JP2005078407A - Data search method, data search device, data search program and recording medium having the program recorded thereon - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To search at high speed a sequence having the highest similarity to a given genome sequence (a base sequence, amino acid) from a database. <P>SOLUTION: An input sequence is defined as sequence 1, and a sequence fetched from the database as sequence 2. A data search method executes: superposition processing (Step 5) to make the sequence 2 in an superposition state by using a quantum algorithm; processing to apply quantum arithmetic operations complying with dynamic programming to the sequence 1 and the superposed sequence 2 by using a quantum query and derive (Step 6) the score of pair-wise alignment between the sequence 1 and the sequence 2; then processing to repeat Grover processes consisting of application of quantum oracle and diffusion conversion to the sequence 2 (Steps 7 to 10); and processing to obtain a candidate y' for a solution by observing a qubit corresponding to the sequence 2 (Step 11). <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a method and apparatus for searching a database for a sequence having the highest similarity to a given genome sequence (base sequence, amino acid sequence) using a quantum computer or applying a quantum algorithm.

  In the field of bioinformatics, in which information science is applied to biology, the decoding of base sequences in DNA and RNA possessed by organisms, amino acid sequences in proteins and peptides, etc., is rapidly progressing. In this specification, a base sequence and an amino acid sequence are collectively called a genome sequence. Decoded sequences are stored in a database. By discovering sequences with high similarity among many sequences, it becomes possible to elucidate new functions and evolutionary systems of living organisms, and to obtain useful knowledge for drug discovery and gene therapy. Can do.

However, it is not realistic to calculate the similarity with high accuracy because a large amount of calculation time and memory space are required and the cost is increased. Therefore, a technique of calculating the similarity between sequences by an algorithm using an approximate calculation method has become mainstream.
SB Needleman and CD Wunsch, Journal of Molecular Biology, Vol. 48, pp. 443-453 (1970) L. Grover, 28th Annual ACM Symposium on the Theory of Computing, pp. 212-219 (1996) M Boyer, G. Brassard, P. Hoyer and A. Tapp, 4th Workshop on Physics and Computation, pp. 36-43 (1996) C. Durr and P. Hoyer, "A quantum algorithm for finding the minimum," [online], January 7, 1999, [August 27, 2003 search], Internet <URL: http://arXiv.org / abs / quant-ph / 9607014>

  Currently, the decoding of genome sequences is rapidly progressing, and the amount of data stored in the database is enormous. When the number of sequences stored in the database is n, the amount of calculation by the similarity calculation algorithm currently used is about O (n), but the amount of data stored in the database, that is, n is accelerated. Since it is increasing, the amount of calculation for calculating the similarity of sequences is expanding. In order to elucidate the features and functions of sequences at higher speed based on the similarity between sequences, the development of algorithms for efficiently searching databases has become an urgent issue.

  The similarity between two sequences is generally calculated through a procedure called pair-wise alignment, and the higher the similarity, the higher the score. In general, DP (dynamic programming) is used in the calculation of pairwise alignment (see Non-Patent Document 1). However, calculation of DP with high reliability requires a large amount of calculation. Therefore, in practice, the calculation speed is often improved by approximate calculation, but there is a problem that the accuracy is inferior.

  Accordingly, an object of the present invention is to provide a data search method and apparatus for searching a database having a highest similarity to a given genome sequence at high speed.

  Computers called “quantum computers” have been proposed that can solve problems much faster in principle than computers that are currently widely used. A quantum computer uses a quantum superposition state, which is a phenomenon peculiar to a physical phenomenon described by quantum mechanics. In order to realize a quantum computer, an element called a qubit that executes an operation while maintaining a quantum superposition state is used. Quantum computers are capable of computing much faster than current computers (classical computers) for problems such as factorization, and are being developed by many research institutions around the world.

  The present invention seeks to obtain an optimal pair-wise alignment score using DP more efficiently than a conventional computer by using an algorithm (see Non-Patent Documents 2 to 4) using a calculation principle of a quantum computer. To do. Therefore, a genome sequence having the highest similarity to the sequence given as an input can be found from the genome sequence database. Here, the similarity is a pair-wise alignment score.

  That is, the data search method of the present invention is a data search method for searching for a sequence most similar to a genome sequence given as an input sequence from a database storing sequences, and taking out the input sequence from the input sequence storage unit. A first qubit string; a superposition stage in which an array is extracted from a database storage unit for storing a database to form a second qubit string, and the second qubit string is superposed; and a first qubit string A quantum query step of performing a quantum operation corresponding to the dynamic programming on the second qubit string in a superposed state and obtaining a score of pairwise alignment between the input array and the array extracted from the database storage unit And having.

  A data search apparatus according to the present invention is a data search apparatus for searching a sequence most similar to a genome sequence given as an input sequence from a database storing sequences, and an input sequence storage for storing an input sequence input An output sequence storage unit that stores an output sequence to be output, a database storage unit that stores a database of genome sequences, an input temporary storage unit that temporarily stores the progress of computation, an input sequence storage unit, An arithmetic unit that writes and reads information between the output array storage unit and the input temporary storage unit and performs an operation by a quantum algorithm, and the arithmetic unit takes out the input array from the input array storage unit and Processing to make one qubit string, taking out the array from the database storage unit as a second qubit string, and superimposing the second qubit string The quantum processing corresponding to the dynamic programming is performed on the second qubit string that is superimposed with the first qubit string, and is extracted from the input array and the database storage unit. Quantum query processing for obtaining a pair-wise alignment score between the sequences.

In the present invention, when a quantum computer is used, a sequence having the highest similarity to the input genome sequence from a database in which a large number of sequences are stored is used using a conventional computer, that is, a classical computer. The effect is that it can be discovered more efficiently. The calculation speed when using the present invention is √n times faster than a conventional computer, where n is the rate of increase in the number of sequences in the database, and the cost of the calculation space is log (n). Compressed. For example, when the number of arrays in the database increases from 100 to 10,000, that is, when n = 100, the conventional computer requires 100 times the computation time and memory space required when the number of arrays is 100. And need. However, according to the present invention, it is only necessary to prepare 10 times the calculation time and log (100), that is, 2 log ₂ (10) times as much memory space. Therefore, according to the present invention, the algorithm execution efficiency increases dramatically as the number of arrays in the database increases as compared to the conventional one.

  Next, embodiments of the present invention will be described with reference to the drawings.

  FIG. 1 is a block diagram showing a configuration of a data search apparatus according to an embodiment of the present invention. This data search apparatus 100 is configured as a quantum computer that executes calculation of similarity in a genome sequence such as a base sequence or an amino acid sequence according to a quantum algorithm, and includes a calculation unit 110 that performs calculation using qubits, and a storage unit 120. The storage unit 120 stores an input array storage unit 121 that stores an input array that has been input, an output array storage unit 122 that stores an output array to be output, and a quantum algorithm for searching for a sequence having the highest degree of similarity. An algorithm storage unit 123, a database storage unit 124 that stores a genome sequence database, and an input temporary storage unit 125 that temporarily stores a calculation process. Here, the input array is an array given as an input when the quantum algorithm is executed, and the output array is an array obtained when the execution of the quantum algorithm is finished. The arithmetic unit 110 writes / reads information to / from the input array storage unit 121, the output array storage unit 122, the database storage unit 124, and the temporary input storage unit 125, and stores the quantum algorithm stored in the algorithm storage unit 123. Quantum operation is performed according to

  In this data search apparatus 100, when an input array is input, it is stored in the input array storage unit 121. The genome sequence database is stored in advance in the database storage unit 124, and an algorithm for searching for a sequence having the highest similarity is stored in the algorithm storage unit 123. Next, the output array storage unit 122 and the input temporary storage unit 125 are initialized. Thereafter, the quantum computation process is repeated in the computation unit 110 according to the quantum algorithm. During this time, the contents of the temporary input storage unit 125 are updated for each calculation step. Finally, the contents of the output array storage unit 122 are output as calculation results. As a result, in the data search apparatus 100, the sequence most similar to the input one sequence is searched from the database and output.

  Hereinafter, the quantum algorithm used in this embodiment will be described in detail.

  First, a quantum algorithm is defined.

Quantum algorithm:
A computer to which a mathematical model is given by a Turing machine, typically a von Neumann computer, is called a classic computer. A concept in a quantum computer that corresponds to a bit in a classical computer is called a qubit. Each qubit holds “1”, “0” in the classical computer, or bit information obtained by extending it (corresponding to the superposition of “1” and “0”). A quantum computer is composed of a finite number of qubits. “Calculate with a quantum computer” means to give each qubit an initial value, manipulate them to manipulate the bit information held by the qubits, and finally take these bit information by observing Means. A series of operations on bit information performed on the qubit is called a quantum algorithm. If the basic unit of operation performed on the qubit is defined, the quantum algorithm is described by this sequence of basic operations.

  Next, a database used in this embodiment is defined.

Let α be an input array (referred to as array 1) and α _i be an array (referred to as array 2) in the database. Assume that the database stores up to _n arrays α _i , that is, α ₁ , α ₂ ,..., Α _n . Each of the arrays 1 and 2 is composed of an index (index) and a table (table: entity of the array) labeled with the index. The index is distinguished by a variable i (1 ≦ i ≦ n) that means a sequence number in the database, and the table stores the encoded sequence α _i . Encoding is defined as expressing a sequence of bases or amino acids constituting a sequence by numbers with a sequence of numbers “0” and “1”.

  Next, a DP (dynamic programming) algorithm in the present embodiment for obtaining the similarity will be described.

One array 1 is given as an input array, which is stored in the input array storage unit 121, and there is an array 2 in the database stored in the database storage unit 124. The calculation of similarity will be described. Here, it is assumed that sequences 1 and 2 are base sequences, but the same procedure is adopted when they are amino acid sequences. When the number of bases constituting the sequence 1 is p and the number of bases constituting the sequence 2 is q, the sequence 1 is x ₁ , ..., x _a , ..., x _p , and the sequence 2 is y _i1 , ..., y _ib , ..., y _iq . a is a variable that designates the a-th base, and ib is a variable that designates the b-th base of the i-th sequence. The number of the base is generally counted from the N-terminal side of the sequence. In the case of an amino acid sequence, the amino acid number is generally counted from the 5 ′ end side of the sequence.

First, consider a matrix in which the base of sequence 1 is the row number of the matrix and the base of sequence 2 is the column number. Each element of this matrix is called a cell, and one alignment score f _i (a, b) is held in one cell (x _a , y _b ).

f _i (a, b) is
f _i (a, b) = max (f _i (a−1, b−1) + s (a, bi), f _i (a−1, b) −d, f _i (a, b−1) − d)… (1)
It is expressed. Here, max (A, B, C) is a symbol meaning that the largest value among A, B, and C is adopted. s (a, bi) is a substitution score when using a substitution matrix of x _{a of} sequence 1 and y _{ib of} sequence 2, and d is a gap score when deletion or insertion occurs in the sequence. is there. The substitution matrix is a table of likelihood combinations of two bases or amino acids, and some are known. In a specific example described later, BLOSUM50 is used as the permutation matrix.

The cells at this time exist from (0, 0) to (p, q), and when the row number or column number includes 0, the boundary condition is
f _i (0,0) = 0,
f _i (0, b) = − bd,
f _i (a, 0) = − ad
It is stipulated. By applying the equation (1) using such a boundary condition, the score of each cell from (1, 1) to (p, q) can be calculated, and finally _fi (p, q) can be calculated. The score f _i (p, q) is a pair-wise alignment score S (α, α _i ) indicating the similarity between the sequences 1 and 2.

  The DP algorithm described above is not an algorithm that itself is based on the application of a quantum algorithm, but is an algorithm that is normally executed on a classic computer. Next, the procedure of the quantum algorithm for executing the above DP algorithm will be described. FIG. 2 is a flowchart showing the quantum algorithm used here.

Step 1. Setting the index and λ of the threshold y:
Array 1 is input as an input array and stored in the input matrix storage unit 121. First, in step 1, the initial value of the threshold value is set to an integer y randomly selected from integers of 1 to n or less. Here, n is the number of sequences in the database. Also, λ, which is a constant that determines the increment of m, is 6/5. The threshold value is a candidate for a solution (sequence having the highest similarity to sequence 1 in sequence 2). As will be apparent from the following description, in this embodiment, the threshold value is updated by updating y, which is the index of the threshold value, when a score greater than the pairwise alignment score for the threshold value is obtained. By repeating such a process, a final solution is obtained. In the following description, a threshold value whose index is y is expressed as a threshold value y.

Step 2. Threshold table and m settings:
After execution of step 1, in step 2, table is set so as to correspond to the index of threshold value y selected in step 1. As an initial value, m, which determines the number of times the algorithm is repeated, is set as m = 1.

Step 3. Calculation of the pairwise alignment score S (α, α _y ):
After execution of step 2, in step 3, a pair-wise alignment score S (α, α _y ) corresponding to the table of the threshold value y is calculated using a classic computer. Since the pair-wise alignment score S (α, α _y ) corresponding to the threshold value y is necessary when evaluating the solution candidate obtained by the quantum algorithm, it is calculated in advance by a classical computer here. Of course, you may calculate using a quantum algorithm. As an algorithm for obtaining the pair-wise alignment score S (α, α _y ), the above-described DP algorithm can be applied.

Step 4. Random selection of an integer j that is less than m and not negative:
After step 3 is executed, step 4 randomly selects an integer j that is smaller than m and not negative. j is a variable set to repeat Steps 7 and 8 described later.

Step 5. Application of Walsh-Hadamard transform H to array 2 index:
After execution of step 4, in step 5, operation unit 110 applies Walsh-Hadamard (Walsh-Hadamard) transformation H to array 2 in database storage unit 124. At this time, the array 2 is represented as a qubit string. Walsh-Hadamard transformation H is a transformation matrix H for one qubit.

It is a conversion expressed as For the input 0, 0 and 1 are output with the same amplitude overlap state, and for the input 1 0 and 1 are output with the same amplitude overlap state that differs only in sign. The amplitude is the square root of each probability of obtaining 0 or 1 when a qubit is observed, and is generally represented by a complex number.

Step 6. Calculation of S (α, α _i ) using a quantum query :
After execution of step 5, in step 6, a score S (α, α _i ) of similarity between the arrays 1 and 2 is obtained using a quantum query. Here, the quantum query is defined as a quantum operation in which DP calculation is performed on two arrays in which both or one of them is in an overlapped state to obtain a pair-wise alignment score.

FIG. 3 shows a quantum query for obtaining the score S (α, α _i ). Here, a quantum algorithm representing a quantum query is composed of a set of qubits having a meaning such as a quantum register and a function corresponding thereto. Quantum operations are applied to qubits in chronological order from the left. In this figure, one horizontal line indicates a plurality of qubits, and a square is a function with respect to the horizontal line (target qubit) that overlaps the qubit. A black circle connected to the function by a vertical line specifies a control qubit necessary for executing the function. The quantum operation shown here is performed by the operation unit 110. The index and table (that is, x ₁ ,..., X _a ,..., X _p ) of the array 1 (array α) are supplied from the input matrix storage unit 121 to the arithmetic unit 110 as a qubit string, and the array 2 (array α _i ) Index and table (ie, y _i1 ,..., Y _{ib ,.} The permutation matrix is given as a qubit from the algorithm storage unit 123 to the calculation unit 110 as a parameter used for quantum calculation. The intermediate results of the calculation, f _i (0,0) to f _i (p, q), are stored in the temporary input storage unit 125.

Step 7. Application of quantum oracle to array 2 index:
After execution of step 6, in step 7, the arithmetic unit 110 applies the quantum oracle to the index of the array 2. A quantum oracle (oracle) is a function that outputs 0 or 1 or an overlap state of 0 and 1 with respect to an input represented by a qubit string of 0 and 1. By this function, only the sign of the index satisfying the set condition among the indexes of the array 2 in the superposed state is inverted.

Step 8. Application of diffusion transformation to array 2 index:
After execution of step 7, in step 8, operation unit 110 applies diffusion transformation to the index of array 2. The diffusion transformation D is an n-order square matrix, and the diagonal components are all -1+ (2 / n), and the other components are all 2 / n.

Step 9. j decrement:
After execution of step 8, in step 9, 1 is subtracted from j. The processing of Steps 7 and 8 described above is a Grover iteration, and by subtracting 1 from j, the iteration is repeated as many times as j determined in Step 4. I have to. As a result of this iterative operation, the difference in amplitude between the index having a higher score and the other index is enlarged.

Step 10. Conditional branch:
After execution of step 9, the value of j is evaluated in step 10, and if j is positive, the process proceeds to step 11; otherwise, the process returns to step 7.

Step 11. Observation of sequence 2:
In step 11, the index of array 2 is observed, and y ′, α _{y ′,} and S (α, α _{y ′} ) that are solution candidates are obtained.

Step 12. Check the number of execution steps of the algorithm:
After execution of step 11, in step 12, the total number of steps T executed in steps 7 and 8 is obtained, and when this T is less than O (√n), the process proceeds to the next step 13. Otherwise, y at this time is output as a solution, and the algorithm is terminated. Here, the total step number T is counted as one step for one operation on one qubit. O (√n) is a polynomial having an order of √n at most. According to the present invention, by using the quantum algorithm, it is possible to reduce the amount of calculation for searching for a sequence having the highest similarity from a database having n sequences to the order of O (√n). Which value is used as the end condition of the algorithm here depends on the implementation of the quantum computer used.

Step 13. Conditional branch:
In step 13, the value of S (α, α _{y ′} ) −S (α, α _y ) is evaluated. If this value is positive, the process proceeds to step 14, and if not, the process proceeds to step 15. Evaluation here is 'score S (alpha, alpha _y corresponding _to') a candidate y of the solution obtained in step 11, the score S (alpha, alpha _y) corresponding to the threshold y at the present time the It is to judge whether it exceeds. When the solution candidate y ′ shows a larger score, that is, when the similarity is higher, the process proceeds to step 14 in order to replace the threshold value y with y ′.

Step 14. Threshold update:
In step 14, the threshold value y is rewritten to y ', and the process returns to step 2.

Step 15. Update m:
In step 15, m is updated by substituting the smaller value of λm and √n for m, and the process returns to step 4.

  When the above quantum algorithm is completed, the threshold value index and table that are candidates for the maximum value are stored in the output matrix storage unit 122 and output.

  FIG. 4 shows, in time series, what kind of processing is performed for each array, each variable, etc. in the quantum algorithm described above. Here again, the quantum algorithm is composed of a set of qubits having a meaning such as a quantum register and a function corresponding thereto. In the figure, one horizontal line indicates a plurality of qubits, and a square is a function with respect to the horizontal line (target qubit) that overlaps the qubit. A black circle connected to the function by a vertical line specifies a control qubit necessary for executing the function. Also, “work space” (work space) in the figure is an auxiliary qubit that is temporarily used during the calculation.

《Specific examples of calculation》
Hereinafter, a specific calculation example based on this embodiment will be described. Here, the genome sequence is assumed to be an amino acid sequence, and a sequence having the highest similarity is found. Further, a permutation matrix called BLOSUM50 is used for calculating the similarity. The number of sequences given as input is 1, the number of sequences n in the database is 4, and the number of amino acids in each amino acid sequence is 3. Since 2 qubits and 20 kinds of amino acids are used to express n, 15 qubits are required to express one amino acid sequence. The sequence α given as an input with a gap score of −8 is “ARN”, and the sequences in the database are (index, table) = {(i, α _i ) | (1, ALK), It is assumed that (2, ANQ), (3, APN), (4, TWY)}. Here, each of the 20 amino acids is represented by an abbreviation consisting of one uppercase letter of the alphabet, and within the ranges described herein, A is alanine, R is arginine, N is asparagine, L is leucine, and K is lysine. , Q is glutamine, P is proline, T is threonine, W is tryptophan, and Y is tyrosine. These amino acids are replaced by a sequence of numbers 0 and 1 as qubits such as A = “00000”, R = “00001”, N = “00010”,..., V = “10011”. Of course, the input is not limited to this example.

  The following is a calculation process for outputting an array having the highest similarity in this example, corresponding to each step in the flowchart shown in FIG.

Step 1. The threshold is set to y = 2 and λ = 5/6.
Step 2. Set α _y = ANQ, m = 1.
Step 3. The alignment score is S (α, α _y ) = 4.
Step 4. Let j = 1.
Step 5. The Walsh-Hadamard transformation is applied to the index of array 2. The quantum state at this time is 1/2 {(1, ALK, 0) + (2, ANQ, 0) + (3, APN, 0) + (4, TWY, 0)}. In the quantum state notation, it is assumed that one quantum state (index, table, score) is superposed in {}.
Step 6. The quantum state when S (α, α _i ) is obtained is 1/2 {(1, ALK, 2) + (2, ANQ, 4) + (3, APN, 9) + (4, TWY, − 5)}.
Step 7. Using the quantum oracle, only the index coefficients satisfying S (α, α _i )> S (α, α _y ) are inverted. The quantum state at this time is 1/2 {(1, ALK, 2) + (2, ANQ, 4) − (3, APN, 9) + (4, TWY, −5)}.
Step 8. A diffusion transformation is applied to the index of array 2. The quantum state at this time is {(3, APN, 9)}.
Step 9. j = 0.
Step 10. Since j ≦ 0, go to Step 11.
Step 11. Observe the index of sequence 2 to obtain (3, APN, 9).
Step 12. If the total number of steps 7 and 8 has not reached O (√n), the process proceeds to step 13.
Step 13. Since S (α, α _{y ′} ) −S (α, α _y ) = 9−4> 0, the process proceeds to step 14.
Step 14. Update to y = 3.
Step 15. Update to m = 6/5 and return to step 4.

After each step is executed as described above, y = 3 and α _y = APN are output, and the algorithm is terminated.

  Here, a specific example in which the genome sequence is an amino acid sequence has been described, but the same procedure is performed when the genome sequence is a base sequence. However, since there are only four types of bases in the case of a base sequence, two qubits are required to describe one base.

  The preferred embodiments of the present invention have been described above. In addition to being executed by the quantum computer as dedicated hardware, the above-described data search apparatus records a program for realizing the function on a computer-readable recording medium, and the program recorded on the recording medium May be read by a computer system and executed. Even with a classical computer, computation time is required, but it is possible to perform verification and simulation of a quantum algorithm, so that a computer system, which is a classical computer, can execute such a program, as described above. Data search may be performed.

  The computer-readable recording medium refers to a recording medium such as a flexible disk (FD), a magneto-optical disk (MO), a CD-ROM, or a storage device such as a hard disk device built in a computer system. Furthermore, a computer-readable recording medium is one that dynamically holds a program for a short time (transmission medium or transmission wave), as in the case of transmitting a program via a network such as the Internet, in which case Some of them hold a program for a certain period of time, such as a volatile memory inside a computer system as a server.

It is a block diagram of the data search device of one embodiment of the present invention. It is a flowchart which shows the algorithm of the data search method based on this invention. It is a figure explaining the quantum query shown as a quantum algorithm. It is the figure which expressed the quantum operation which comprises a quantum algorithm in time series.

Explanation of symbols

1 to 15 Steps 100 Data Search Device 110 Arithmetic Unit 120 Storage Unit 121 Input Array Storage Unit 122 Output Array Storage Unit 123 Algorithm Storage Unit 124 Database Storage Unit 125 Input Temporary Storage Unit

Claims (10)

A data search method for searching a sequence most similar to a genome sequence given as an input sequence from a database storing sequences,
Extracting the input array from the input array storage unit into a first qubit string;
A superposition step of taking an array from a database storage unit that stores the database to form a second qubit string, and superimposing the second qubit string;
A quantum operation corresponding to dynamic programming is performed on the first qubit string and the second qubit string in the overlapped state, and between the input array and the array extracted from the database storage unit A quantum query stage to determine a pairwise alignment score;
A data search method comprising: Using a candidate sequence for the most similar sequence as a threshold,
Determining an initial value of the threshold, and determining a score by pairwise alignment between the array of the initial value and the input array;
Quantum oracle and diffusion for the second qubit string corresponding to the sequence having a score obtained in the quantum query step larger than the pairwise alignment score between the sequence corresponding to the threshold and the input sequence An iterative phase to apply the transformation;
An observation step of observing the second qubit string that has undergone the iterative operation step to obtain a solution candidate;
The data search method according to claim 1, further comprising:   If the number of sequences stored in the database storage unit is n, and the pairwise alignment score corresponding to the solution candidate is greater than the pairwise alignment score corresponding to the threshold, the solution candidate Updating the threshold, otherwise, repeating the superposition step, the quantum query step, the iterative operation step and the observation step by the number of operations increasing in the order of O (√n) with respect to n, 3. The data search method according to claim 2, wherein after that, a final solution candidate is stored in the output array storage unit as an output array. A data search device for searching a sequence most similar to a genome sequence given as an input sequence from a database storing sequences,
An input array storage unit for storing input input arrays;
An output array storage unit for storing an output array to be output;
A database storage unit for storing a genome sequence database;
An input temporary storage unit for temporarily storing the progress of the calculation;
An arithmetic unit that performs operations by writing and reading information between the input sequence storage unit, the output sequence storage unit, the database storage unit, and the input temporary storage unit, and a calculation by a quantum algorithm,
The operation unit extracts the input array from the input array storage unit to form a first qubit string, extracts the array from the database storage unit to form a second qubit string, and superimposes the second qubit string Superimposing processing for combining states, performing quantum operation corresponding to dynamic programming on the first qubit string and the second qubit string in the overlapping state, and the input array and the database storage unit Performing a quantum query process for determining a pairwise alignment score between the sequence extracted from
Data search device.   Using the sequence that is a candidate for the most similar sequence as a threshold value, the calculation unit further obtains an initial value of the threshold value, and a pairwise alignment score between the sequence of the initial value and the input sequence And after the execution of the quantum query process, the score obtained in the quantum query stage is larger than the pair-wise alignment score between the array corresponding to the threshold and the input array. Performing an iterative operation process that applies a quantum oracle and a diffusion transformation to the second qubit string to be observed, and an observation process that obtains a solution candidate by observing the second qubit string that has undergone the iterative operation process. The data search device according to claim 4.   When the number of sequences stored in the database storage unit is n, the calculation unit, if the pair-wise alignment score corresponding to the solution candidate is larger than the pair-wise alignment score corresponding to the threshold, Updating the threshold with candidate solutions, otherwise, the superposition process, the quantum query process, the iterative operation process, and the number of operations increasing in the order of O (√n) to n; The data search apparatus according to claim 5, wherein the observation process is repeated, and then a final solution candidate is stored in the output array storage unit as an output array. On the computer,
Processing to extract the input array from the input array storage unit and form a first qubit string;
A superposition process for taking out the array from the database storage unit for storing the database storing the array and setting it as a second qubit string, and setting the second qubit string to a superposed state;
A quantum operation corresponding to dynamic programming is performed on the first qubit string and the second qubit string in the overlapped state, and between the input array and the array extracted from the database storage unit Quantum query processing to find a pairwise alignment score,
A data search program that executes Using a sequence that is a candidate for the most similar sequence as a threshold,
A process for obtaining an initial value of the threshold value and obtaining a score of a pairwise alignment between the array of the initial value and the input array;
Quantum oracle and diffusion for the second qubit string corresponding to the sequence having a score obtained in the quantum query step larger than the pairwise alignment score between the sequence corresponding to the threshold and the input sequence Iterative processing to apply transformations,
An observation process for observing the second qubit string that has undergone the iterative calculation process to obtain a solution candidate;
The data search program according to claim 7, further executing: In the computer, n is the number of arrays stored in the database storage unit.
If the pairwise alignment score corresponding to the solution candidate is greater than the pairwise alignment score corresponding to the threshold, update the threshold with the solution candidate, otherwise for n The superposition process, the quantum query process, the iterative calculation process, and the observation process are repeated as many times as the number of operations increases in the order of O (√n), and then the final solution candidate is stored as an output array. The data search program according to claim 8, further causing the processing to be stored in the section to be executed. A computer-readable recording medium on which the data search program according to any one of claims 7 to 9 is recorded.

Images