[0048] Candidates for ε: 6 values ​​from 0.001 to 1000 in 10-fold increments Candidates for γ: 6 values ​​from 0.001 to 1000 in 10-fold increments Candidates for C: 6 values ​​from 0.001 to 1000 in 10-fold increments

[0049] Then, for each combination of ε, γ, and C candidates (6 × 6 × 6 = 216 combinations), Under these conditions, 5-fold cross-validation was applied to a total of 172 types of training data. By doing so, the coefficient of determination r 2 By calculating the optimal value of ε, the optimal value of γ, and Consider the optimal value of C. Specifically, we divide a total of 172 types of training data into 5 parts and then divide them into 4 parts. By using a subset of the above, machine learning is performed to verify the regression equation f(x) in the same way as the regression equation f(x) is obtained. The regression equation for the validation is obtained, and the remaining validation subset DNA sequence is transformed into the regression equation for validation. Dissociation constant K of compound 6OTD-Py d The process of calculating the predicted value of is performed on one subset for validation. The experiment was repeated five times, changing the number of the predicted DNA sequences. -Py dissociation constant K d (predicted value) and the actual value (compound of DNA sequence in training data) Dissociation constant K of 6OTD-Py d Using the value of , the coefficient of determination r 2 , correlation coefficient, or RMSE Calculate the Root Mean Squared Error (RME). 2 In the calculation of the predicted value, The coefficient of determination r from the measured values 2 can be calculated directly, but the predicted value itself and the actual value itself Instead of 2 Calculate Alternatively, the RMSE may be calculated directly from the predicted values ​​and the actual measured values. However, instead of the predicted values ​​and the actual values ​​themselves, the logarithms of the predicted values ​​and the actual values ​​are used. The RMSE may be calculated using the predicted values ​​and the actual values. Alternatively, the correlation coefficient may be calculated to represent the relationship between the logarithm of the predicted value and the logarithm of the actual value. You can also calculate the correlation coefficient, which represents the coefficient of determination, r 2 , the correlation coefficient, or RMSE, is The values ​​of ε, γ, and C that result in the highest values ​​are set as the optimal values ​​of ε, γ, and C, respectively.

[0050] In this way, we constructed a prediction model that constructs a regression equation to predict the stability of G4 in DNA sequences. Build a rule.

[0051] G4-binding ligands The G4 binding ligand according to the first embodiment will be described. The synthetic ligand has the following structure: [ka] The present invention relates to a G4 binding ligand consisting of a compound (compound 6OTD-Py) or a salt thereof having the formula: The G4-binding ligand according to the first embodiment is a DNA that has formed G4 in a sample to be measured. When the A sequence is present, it selectively binds to the G4, and the change in its fluorescent properties is used as an indicator. The presence and amount of such DNA sequences can be detected.

[0052] B. Second embodiment Furthermore, the G4 formation sequence prediction device and machine learning method according to the second embodiment, and the G4 formation The synthetic ligand will now be described by way of example.

[0053] <G4 formation sequence prediction device> FIG. 3 is a diagram for explaining the outline of the G4 formation sequence prediction device according to the second embodiment.

[0054] As shown in FIG. 3, the G4 formation sequence prediction device 20 according to the second embodiment includes an input unit 12 and A memory unit 14, a processing unit 16 including a determination unit 16a and a G4 formation prediction unit 16b, and an output unit 18 The G4 formation sequence prediction device 20 includes, for example, the computer shown in FIG. This can be achieved using a PC100 or similar device.

[0055] The input part of the G4 formation sequence prediction system is 5'-N y0 G x1 N y1 G x2 N y2 G x3 N y 3G x4 N y4 -3' DNA sequence (base sequence) (in the DNA sequence, N y0 ~N y 4 represents any sequence fragment with 0 or more bases, and G x1 ~G x4 Only G with 2 or more bases The sequence data of the sequence fragment containing the sequence is input. The memory unit stores prediction models and other The determination unit of the processing unit determines a feature vector x that indicates the characteristics of the DNA sequence. The G4 formation prediction unit of the processing unit uses a prediction model to determine the G4 formation from the sequence data of the DNA sequence. By this, the presence or absence of G4 formation in the DNA sequence (the DNA sequence is G4-shaped) can be determined from the feature vector x. The output unit outputs the prediction results from the G4 formation prediction unit. .

[0056] The prediction model stored in the memory is a feature vector x i and D.N. Classification label for classifying the presence or absence of G4 formation in the A sequence (whether the DNA sequence is a G4 forming sequence or not) (Classification label for classifying whether i By performing machine learning using training data that is a pair of The prediction model is constructed based on the machine learning method according to the second embodiment described later. The method was constructed using the method.

[0057] Formation of G4 in DNA sequence by using the G4 formation sequence prediction device according to the second embodiment In the prediction method for predicting the presence or absence of G4 formation, first, a 5'-N y0 G x1 N y1 G x2 N y2 G x3 N y3 G x4 N y4 The DNA sequence represented by -3' ( In the DNA sequence, N y0 ~N y4 represents any sequence fragment with 0 or more bases, and G x1 ~G x 4 represents a sequence fragment containing only G with two or more bases.)

[0058] Next, sequence data of the DNA sequence is input via the input unit of the G4 formation sequence prediction device. The input sequence data is stored in the storage unit by the processing unit.

[0059] Next, a feature vector showing the characteristics of the DNA sequence is generated by the determination unit of the processing unit of the G4 formation sequence prediction device. The feature vector x is determined from the sequence data. The determined feature vector x is stored in the memory by the processing unit. The method of determining the feature vector x from the array data is described in the second embodiment. The feature vector x in the machine learning method according to the embodiment i The method is similar to the method for determining

[0060] Next, the G4 formation prediction unit of the processing unit of the G4 formation sequence prediction device predicts the predicted By using the prediction model, we can predict whether or not G4 is formed in the DNA sequence from the feature vector x. Specifically, the discrimination function of the prediction model expressed by the formulas (7-1) to (7-3) described later is The value calculated by inputting the feature vector x into the function f(x) is the predicted value of the presence or absence of G4 formation. Obtain as.

[0061] Next, the output unit of the G4 formation sequence prediction device outputs the following as a prediction result by the G4 formation prediction unit: The predicted value of G4 formation is output. According to the predicted value of G4 formation, the boundary is 0.5. As a boundary, the closer the predicted value is to 1, the more likely the DNA sequence is to be a G4-forming sequence. The closer the predicted value is to 0, the more likely the DNA sequence is to be a non-G4 sequence. .

[0062] <Machine learning method> A machine learning method according to the second embodiment will be described. The method includes a teacher data acquisition process and a machine learning process.

[0063] [Teacher data acquisition process] In the teacher data acquisition process according to the second embodiment, first, 5'-N y0 G x1 N y1 G x 2N y2 G x3 N y3 G x4 N y4 -3' DNA sequence (base sequence) In row, N y0 ~N y4 represents any sequence fragment with 0 or more bases, and G x1 ~G x4 is a base A feature vector that represents a sequence fragment containing only Gs with a number of 2 or more, and a DNA sequence Classification label for classifying the presence or absence of G4 formation in the DNA sequence (whether the DNA sequence is a G4 forming sequence or not) The pair of the classification label to be used for classification is obtained as training data.

[0064] [Selection of G4-forming and non-G4-forming sequences for training data] In the training data acquisition process, specifically, first, a DNA microarray obtained in advance is Regarding the various types of DNA sequences arranged, the G4 bonds according to the second embodiment described below are A method is used to detect the presence of G4-forming base sequences in a sample using a cyclic ligand. By this, 2000 types of sequences to which G4-binding ligands bind were identified as G4-forming sequences (G 4) and 2000 types of sequences to which G4-binding ligands do not bind. These are detected as non-G4 forming sequences (sequences that do not form G4). For the sequence, we will analyze the DMS footprint, EMSA, CD spectrum, etc. Using the method described above, we experimentally verify whether G4 formation occurs. Of the 2,000 sequences that were found, 130 were experimentally verified to form G4. 2000 types of G4-forming sequences were selected as training data and detected as non-G4-forming sequences. 130 sequences that were experimentally verified not to form G4 were used as training data for G4 non- Among the sequences that are verified to form G4, 5'-N y 0G x1 N y1 G x2 N y2 G x3 N y3 G x4 N y4 - The sequence not represented in 3' is not included. Among the sequences that were verified not to form G4, 5'-N y0 G x1 N y1 G x 2N y2 G x3 N y3 G x4 N y4 -3' may contain sequences not represented in the 13 The 0 type is the G4 non-forming sequence for the training data, which is a sequence verified not to form a G4. From the middle, 5'-N y0 G x1 N y1 G x2 N y2 G x3 N y3 G x4 N y4 -3' Exclude sequences that are not included in the 5'-N y0 G x1 N y1 G x2 N y2 G x3 N y3 G x4 N y4 Only the sequences represented in -3' are selected.

[0065] [Acquisition of teacher data] Next, we created 130 types of training data for 130 types of G4-forming sequences. For each of the 130 types of G4 non-forming sequences, 130 types of training data were obtained. By acquiring each of these data, a total of 260 types of training data are obtained. As training data for the G4-forming sequence, a feature vector showing the characteristics of the G4-forming sequence is x i and a classification label t that classifies the DNA sequence as a G4-forming sequence. i Get pair with "1" As training data for various types of non-G4 sequences, the characteristics of the non-G4 sequences are Denote the feature vector x i and a classification label t that classifies the DNA sequence as a non-G4 sequence. i Get a pair with "-1".

[0066] Here, various DNA sequences of 130 types of G4 forming sequences and 130 types of non-G4 forming sequences were A feature vector x that describes the characteristics of the columns iA method for determining the above will be described.

[0067] (Obtaining components of feature vectors from DNA sequences) A feature vector x that indicates the characteristics of various DNA sequences i The method for determining the Based on the rule, an initial feature vector x is generated from the sequence data of the DNA sequence. p Ingredient x pn Get Specifically, the DNA sequence is subjected to a G4 formation test based on all arbitrary five base pairs. N for any of the base sequences y0 ~N y4 Along with the homology of the sequence fragments, x1 ~G x Since the number of G bases in 4 is expected to be an important factor, the DNA sequence From the column data, N for any 5-base sequence y0 ~N y4 Homology of sequence fragments of Gender score and G x1 ~G x4 The length of the sequence fragment and the initial feature vector x p Ingredient x p n Obtain as.

[0068] N for any 5-base sequence y0 ~N y4 The homology score of the sequence fragment and The method of obtaining the sequence and the sequence of all five bases are described below. How to arrange five letters selected from the four letters A, T, C, and G. Because it is a type 4 5 = 1024 types. Therefore, for any arbitrary 5-base sequence, N y0 ~N y4 When obtaining the homology score for each sequence fragment, 1024 types of This will give a score for the homology of each sequence fragment to any 5-base sequence. , 1024 scores are obtained. Therefore, N y0 ~N y4 The homology score for each sequence fragment is 1024 × 5 = 5120 scores. These 5120 scores are then used to calculate the initial feature vector x p 512 0 components x pn It becomes.

[0069] In addition, the N y0 ~N y4 Homology score for each sequence fragment If you want to obtain any 5 salts by using the Smith-Waterman method, After aligning each sequence fragment with respect to the base sequence, the sequence is If the characters at the same position in each sequence fragment match (match(m)), a score of +2 is given. If the characters in the same position of the string fragment do not match (mismatch(i)), the score is -1. When a character is deleted in the base sequence or each sequence fragment (gapopen(o)), a gap of -1 is added. Core, any 5-base sequence or sequence fragment with consecutive character deletions (gapex on(x)) is given a score of -0.1, and the total score is calculated for all adjacent putts. The system finds the arrangement pattern that maximizes the total score among all the patterns. The maximum value of the total is the N y0 ~N y4 The homology of each sequence fragment Get it as a score.

[0070] Here is an example of how to obtain a homology score. For example, a base sequence of 5 bases, "ATGCA " for N y0 ~N y4 An example of each sequence fragment, "ATCCGCA", is shown in Table 3 below. When juxtaposed, the sum of the relationship scores for characters in the same position is 2.9.

[0071] [Table 3]

[0072] Then, for an arbitrary 5-base sequence, such as "ATGCA", y0 ~N y4 Each distribution In all alignment patterns of the sequence fragment "ATCCGCA", The alignment pattern that maximizes the total score of the character relationships is shown in Table 4 below. The total score of the relationships of the letters in the same position in the juxtaposition pattern is 8.9. Therefore, for an arbitrary 5-base sequence such as "ATGCA", y0 ~N y4 Each array of An example fragment "ATCCGCA" receives a homology score of 8.9.

[0073] [Table 4]

[0074] On the other hand, G x1 ~G x4 The length of each sequence fragment is the number of G bases contained in the fragment. Yes. x1 ~G x4 To obtain the length of a sequence fragment, use G x1 ~G x4 Four types: The base lengths of the four sequence fragments are the initial characteristic lengths. Characteristic vector x p 4 components of x pn It becomes.

[0075] Initial feature vector x p Ingredient x pn In the acquisition of the above, all 5 N for the base sequence y0 ~N y4 Homology scores for sequence fragments (5120 components) x pn ) and G x1 ~G x4 The length of the sequence fragment (four components x pn ) to be integrated Therefore, the initial feature vector x p 5124 components x p1 ~x p5124 Get the.

[0076] (Feature Selection) The above 5124 components x p1 ~x p5124 The initial feature vector of 5124 dimensions is x p The number of dimensions is larger than the total number of training data, 260. Therefore, the initial Period feature vector x p Machine learning is performed by using training data that is directly adopted from the When constructing a G4 formation sequence prediction device using this method, the generalization performance may be low. Considering this, we use the 5124-dimensional initial feature vector x p By performing feature selection, 5124 x components p1 ~x p5124 Predict whether the DNA sequence is a non-G4 forming sequence from the By selecting components that contribute highly to the prediction accuracy of the measured data, The feature vectors that represent the characteristics of various DNA sequences are denoted by x i It is determined as follows.

[0077] 5124-dimensional initial feature vector x p When selecting features, we first use a linear SVM ( A total of 260 types of training data are generated by setting appropriate constraints using a vector machine (VVM). The feature vector x consisting of only selected components in the data i Instead, we use a 5124-dimensional initial feature vector. Ktor x p By using 260 types of provisional training data that were directly adopted from the A feature selection prediction model that constructs a discriminant function for separating G4-forming sequences and G4-non-forming sequences is The classification function of the feature selection prediction model is expressed by the following formula (4).

[0078]

number

[0079] D(x p ): Discriminant function of predictive model for feature selection w: 5124-dimensional weight vector x p : 5124-dimensional initial feature vector w1~w 5124 :5124-dimensional weight vector components x p1 ~x p5124 :5124-dimensional initial feature vector components b: constant term

[0080] In the classification function of the feature selection prediction model expressed by the above formula (4), the weights of 5124 dimensions are Each component of the vector w n The magnitude of the absolute value of (n=1 to 5124) is 512 4-dimensional initial feature vector x p The component w of the weight vector w in n Component x corresponding to pn This indicates the degree to which the prediction accuracy is affected.

[0081] So, the initial feature vector x of 5124 dimensions p When selecting features, the initial feature vector Tor x p 5124 components x p1 ~x p5124 From the weight vector w, 5124 Components w1~w 5124 The 84 components with absolute values ​​of 0.0175 or more in The corresponding 84 components are selected as components that contribute highly to the prediction accuracy. Then, we obtain a feature vector x i As a result, we have created an 84-dimensional feature set consisting of only selected components. Characteristic vector x i Determine.

[0082] [Machine learning process] Next, in the machine learning process, the above 84-dimensional feature vector x i A total of 260 By using machine learning with training data of various species, y0 G x1 N y1 G x2 N y2 G x3 N y3 G x4 N y4 -3' DNA sequence (in the DNA sequence, N y0 ~N y4 represents any sequence fragment with 0 or more bases, and G x1 ~G x4 is a G with 2 or more bases The presence or absence of G4 formation in the DNA sequence (whether the DNA sequence is a G4 forming sequence or not) A predictive model is constructed to predict whether the product will be affected (machine learning process). By using SVM with RBF kernel, a total of 260 types of training data were used. By performing machine learning on the DNA sequences, we can separate the DNA sequences into G4 forming sequences and non-G4 forming sequences. We will construct a prediction model that configures a classification function that uses the RBF kernel. The process flow for constructing a prediction model will now be described in detail.

[0083] In the process of constructing a predictive model, first, an 84-dimensional feature vector x i Take A total of 260 types of training data were used. i (i=1~N, N=260) J multiplier α i The objective function L expressed by the following equations (5-1) and (5-2) D (α ) is the lag that is maximized within the range that satisfies the constraints expressed by the following formulas (6-1) and (6-2). Lange multiplier α i This machine learning uses RBF filters to determine the combination of The parameter γ of the control is set to an optimal value in the range of 0.001 to 1000, and The test parameter C is set to an optimal value in the range of 0.001 to 1000, and the objective function L D The Lagrange multiplier α that maximizes (α) within the range where the constraint is satisfied. i A combination of A total of 260 types of training data Tr i The corresponding Lagrange multiplier α i A vector that is not 0 is specifically called a support vector.

[0084]

number

[0085]

number

[0086] α i :Lagrange multiplier for each training data x i : 84-dimensional feature vector for each training data ti : Classification label of each training data ("1" or "-1") γ: RBF kernel hyperparameter C: Cost parameters

[0087] Next, the Lagrange multiplier α determined by the above machine learning is i Using a combination of By using this method, the DNA sequence to be predicted, which is different from the sequence used for training data, can be used as a G4-forming sequence and The discrimination function for separating the G4 non-forming sequence is f represented by the following formulas (7-1) to (7-3). (x) is obtained by dividing the feature vector f(x) by the discrimination function f(x) The value calculated by inputting x is normalized to the range of 0 to 1 using a sigmoid function. This is a predictive value for predicting whether or not G4 is formed in the target DNA sequence. According to the predicted value of the presence or absence of G4, the boundary is 0.5, and the closer the predicted value is to 1, the more likely the DNA sequence is G4. The closer the predicted value is to 0, the more likely the DNA sequence is to be a G4 non-forming sequence. It is predicted that it is likely to be a sequence.

[0088]

number

[0089] x: A feature vector that indicates the characteristics of the DNA sequence to be predicted α i :Lagrange multiplier for each training data x i : 84-dimensional feature vector for each training data t i : Classification label of each training data ("1" or "-1") γ: RBF kernel hyperparameter C: Cost parameters ρ: Threshold constant

[0090] In addition, the optimal value of the RBF kernel hyperparameter γ and the optimal value of the cost parameter C are The value is examined using the grid search method. In this case, γ and C are set separately to 0. A total of 260 types of conditions were set for each value in 10-fold increments from 001 to 1000. By applying 5-fold cross-validation to the training data, the accuracy rate (5-fold Cro By calculating the ss Validation Accuracy, the optimal values ​​of γ and C are considered. Specifically, we divided a total of 260 types of training data into five parts, and then used four subsets to evaluate the function. By performing machine learning, a validation discriminant function is obtained in the same way as the discriminant function f(x) is obtained. Then, the DNA sequence of the remaining one subset for verification is classified into G4 forming sequence and The process of separating the G4 non-forming sequences is repeated five times, changing one of the subsets for verification. As a result of the investigation, the values ​​of γ and C that give the highest accuracy rate are set as the optimal values ​​of γ and C, respectively. Determine.

[0091] In this way, a discrimination function for separating DNA sequences into G4-forming sequences and non-G4-forming sequences is obtained. We build a predictive model that

[0092] G4-binding ligands The G4 binding ligand according to the second embodiment will be described. The synthetic ligand is a G4-binding ligand similar to the G4-binding ligand according to the first embodiment. When a DNA sequence that forms G4 is present in the sample to be measured, the G4 is selected. The presence and amount of such DNA sequences can be determined by selectively binding to them and observing the change in their fluorescent properties. It can be detected.

[0093] C. Effects Viral RNA or DNA is known to form a G4 structure, and The G4 structure is involved in transcription, translation, replication, genome stability, etc. This is a target for antiviral drug discovery. According to the first and second embodiments, G4 in the base sequence of nucleic acid such as DNA or RNA The formation state of G4s, such as the presence or absence of G4 formation and the stability of G4s, is determined in the genome independently of the organism species. This information allows for highly accurate predictions. Therefore, it is possible to take prompt action against emerging and re-emerging infectious diseases. It is possible to present drug discovery targets to

[0094] In particular, according to the embodiment, as in the first and second embodiments, Based on statistical methods, not rules, the presence or absence of G4 formation in the base sequence and the stability of G4 are determined. It is possible to predict the formation state of G4. This makes it possible to predict the state of G4 formation, such as the presence or absence of G4 formation and its stability, and therefore allows for the practical application of It is possible to predict the formation process of G4 in vivo. The content of G (guanine) that had been considered to be high and the base sequence other than the base sequence of uniform length were Since it is possible to predict the state of G4 formation, such as the presence or absence of G4 formation and its stability, It is easy to obtain target specificity for G4-forming sequences that can be uniquely predicted by the method according to the embodiment. This also provides the advantage that

[0095] Therefore, in an embodiment, prediction of higher-order structures of nucleic acid structures, particularly those with atypical base pairs, is performed. This makes it possible to present highly unique drug discovery targets.

[0096] Furthermore, according to the compound or a salt thereof and the G4 binding ligand of the embodiment, the first and As in the second embodiment, when a base sequence forming G4 is present in a sample to be measured, The nucleotide sequence of the present invention is selected by selectively binding to the G4 and detecting the change in its fluorescent properties as an indicator. The presence and amount of the nucleotide sequence in the sample to be measured can be detected. The formation / absence of G4 formation can be detected rapidly and sensitively, making it a ligand-based disease marker. The compound according to the embodiment or a salt thereof and G4 The binding ligand eliminates the need for a separation step between the base sequence and the unbound ligand. Therefore, it can be suitably used for imaging in living cells.

[0097] D. Detailed Description of the Embodiments Next, the G4 formation sequence prediction device and machine learning method according to the embodiment, and the compound or The salt and the composition of the G4 binding ligand will now be described in detail.

[0098] 1. G4 formation sequence prediction device The G4 formation sequence prediction device according to the embodiment is characterized by comprising a determination unit and a prediction unit. do.

[0099] The determining part of the G4 formation sequence prediction system is the 5'-N y0 G x1 N y1 G x2 N y2 G x3 N y 3G x4 N y4 -3' (in the base sequence, N y0 ~N y4 The number of bases is 0 or less represents any sequence fragment above, and G x1 ~G x4 is a sequence fragment containing only G with 2 or more bases. A feature vector showing the features of the base sequence is determined from the sequence data of the base sequence.

[0100] The base sequence is 5'-N y0 G x1 N y1 G x2 N y2 G x3 N y3 G x4 N y4 There is no particular limitation as long as the base sequence is represented by -3', and it may be a DNA sequence or an RNA sequence. "Sequence data of a base sequence" refers to the order of letters representing bases in a base sequence. For example, the base sequence of a sequence is a DNA sequence. In some cases, the order of the four letters A, T, C, and G that represent bases in a DNA sequence For example, if the base sequence is an RNA sequence, the This data represents the order of the four letters A, U, C, and G that represent the bases in a DNA sequence.

[0101] The method for determining the feature vector showing the characteristics of the base sequence is not particularly limited, but for example, For example, as in the first embodiment, the salts of all the five arbitrary bases are used as components of the initial feature vector. N for base sequence y0 ~N y4 The homology score of the sequence fragment and G x1 ~G x4 Sequence of The length of each base pair is obtained, and an initial feature vector is obtained by integrating the two. By performing feature selection on the vector, we select components that contribute highly to the prediction accuracy, and the selection results A feature vector consisting of the above components may be determined as a feature vector showing the characteristics of the base sequence. or N for any 5-base sequence y0 ~N y4 The sequence fragment homology Core and G x1 ~G x4 The base lengths of the sequence fragments are obtained, and the feature vectors are integrated. may be obtained and the feature vector may be determined as a feature vector indicating the characteristics of the base sequence.

[0102] The prediction unit of the G4 formation sequence prediction device uses a prediction model to calculate the feature vector The formation state of G4 in the above base sequence is predicted based on the above. "G4 non-forming sequence" refers to a sequence that does not form G4 in the base sequence. It points to an array that does not

[0103] The prediction model is based on the feature vector that shows the characteristics of the base sequence and the G4 shape in the base sequence. By performing machine learning using training data that is a pair of G4 formation state information that indicates the formation state, The prediction model is constructed by, for example, the method described in "2. Machine Learning Method" below. It was constructed by using the machine learning methods described in the previous section.

[0104] The prediction unit of the G4 formation sequence prediction device is not particularly limited, but for example, As shown above, the stability of G4 in the above base sequence is a G4 stability prediction unit for predicting a G4 bond of the base sequence, The G4 stability prediction part, which is the value of the dissociation constant of the combined ligand, is preferred. This is because it is possible to predict not only the formation but also the stability of G4. The prediction unit of the formation sequence prediction device may be, for example, as in the second embodiment, The formation state of G4 is whether or not G4 is formed in the base sequence (if the base sequence is G4 type). The G4 formation prediction unit predicts whether the G4 formation state information is a G4 formation sequence or not. A classification label to classify the presence or absence of G4 formation in the above base sequence (the above base sequence is a G4 forming sequence) It may also be a G4 formation prediction unit, which is a classification label that classifies whether or not there is a G4 formation.

[0105] The G4 formation sequence prediction device is not particularly limited as long as it includes a determination unit and a prediction unit. However, as in the first embodiment, an input section for inputting sequence data of a base sequence, a prediction model, The prediction unit may include a memory unit in which the prediction rule is stored, and an output unit that outputs the prediction result by the prediction unit. .

[0106] 2. Machine Learning Methods The machine learning method according to the embodiment includes a teacher data acquisition step and a machine learning step. It is characterized by:

[0107] (1) Training data acquisition process In the training data acquisition process, 5'-N y0 G x1 N y1 G x2 N y2 G x3 N y3 G x4 N y4 -3' (in the base sequence, N y0 ~N y4 is any number of bases 0 or more represents a sequence fragment of x1 ~G x4 represents a sequence fragment containing only G with two or more bases.) A feature vector showing the characteristics of the G-quadruplex in the above base sequence and G 4. The combination with the formation state information is obtained as training data.

[0108] The base sequence is based on the base sequence explained in the above section "1. G4 formation sequence prediction device." Since it is similar to a column, the explanation will be omitted here.

[0109] The G4 formation state information is not particularly limited, but for example, as in the first embodiment, The dissociation constant of the G-quadruplex binding ligand having the above base sequence is preferable. We were able to construct a predictive model that could predict not only the formation of G4 but also the stability of G4. In addition, as the G4 formation state information, as in the second embodiment, A classification label for classifying the presence or absence of the above G-quadruplex formation in the A classification label that classifies whether a row is a column or not may also be used.

[0110] A feature vector showing the characteristics of the base sequence and a classification to classify the presence or absence of G4 formation in the base sequence. The pair of the label (classification label that classifies whether the base sequence is a G4 forming sequence or not) is used as training data. An embodiment of a method for obtaining the data will be described. In this embodiment, first, G4 is experimentally formed. Several types of G4-forming sequences that were verified to be G4-forming sequences were selected as training data and used in experiments. We used multiple types of G4 non-forming sequences that were verified not to form G4s as training data. Next, multiple G4-forming sequences for the training data are selected as training sequences. The teacher data were obtained from each group, and multiple types of G4 non-forming sequences were used for the teacher data. In this case, the teacher data for each G4 formation sequence are is a feature vector showing the characteristics of the G4 forming sequence and a base sequence is a G4 forming sequence (G4 forming A pair of the classification label "1" that classifies the sequence as a G4 non-forming sequence is obtained. The training data for this are a feature vector that shows the characteristics of the G4 non-forming sequence and a base sequence. The classification label "-1" classifies the sequence as a non-G4 sequence (a sequence that does not form a G4). A feature vector showing the characteristics of the base sequence including the G4 forming sequence and the G4 non-forming sequence is obtained. The method for determining the sequence is not particularly limited, but may be the same as that described in "1. G4 formation sequence prediction device" above. Since this method is similar to the method for determining the feature vector described in , the description thereof will be omitted here.

[0111] (2) Machine learning process In the machine learning step, machine learning is performed using the training data to generate the base sequence. A predictive model is constructed to predict the formation state of the above-mentioned G-quadruplex in the

[0112] The prediction model is not particularly limited, but for example, as in the first embodiment, As the formation state of the G-quadruplex in the sequence, A predictive model for predicting strand stability is preferred. Only the presence or absence of G4 formation in the base sequence This is because we can construct a predictive model that can predict the stability of G4. As a prediction model, as in the second embodiment, As the formation state, the presence or absence of the formation of the G-quadruplex in the base sequence (the base sequence The present invention may also be a prediction model for predicting whether a given sequence is a G4-forming sequence.

[0113] There are few methods for constructing a predictive model by using machine learning with training data. Although it is not fixed, a method that can handle nonlinear relationships in training data is preferable. Possible methods include, for example, SVM (support vector machine) and neural network. These include networking, decision tree regression, random forests, and gradient boosting. For example, as in the SVM used in the first embodiment, a kernel function such as an RBF kernel may be used. A method such as SVM with applied algorithms is preferred.

[0114] (3) Machine learning methods The machine learning method may be the following method. First, a G4 formation sequence prediction device The G4-forming sequence candidates and the G4 non-forming sequence candidates predicted and extracted using the By conducting verification experiments of the formation of G4, new G4 forming and non-G4 forming sequences will be found. New training data was obtained from the newly discovered G4-forming and non-G4-forming sequences, Machine learning is performed using new training data together with existing training data. This allows G4 The prediction accuracy of the formation sequence prediction device can be improved.

[0115] 3. Compound or its salt The compound according to the embodiment, or a salt thereof, has the following general formula (I): [ka] [In the formula, R 2 is H, or [ka] In this case, R 2 R in 1 consists of 2 to 4 conjugated planar rings which may be substituted is an aromatic fused ring R 3 and R 4 are each independently [ka] [ka] It is.] or a salt thereof.

[0116] The above compound or a salt thereof selectively binds to G4 and increases fluorescence depending on the concentration of G4. In order to show whether G4 exists in the base sequence of the DNA sequence in the sample, the presence or absence of G4 is detected by fluorescence. Those skilled in the art will be able to easily understand the present disclosure and the related art. Based on the common general technical knowledge in the field, the compound of the above structure is the compound specifically shown in the examples. It can be seen that the above-mentioned compounds can be suitably used as ligands for detecting G4. The above compound or a salt thereof can further detect G4 by Stokes shift (maximum The difference between the maximum excitation wavelength and the maximum emission wavelength is large, the brightness is high, and the presence or absence of binding to G4 The advantage of this method is that there is a large difference in fluorescent brightness between the two.

[0117] R in general formula (I) 2 R in 1 is, but is not limited to, unsubstituted or may be substituted with one substituent; a naphthyl group, an anthracenyl group, a pyrenyl group, Quinolyl group, isoquinolyl group, 1-methylquinolinium group, acridinyl group, indolyl group group, isoindolyl group, benzothiophenyl group, benzofuranyl group, isobenzofuranyl group group, benzimidazolyl group, benzoxazolyl group, benzisoxazolyl group, and benzyloxy group. The aromatic condensed ring group may be selected from the group consisting of benzothiazolyl groups. The aromatic fused rings are each composed of 2 to 4 conjugated planar rings, and each ring forming the fused ring is a carbon atom. The ring atom is selected from carbon (C), oxygen (O), nitrogen (N) and sulfur (S). These aromatic fused rings have a conjugated ring structure, Its planar structure was found to be favorable for its interaction with G4.

[0118] R 1 When has a substituent, the substituent is preferably a hydroxyl group (-OH), a methoxy group, group (-OMe), acetoxy group (-OAc), mesyloxy group (-OMs), trifluoro dimethylsulfonyloxy group (-OTf), amino group (-NH2), dimethylamino group ( The substituent is selected from the group consisting of a diethylamino group (-NMe2), and a diethylamino group (-NEt2). may be two or more, but it is particularly preferable for it to be substituted with one substituent.

[0119] More specifically, R in general formula (I) 2 R in 1 Examples include, but are not limited to: For example, it may have the structure shown in Table 5 below.

[0120] [Table 5]

[0121] The salt of the above compound is not particularly limited, but may be, for example, hydrochloride, trifluoro In addition, the above compounds can be used as a sample for detecting G4, for example. When the sample is brought into contact with the substance derived from the living body or the buffer, etc., contained in the sample, Salts or solvates may also be formed.

[0122] Compounds according to the invention have the following structure: [ka] In the present specification, this compound is referred to as "compound 6OTD-Py." Write down.

[0123] The compounds according to the embodiments, or salts thereof, can be prepared as described herein, including in the Examples below, and in the art. The compound can be suitably synthesized according to the common technical knowledge in the art.

[0124] 4. G4-binding ligands The G4 binding ligand according to the embodiment is a G4 binding ligand comprising a compound according to the embodiment or a salt thereof. 4-binding ligand.

[0125] The G4-binding ligand according to the embodiment is a salt of a DNA sequence or the like that forms G4 in a sample. When the base sequence is present, it selectively binds to the G4, and the change in its fluorescent properties is used as an indicator. The presence and amount of such a base sequence can be detected. The gand is a compound or compounds that do not emit fluorescence alone (i.e., when not bound to G4). It is a "light-up ligand" consisting of a salt of

[0126] Thus, embodiments also provide for the detection of G4 binding ligands in a sample using embodiments of the G4 binding ligands. A method for detecting the presence of a base sequence that forms G4 is provided. More specifically, the method of the present invention comprises the following steps: (i) Fluorescence generated when the above-mentioned G4-binding ligand of the present invention is brought into contact with a sample Measure (n) Compare with the fluorescence in the absence of sample A method for detecting a base sequence that forms G4 in a sample, comprising: The method may be as described above, but does not include a step of separating unbound ligand.

[0127] Since the G4-binding ligand according to the embodiment is a "light-up type ligand," Little or no fluorescence is detected in the absence of a sample in step (n). Therefore, it can be ignored in the above method, i.e., it does not include step (n), It can simply be a method for carrying out step (i). In this detection method, if fluorescence is generated in the presence of a sample, it is determined that a base sequence that has formed G4 is present in the sample. It may indicate that a column exists.

[0128] The detection method according to the embodiment is to visually confirm an increase or change in fluorescence. In the detection method according to the embodiment, a "light-up type" G4-binding ligand is used. Because the enzyme uses a “ligand,” the presence of a base sequence that forms G4 in the sample is detected by the generation of fluorescence. Indicates presence.

[0129] Alternatively, the detection method according to the embodiment may measure the fluorescence intensity at a specific wavelength. In addition, the detection method according to the embodiment can obtain a fluorescence spectrum in a specific wavelength range. The specific wavelength or range of wavelengths to be measured may vary depending on the embodiment. This can be appropriately determined depending on the compound or a salt thereof used as the ligand.

[0130] The G4 binding ligand according to the embodiment has a fluorescence wavelength of 510 nm when excited at 406 nm. Since the length of the nucleotide sequence is longer than 100 nm, the Stokes shift exceeds 100 nm, indicating the presence of a G4-forming base sequence. The presence of the object can be more easily detected.

[0131] For example, the compound 6OTD-Py was used to detect the presence of a base sequence that formed G4 in a sample. When detection is intended, an excitation wavelength of 406 nm is used, but is not limited thereto. In this regard, it is preferable to measure the fluorescence intensity at a wavelength in the range of 500 to 600 nm.

[0132] The concentration of the ligand used for detection is determined based on the salts of DNA sequences and the like that may be contained in the sample to be measured. The optimum ligand concentration for fluorescence measurement can be adjusted depending on the concentration of the base sequence. The degree of amplification is not particularly limited, but is preferably about 1 / 70 to about 50 times the base sequence. The concentration of the ligand in the fluorescence measurement sample is not particularly limited, but may be: A suitable range for measurement is 0.1 μM to 0.5 μM, preferably 0.1 μM to 0.2 μM. It can be determined.

[0133] The measurement of the fluorescence intensity in the detection method according to the embodiment is not particularly limited, but for example, For example, under pH conditions in the range of pH 7.0 to 8.0, preferably pH 7.3 to 7.7, The reaction can be carried out at room temperature. The salt concentration in the sample is set at a level that allows G4 to form. There are no particular limitations, but the concentration is preferably 10 to 2000 mM, more preferably 20 to 15 It is preferable to adjust the concentration within the range of 0 mM.

[0134] The G4-binding ligand according to the embodiment is different from base sequences such as DNA sequences that do not form G4. No significant binding was observed, and therefore the presence of the G4-forming base sequence could be specifically detected. In addition, the fluorescence obtained by the G4-binding ligand according to the embodiment can be obtained by forming G4. Since the fluorescence intensity increases depending on the concentration of the nucleotide sequence, the fluorescence intensity in the absence of the sample (nucleotide sequence) By comparing with light, it can be used to quantify the base sequences that form G4 in a sample. In this case, the fluorescence intensity depending on the concentration of the base sequence that formed G4 was calculated for each compound used as a ligand. The intensity can be measured in advance and a calibration curve can be prepared.

[0135] As described above, in the detection method according to the embodiment, the G4-binding ligand is “light-up Since the "G4-type ligand" is used, the fluorescence occurs when the salt that formed G4 in the sample is This indicates the presence of a base sequence, so the concentration of the base sequence that formed G4 can be quantified more accurately. It can provide rapid and accurate measurement results. The fluorescence emitted by the nucleotide sequence is generated only by binding to the G4-forming base sequence, and the intensity of the fluorescence is Since the fluorescence increases depending on the concentration, it should be compared with the fluorescence in the absence of sample (base sequence). (Step (n)) can be used to quantify the base sequence that forms G4 in a sample. Cut. EXAMPLES

[0136] Hereinafter, the embodiment of the present invention will be further described with reference to examples.

[0137] [Example 1: Construction of a prediction device] A G4 formation sequence prediction device according to the second embodiment was constructed. The prediction model stored in the storage unit of the device is subjected to the machine learning method according to the second embodiment. It was constructed by:

[0138] When the predictive model is constructed by using the machine learning method, the predictive model is used for machine learning. A portion of the G4-forming sequences and non-G4-forming sequences from which training data was obtained (SEQ ID NO: 1 to SEQ ID NO: The number 20) is as follows. The training data used for machine learning was obtained from the G4 non-tangible The synthetic sequence is a sequence that has been verified not to form a G4 and has a 5'-N y0 G x1 N y1 G x2 N y2 G x3 N y3 G x4 N y4 -3'(N y0 ~N y4is any sequence with 0 or more bases G stands for fragment x1 ~G x4 represents a sequence fragment containing only G with two or more bases. Only sequences that can be identified were selected.

[0139] (G4 forming sequence) SEQ ID NO: 1: GTGAGGCTAGGGGGTGGTGGTGGTGGGGGTGAGGAAGGGAAAGAT SEQ ID NO: 2: ATAGCGCCAGTGGGTGGGCTTAGATCTGGGAAGGGCGGGACAGAG SEQ ID NO: 3: GTCCAGGAAGGCTAGAGGTGGGGGGGCGGGTACCGGTGAAGGGAG SEQ ID NO: 4: TAAAGAGGAGGGTGGCGGTAGTGGAGGGGGGGGTTGGAGTTGGTT SEQ ID NO: 5: CTTGTGTTCGGGAAGGGGGGGGGGGGAGCGTTGGAAACGCATGC SEQ ID NO: 6: ACTTGGGTAGGCGGGCTTGGGAGGCTCCGTTGGACGTGGGGTCTA SEQ ID NO: 7: CACTCTTTTTGGGGTTGGGACTGAGGTGGCGGTCACGCGAGGATC SEQ ID NO: 8: AGTAGGGAGCTTTGGAATGAGGGATTAACACTTTGGGGGACTTAGTC SEQ ID NO: 9: AGGGGTAGGTTGGGTGGTAAGAGGTGGTAAGCGGAGCGGCTGCTG SEQ ID NO: 10: TGGGGAGGCTGGGTGGAGGACAAAGTAGAAATAAAAATACGGAAGTAGAT

[0140] (G4 non-forming sequence) SEQ ID NO: 11: CAAATACATACCGACTCTCCGGGGACCAAGGGGAA SEQ ID NO: 12: TTCCGCGCCTAGCCAAGGGGAGGAACGGGGCAGGA SEQ ID NO: 13: CACCCACAAGCTCTGGGATGGGGCTGGGAGTGGTC SEQ ID NO: 14: CTACAGCAAATGATGGAAATTAACAAGGCAGGAGG SEQ ID NO: 15: TTTCCACCTCAGGAGAAAGGGCGGGAGGGAAGGAG SEQ ID NO: 16: CTAGACTGCTGGAAGGGGGAGACCGGCGTCGACAG SEQ ID NO: 17: ACGCAAACGGGGAGGGCCGAGACGCGGATTCTCGA SEQ ID NO: 18: ACCTGAGAGGCACCCGGGGAGGAAGGAAAATGTGT SEQ ID NO: 19: AGTTCCGGAGGTAGGCGCGCACTTGAAGGAGAAG SEQ ID NO: 20: CCAGAGTCAGGCTGCGGGCCCACGGGGGTGCAGGC

[0141] Furthermore, when the predictive model is constructed by using the machine learning method, Regarding the optimal value of the RBF kernel hyperparameter γ and the optimal value of the cost parameter C, As a result of the grid search method, the optimal value of γ was found to be γ=0.01 and the optimal value of C was found to be The value is C=1000, and the correct answer rate at γ=0.01 and C=1000 is 96.45% (maximum High price).

[0142] [Evaluation of prediction accuracy of the prediction device of Example 1] The G4 formation sequence prediction device constructed in Example 1 (hereinafter, sometimes abbreviated as "prediction device") The prediction accuracy of the prediction device constructed in Example 1 was evaluated. A method for predicting the presence or absence of G4 formation in a DNA sequence (whether a DNA sequence is a G4 forming sequence or not) FIG. 5 is a diagram showing a process flow of the prediction result of the prediction device constructed in the first embodiment. The following is a detailed explanation of the results of the verification experiment. Reveal.

[0143] [Prediction of G4-forming and non-G4-forming sequences] In the evaluation of the prediction accuracy of the prediction device constructed in Example 1, first, 5'-N, which can form G4 y0 G x1 N y1 G x2 N y2 G x3 N y3 G x4 N y4 -3' DNA sequence (in the sequence, N y0 ~N y4 is any number of bases 0 or more represents any sequence fragment, and G x1 ~G x4 represents a sequence fragment containing only Gs having two or more bases. ) and searched for DNA sequences of 15 to 40 bases. In the CoV2 genome, the predicted DNA sequence is divided into two parts, excluding the Tether sequences at both ends, G x1 N y1 G x2 N y2 G x3 N y3 G x4 There are 132 places where the part " In contrast, there were 5,726 types of DNA sequences to be predicted.

[0144] Next, by using the prediction device of Example 1, 5726 types of DNA sequences to be predicted were For each sequence, the presence or absence of G4 formation (G4 forming sequence or not) was predicted.

[0145] At this time, the presence or absence of G4 formation in each DNA sequence (whether each DNA sequence is a G4 forming sequence or not) When predicting whether a DNA sequence is a nucleotide sequence or not, first, sequence data of the DNA sequence is input via the input unit of the prediction device. Next, a feature vector showing the characteristics of the DNA sequence is generated by the determination unit of the processing unit of the prediction device. The loop x was determined from the sequence data. Next, the G4 formation prediction unit of the processing unit of the prediction device By using the prediction model stored in the part, G4 in the DNA sequence is calculated from the feature vector x. The presence or absence of G4 formation (whether the DNA sequence is a G4 forming sequence or not) was predicted. Input the feature vector x to the classification function f(x) expressed by the formulas (7-1) to (7-3). The value calculated by the above was used as a prediction value for the presence or absence of G4 formation.

[0146] In this way, the presence or absence of G4 formation was determined for the 5,726 types of DNA sequences to be predicted. (G4 forming sequence or not) and predict whether G4 is formed or not for each DNA sequence. As mentioned above, the predicted value of the presence or absence of G4 formation was set at 0.5. The closer the predicted value is to 1, the more likely the DNA sequence is to be a G4-forming sequence. The closer the prediction value is to 0, the more likely the DNA sequence is to be a non-G4 forming sequence.

[0147] Next, the predicted DNA sequence was divided into two parts, the Tether sequences at both ends of which were removed, called "G x1 N y1 G x 2N y2 G x3 N y3 G x4 " is a G4-forming sequence. The sequence of the portion predicted to form a G4 is determined based on the predicted DNA sequence. In order to verify both the predicted DNA sequence and whether it is a sequence, G4-forming and non-G4-forming sequences were predicted from 26 types of DNA sequences, and the G4-forming sequences were Candidates and candidates for non-G4 forming sequences were extracted.

[0148] Specifically, the "G" in the 5726 types of DNA sequences to be predicted x1 N y1 G x2 N y 2G x3 N y3 G x4 From each of the 132 groups that have the same part represented by ", The DNA sequence with the predicted value of the presence or absence of the gene formation closest to 1 was selected. Two types of DNA sequences were predicted as G4-forming sequences, and 132 types of G4-forming sequence candidates were extracted. I put it out.

[0149] Furthermore, the G x1 N y1 G x2 N y2 G x3 N y3 G x4 All DNs belonging to 132 groups that have the same part represented by " Select at least one group with a predicted value of less than 0.5 for the presence or absence of G4 formation in the A sequence. Then, all DNA sequences belonging to at least one of the selected groups were classified as G4 non-G4 It was predicted to be a G4 forming sequence and extracted as a candidate for a G4 non-forming sequence.

[0150] [Verification experiment] From the 132 G4 formation sequence candidates above, 13 G4 formation sequence candidates ( Candidates 1-13) for validation of G4-forming sequences were selected, and the selected candidates for validation were confirmed to be G4-forming sequences. We also experimentally verified whether there are any G4 non-forming sequences among the nine candidates listed above. Candidates for validation of G4 non-forming sequences (validation candidates 1 to 9 of G4 non-forming sequences) were selected. We experimentally verified whether the candidates for validation were non-G4 forming sequences.

[0151] Specifically, first, 13 types of candidates for G4 formation were selected for validation (candidates for validation of G4 formation sequences). Thirteen types of oligonucleotides (complements 1 to 13) were prepared, and nine types of non-G4 sequences were Candidates for validation (validation candidates 1 to 9 for G4 non-forming sequences) were prepared as nine types of oligonucleotides. It was made by:

[0152] Next, 13 candidates for G4 formation and 9 candidates for non-G4 formation were selected. For each of the 22 candidate oligonucleotides, samples were prepared at various concentrations. In this case, the fluorescence intensity of lithium cacodylate (10 mM, pH 7.4, 1 6OTD-Py (0.2 μM) was dissolved in 0.00 mM KCl, 0.01% Tween-20. Then, oligonucleotides were added at various concentrations (0-5 μM) and incubated for 24 hours. Subsequently, the fluorescence intensity of the samples thus obtained at various concentrations was measured. The fluorescence intensity was measured using a 96- or 384-well plate (Perkin Elmer, 6002290; Greiner , 781900) and measured with a microplate reader (TECAN, Spark). Put 40 μL of samples of various concentrations into each well of a 96- or 384-well plate. The fluorescence intensity was measured at an excitation wavelength of 406 nm and an emission wavelength of 510 nm.

[0153] Next, we prepared 13 types of G4 formation sequence candidates for validation (G4 formation sequence candidates 1 to 13). And the total of nine types of candidates for validation of non-G4 sequences (candidates for validation of non-G4 sequences 1 to 9) A total of 22 oligonucleotides were measured in samples of various concentrations. From the fluorescence intensity, the relationship between the fluorescence intensity and the sample concentration is expressed by the following regression equation (8). By fitting the curve, the logarithm of the dissociation constant of 6OTD-Py, log(K d [M]) Then, the logarithm of the dissociation constant, log(K d [M]) < -7, The G4-forming sequence was determined, and the logarithm of the dissociation constant, log(K d [M]) ≧ -7 The complement was determined to be a non-G4 forming sequence.

[0154]

number

[0155] F: Fluorescence intensity m0: [OLG] (oligonucleotide concentration [μM]) m1: [6OTD-Py] (6OTD-Py concentration [μM]) m2:K d (Dissociation constant [μM]) A: Maximum fluorescence intensity of the complex of 6OTD-Py and G4

[0156] As the G4-forming sequence, the following known telo24 (SEQ ID NO: 21) and bcl2 (SEQ ID NO: 22) For k-ras (SEQ ID NO: 23), the logarithm of the dissociation constant log(K d [M]) is required, and both are log(K d [M])=-9~-8(<-7) The G4 non-forming sequences, triple-2, triple-3, and dsDNA, are known to be A. For ssDNA, the logarithm of the dissociation constant, log(K d [M]) is required Both are log(K d[M] was ≥-7.

[0157] telo24 (SEQ ID NO: 21): TTAGGGTTAGGGTTAGGGTTAGGG bcl2 (SEQ ID NO: 22): GGGCGCGGGGAGGAATTGGGCGGG k-ras (SEQ ID NO: 23): AGGGCGGTGTGGGAAGAGGGAAGAGGGGGAGG

[0158] Triplex-2 (SEQ ID NO: 24): CTGGGAGGGAGGGA Triplex-3 (SEQ ID NO: 25): GGTTGGTGTGG dsDNA (SEQ ID NO: 26): TATAGCTATATTTTTTTATAGCTATA ssDNA (SEQ ID NO: 27): TTAGAGTTAGAGTTAGAGTTAGAG

[0159] [Evaluation of prediction accuracy] As a result of the validation experiment, 13 types of G4-forming sequences were selected as candidates for validation, as shown in Figure 5. Eleven of the candidates (1 to 13 for validation of G4-forming sequences) were determined to be G4-forming sequences as expected. The nine candidates for validation of non-G4 sequences (candidates 1 to 9 for validation of non-G4 sequences) were selected. Of these, eight were determined to be non-G4 sequences as expected. In the measurement, the AUC (Aria under ROC curve) was 0.90, and the accuracy rate was 86.36%. I was able to confirm that it was possible.

[0160] [Example 2: Construction of a prediction device] The G4 formation sequence prediction device according to the first embodiment was constructed. The prediction model stored in the storage unit of the device is subjected to the machine learning method according to the first embodiment. It was constructed by:

[0161] The prediction model is constructed by using the machine learning method using training data. When performing machine learning on various G4-forming sequences, the G4-formation sequences in the training data are The dissociation constant K of the compound 6OTD-Py d The value of the oligonucleotide of the G4-forming sequence is The value [M] of the nucleotide that is actually obtained by the general method was set. The compound 6OTD- of the G4 non-forming sequence in the teacher data for the G4 non-forming sequence of the species Dissociation constant K of Py d The value of the dissociation constant K d The value of is accurately calculated. Considering the possibility that the sequence may not be able to be determined, we prepared training data for all types of G4 non-forming sequences. The dissociation constant K of the compound 6OTD-Py having the G4 non-forming sequence in d The value of 1×10 -5 .2 A constant value greater than M (1×10 -4 M) was set.

[0162] Furthermore, when the predictive model is constructed by using the machine learning method, , the optimal value of the error deadband parameter ε, and the optimal value of the RBF kernel hyperparameter γ The optimal values ​​of the cost parameter C and the cost function C were examined using a grid search method. , the optimal value of ε is ε = 0.1, the optimal value of γ is γ = 0.01, and the optimal value of C is C = The values ​​of ε, γ, and C were then set to the optimal values ​​of ε, γ, and C, respectively. The dissociation constant K of the compound 6OTD-Py with the DNA sequence calculated when d The predicted value of [M ] and the actual value (the dissociation constant K d of The correlation coefficient calculated using the value [M] was 0.7826 (the highest value). The correlation coefficient is K d Predicted value x 10 6 Logarithm of [μM] (log(K d Predicted value x 1 0 6 [μM]) and K d The logarithm of the measured value (log(K d [M])) The correlation coefficients representing the optimal values ​​of ε, γ, and C were calculated. When each is set, K d The logarithm of the predicted value of (log(K d The predicted value of [M]) and K d The logarithm of the measured value (log(K d The RMSE calculated from [M]) is 2.5073 ( The highest value.

[0163] [Evaluation of prediction accuracy of the prediction device of Example 2] The G4 formation sequence prediction device constructed in Example 2 (hereinafter, sometimes abbreviated as "prediction device") The prediction accuracy was evaluated.

[0164] [Prediction of G4-forming and non-G4-forming sequences] In the evaluation of the prediction accuracy of the prediction device constructed in the second embodiment, the prediction device of the second embodiment was used. As a result, verification was performed in the [Verification Experiment] in the above [Evaluation of the Prediction Accuracy of the Prediction Device of Example 1]. 22 consisting of validation candidates 1 to 13 for G4 formation sequences and validation candidates 1 to 9 for non-G4 formation sequences The stability of G4 in each of the DNA sequences (base sequences) was predicted.

[0165] At this time, first, the sequence data of the DNA sequence was input via the input unit of the prediction device. The determination unit of the processing unit of the prediction device determines the feature vector x indicating the characteristics of the DNA sequence from the sequence data. Next, the G4 stability prediction unit in the processing unit of the prediction device uses a prediction model to By doing this, we predicted the stability of G4 in the DNA sequence from the feature vector x. The feature vector is added to the regression equation f(x) of the prediction model represented by the above equations (3-1) and (3-2). Enter the value of the DNA sequence of the compound 6OTD-Py to calculate the dissociation constant K d The predicted value [M] of the DNA sequence was then output by the output section of the prediction device. Dissociation constant K of compound 6OTD-Py d The predicted value was output.

[0166] In this way, the G4 sequence of each of the 22 types of DNA sequences to be predicted was We predicted the stability of the compound 6OTD-Py for each DNA sequence by using the dissociation constant K d Prediction The value [M] was obtained. As mentioned above, the dissociation constant K d According to the predicted value of The more likely it is that a G4 will form in the DNA sequence, the more stable the G4 will be. The higher the predicted value, the higher the possibility of G4 formation in the DNA sequence. Even if a G4 is formed in the DNA sequence, the stability of the G4 formed in the DNA sequence is low. is predicted to be low.

[0167] [Evaluation of prediction accuracy] For the 22 types of DNA sequences to be predicted, the predicted G4 stability for each DNA sequence was and the dissociation constant K d Predicted value x 10 6 Logarithm of [μM] log(K d Predicted value x 10 6 [μM ]) was calculated. FIG. 6 shows the stability of G4 in a DNA sequence by using the prediction device of Example 2. Dissociation constant K of training data when predicting qualitatived The values ​​of and the predicted stability values ​​of G4 are shown in Example The results obtained when predicting the presence or absence of G4 formation in a DNA sequence using the prediction device in FIG. 7(a) is a table showing the results of the prediction of the presence or absence of G4 formation. The dissociation constant K of the training data when predicting the stability of G4 in DNA sequence using d The logarithm of the value of log(K d 1 is a graph showing the predicted stability of G4 against [M]. FIG. 7(b) shows the effect of G4 formation in a DNA sequence using the prediction device of Example 1. Dissociation constant K of training data when null is predicted d The logarithm of the value of log(K d [M]) 7(a) and 7(b) are graphs showing predicted values ​​of the presence or absence of G4 formation. In the graph, the horizontal axis shows log(K d Any value of [M] greater than -6 will be treated as -6. It was shown.

[0168] As shown in FIG. 6 and FIG. 7(a), by using the prediction device of Example 2, each DNA When predicting the stability of G4 in a sequence, log(K d Predicted value x 10 6 [μM] The predicted stability of a G4 is the dissociation constant K d The logarithm of the value of log(K d [ The predicted stability of G4 and log(K d [M]) was positively correlated. And, especially log(K d [M]) < -7, G4 Stability prediction value (log(K d Predicted value x 10 6 [μM]) and log(K d [M]) The correlation coefficient between the two was 0.7826. d [M]) <-7, K d The logarithm of the predicted value of (log(K d The predicted value of [M]) and K d of The logarithm of the measured value (log(K d The RMSE calculated from [M]) is 2.5073. Therefore, the prediction device of Example 2 can predict the stability of G4 in a DNA sequence with high accuracy. On the other hand, as shown in FIG. 6 and FIG. 7(b), the results of Example 1 When the prediction device was used to predict whether or not G4 was formed in each DNA sequence, Prediction of the formation of and log(K d There was no correlation between [M].

[0169] [Verification experiment of G4 stability] 83 types of G4 type from which the training data used to construct the prediction model in Example 2 was obtained The stability of G4 was experimentally evaluated for the following six types of G4 formation sequences in the synthetic sequence. The following six types of G4-forming sequences were verified by the dissociation constant K d Low K value d Three types of value groups (NCOR2(log(K d [M])=-9.7), telo24(l og(K d [M])=-7.1), and BIN1(log(K d [M])=-7.5)), and the compound Dissociation constant K of 6OTD-Py d High value of high K d Three types of value groups (OTP(log(K d [ M])=-6.0), WT1(log(K d [M])=-6.1), ZN318(log(K d [M])= -5.9)).

[0170] NCOR2 (SEQ ID NO: 28): ACTGGCATGGTGGGGTGCAAGGTTCTTTGCTGGGGTGCGGGCGAT telo24 (SEQ ID NO: 29): TTAGGGTTAGGGTTAGGGTTAGGG BIN1 (SEQ ID NO: 30): AGGATGGCAGAGATAGGGGAGCAAGGGGGTGACGGCGGGGAAGATC OTP (SEQ ID NO: 31): AAGGAGGGAAGGGGAGGATTAGAGCACGGAGTAGGTATTGTGCTC WT1 (SEQ ID NO: 32): AGTAGGGAGCTTTGGAATGAGGGATTAACACTTTGGGGGACTTAGTC ZN318 (SEQ ID NO: 33): AGGAGAGGGGTGGTCTCTCTGCATCACGTGGTCTCCGGGGGAGTA

[0171] When the stability of the G4s in the above six types of G4-forming sequences was experimentally verified, Specifically, the UV absorption of each of the six types of G4-forming oligonucleotides was The spectrum was measured in a lithium cacodylate buffer (10 mM, pH 7. 4) Add KCl (100 mM) to the solution and add each oligonucleotide (10 μM). The sample solution thus obtained was then analyzed using an ultraviolet-visible spectrometer manufactured by JASCO Corporation. The UV absorption spectrum (2 The UV absorption difference spectrum was obtained under these two temperature conditions. I got it. In addition, high K d The sample solution was added with three types of values ​​(OTP, WT1, and ZN318). For this purpose, the compound L2H2-6OTD (20 μM), a known G4-binding ligand, was used. The sample solution after addition was subjected to the same measurement, and the UV absorption difference spectrum was obtained. Figure 8 shows the results of six G4-forming sequences (NCOR2, telo24, BIN1, OTP, WT1, and ZN3). 18) The solid lines represent the UV absorption difference spectra of the compound L2H2-6OTD before and after its addition. This is a diagram showing the roughness of high K d The graphs for the three types of value groups (OTP, WT1, and ZN318) The UV absorption difference spectra after addition of compound L2H2-6OTD are shown by dashed lines.

[0172] As shown in Figure 8, the low K d Three types of value groups (NCOR2, telo2 4, BIN1) before the addition of L2H2-6OTD, the UV absorption difference spectrum (solid line) shows a m, indicating the formation of G4. d Three types of value groups (OT P, WT1, and ZN318) before and after the addition of L2H2-6OTD (solid line). In contrast, no UV absorption was observed at 295 nm. d Three types of value groups UV absorption difference spectra (dashed lines) of (OTP, WT1, and ZN318) after addition of L2H2-6OTD. In the case of the 295 nm UV absorption was observed, indicating the formation of G4. et al., low K d The three oligonucleotides in the value group spontaneously formed G4, whereas High K d The three oligonucleotides in the value group did not spontaneously form G4s and were G4-binding ligases. It was found that G4 was formed only when nitrile was added. For the sequence, the dissociation constant K of the G4-binding ligand d The lower the value, the more stable the G4 is. was confirmed.

[0173] [Example 3: Synthesis of 6OTD-Py] A new compound, 6OTD-Py, was synthesized according to the following method.

[0174] 1. Synthesis of Compound 16 As described in the column "Example 1: Synthesis of Compound 1" in JP2020-176108A, According to the method described in JP 2020-176108 A, compound 16 (large ring) We synthesized a crystalline substance16).

[0175] 2. Synthesis of compound New-1 [ka]

[0176] The compound was mixed in dry THF (tetrahydrofuran) (500 L) at room temperature under an argon atmosphere. A solution of 16 (5.00 mg, 5.56 mol) was added to 1-ethenylpyrene (12.8 mg , 55.6 mol) and Grubbs second generation catalyst (1.50 mg, 1.70 mol). The reaction mixture was then stirred for 29 hours, after which the mixture was cooled to 90° C. The residue was purified by silica gel column chromatography (chloroform / methanol = The compound New-1 (4.20 mg, 69%) was obtained by purification using a 200:1 mixture.

[0177] TIFF2025069236000021.tif43170

[0178] 3. Synthesis of 6OTD-Py [ka]

[0179] Compound New-1 (4.20 mg, 3.82 mol) in dichloromethane (2 mL) Trifluoroacetic acid (1 mL) was added to the mixture at room temperature and stirred for 10 minutes. The reaction mixture was concentrated in vacuo. The mixture was concentrated with ethyl acetate to give 6OTD-Py (3.00 mg, 87%).

[0180] TIFF2025069236000023.tif43170

[0181] Although the embodiment of the present invention has been described in detail above, the present invention is not limited to the above embodiment. However, various modifications are possible without departing from the spirit of the present invention as described in the claims. It is possible to make design changes to the product. [Industrial Applicability]

[0182] This invention is aimed at developing new drugs such as antiviral drugs, antibacterial drugs, and antifungal drugs for emerging and re-emerging infectious diseases. It can be used for research. It can also be used for drug development research for human genetic diseases (cancer, neurodegenerative diseases, etc.). can be used. [Explanation of symbols]

[0183] 10 G4 formation sequence prediction device 2 Input section 4 Storage section 6 Processing section 6a Decision section 6b G4 stability prediction section 8 Output section 20 G4 formation sequence prediction device 12 Input section 14 Storage section 16 Processing section 16a Decision section 16b G4 formation prediction part 18 Output section 100 Computers 110 Processing device (computer) 120 Storage device 130 Input Device 140 Output Device 150 Input / Output Interface (I / F)

Claims

1. The following general formula (I): 【Chemistry 1】 [In the formula, R ２ is H, or 【Chemistry 2】 In this case, R ２ Inside R １ is an aromatic fused ring system consisting of 2 to 4 conjugated planar rings which may be substituted, R ３ and R ４ are each independently 【Chemistry 3】 【Chemistry 4】 It is.] or a salt thereof.

2. The structure: 【Chemistry 5】 A compound or a salt thereof having the formula:

3. A G-quadruplex binding ligand comprising the compound according to claim 1 or a salt thereof.

Images