Quantum computing methods and systems for protein structure prediction

Abstract

The present invention discloses a quantum computing method for predicting protein structure, comprising obtaining a protein amino acid sequence; performing coordinate generation and polar coordinate conversion on each residue and calculating interaction strength parameters to complete sequence pre-processing; constructing a continuous coordinate quantization encoder quantum circuit and realizing quantization encoding of protein continuous spatial coordinates; constructing an interaction strength quantization encoder quantum circuit and realizing quantization encoding of interaction strength between residues; iteratively updating the obtained data information to optimize the coordinate information of all residues; and performing standardization processing on the obtained data information to obtain a prediction result of the protein structure. The present invention also discloses a system for realizing the quantum computing method for predicting protein structure. The present invention not only realizes the prediction of protein structure through the quantization encoding of protein continuous spatial coordinates, the quantization encoding of interaction strength between residues and the iterative update process, but also has higher reliability, better accuracy and less computing resources occupied.

Description

Technical Field

[0001] The present invention belongs to the field of quantum technology, and specifically relates to a quantum computing method and system for protein structure prediction. Background Art

[0002] Protein structure prediction is a complex and important research field. However, protein structures are extremely complex, and conventional classical methods have always been faced with the problem of high complexity when faced with such complex technical applications.

[0003] Quantum machine learning is considered one of the cutting-edge solutions for protein structure prediction on medium-scale noisy quantum devices. However, currently used protein structure prediction methods all suffer from several drawbacks: Hamiltonian-based methods face low coding efficiency and insufficient internal information representation in practical applications. For example, the VQE (Variational Quantum Eigensolver) algorithm typically requires at least 2(n-3) qubits, and this resource requirement increases linearly with the length of the protein sequence; this limits current quantum computing devices to processing small proteins consisting of only tens of residues. Furthermore, lattice-based models struggle to effectively predict the actual spatial arrangement of amino acids in continuous space, resulting in significant differences between the encoded conformation and the native protein structure. For example, the VQE method primarily considers interactions between nearest and next-nearest neighbor residues, which limits its ability to capture information about long-range interactions. Furthermore, Hamiltonian expressions often use static parameters and lack dynamic adaptability. Summary of the Invention

[0004] One of the objectives of the present invention is to provide a quantum computing method for protein structure prediction with high reliability, good accuracy and high efficiency.

[0005] A second object of the present invention is to provide a system for implementing the quantum computing method for protein structure prediction.

[0006] The quantum computing method for protein structure prediction provided by the present invention comprises the following steps:

[0007] S1. Obtain the amino acid sequence of the protein to be processed;

[0008] S2. For the protein amino acid sequence obtained in step S1, coordinate generation and polar coordinate conversion are performed for each residue therein, and interaction strength parameters are calculated to complete sequence pre-processing;

[0009] S3. Construct a continuous coordinate quantization encoder quantum circuit, and based on the data information obtained in step S2, realize the quantization encoding of the continuous spatial coordinates of the protein;

[0010] S4. Constructing a quantum circuit for the interaction strength quantization encoder and realizing the quantization encoding of the interaction strength between residues based on the data information obtained in step S2;

[0011] S5. Iteratively updating the data information obtained in step S3 and step S4 to optimize the coordinate information of all residues;

[0012] S6. Perform normalization processing based on the data information obtained in step S5 to obtain a prediction result of the protein structure.

[0013] Step S2 is to generate coordinates and perform polar coordinate conversion on each residue in the protein sequence obtained in step S1, and calculate the interaction strength parameter to complete sequence pre-processing, which specifically includes the following steps:

[0014] Random coordinate generation and polar coordinate conversion:

[0015] For a protein of length N, the Cartesian coordinates of the corresponding structure are represented by R N×3 , Cartesian coordinates are used to represent the coordinates of each residue in three-dimensional space;

[0016] In order to describe the relative coordinates between two consecutive residues, polar coordinates are used for generation; for the i-th residue r i and the i+1th residue r i+1 , the relative polar coordinates are expressed as (ρ i ,θ i ,φ i ); where ρ i represents the radial distance, θ i represents the direction angle in spherical coordinates, φ i represents the polar angle in spherical coordinates;

[0017] In order to obtain the spatial coordinate distribution of each residue, the bond length distribution between each residue is randomly initialized using statistical prior knowledge; after normal distribution evaluation, it is ensured that the bond length above the set ratio meets the following is a normal distribution with parameters, where is the residue r j With residue r k The mean of the normal distribution of bond lengths between is the residue r j With residue r k The standard deviation of the normal distribution of bond lengths between ; μ and σ of all bond lengths are taken as the hypothesis distribution of test failure, where μ is the mean of the overall distribution of protein bond lengths and σ is the standard deviation of the overall distribution of protein bond lengths; Randomly generate ρ from the normal distribution of the characteristics i The value of θ is randomly generated from the interval [0,π]. iThe value of φ is randomly generated from the interval [0,2π] i The value of

[0018] Use the following formula to convert polar coordinates:

[0019]

[0020] Where (x, y, z) are Cartesian coordinates in three-dimensional space; (ρ, θ, φ) are spherical coordinates in three-dimensional space;

[0021] Calculate the interaction strength parameter:

[0022] set up Represents the properties of the residue itself, where is the i-th amino acid, A is a set of 22 amino acid types; the initial parameters of the quantum gate are calculated using the following formula:

[0023]

[0024] Where θ k,i is the interaction strength parameter between the amino acid at position k and the amino acid at position i in the protein; is the property corresponding to the kth amino acid in the protein sequence; is the property corresponding to the i-th amino acid in the protein sequence; σ is the adaptive scaling parameter; d k,i is the spatial Euclidean distance between two amino acids; ρ k-i| is the adjacency relationship between residues;

[0025] The initial parameters of the obtained quantum gate are normalized to complete the sequence pre-processing.

[0026] The step S3 of constructing a continuous coordinate quantized encoder quantum circuit and realizing the quantized encoding of the continuous spatial coordinates of the protein based on the data information obtained in step S2 specifically includes the following steps:

[0027] In order to i ,θ i ,φ i ) is encoded into three different qubits, and the coordinates are normalized to the interval [-1,1] to ensure that the Pauli Z operator σ z The values ​​on the projection can represent scaled polar coordinates; ρ i ∈[v min -ε,v max +ε], where v min For all ρ i The minimum value of v max For all ρ iThe maximum value of , ε is the set floating value; θ∈[0,π], φ∈[0,2π]; the following formula is used to convert the starting parameters:

[0028]

[0029] Where (ω p ,ω θ ,ω φ ) are the parameters of the quantum gate corresponding to the relative coordinate line (ρ, θ, φ) when it is initialized to the quantum state;

[0030] For quantum bit θ: the initial state is |ψ>=|0>; after the Hadamard gate processing, the state is converted to the Bell ground state The obtained Bell ground state is converted through the RY gate and expressed as Calculate the expectation value of the quantum system on the Pauli Z operator, expressed as Where E is the expected value of the quantum system, ψ is the final state of the quantum system before measurement; finally, the expected value of the θ quantum bit is post-processed, and the post-processed expression is θ' is the final classical polar coordinate azimuth;

[0031] For the quantum bit ρ: the initial state is |ψ>=|0>; after the Hadamard gate processing, the state is converted to the Bell ground state The obtained Bell ground state is converted through the RY gate and expressed as Calculate the expectation value of the quantum system on the Pauli Z operator, expressed as Where E is the expected value of the quantum system, ψ is the final state of the quantum system before measurement; finally, the expected value of the ρ quantum bit is post-processed, and the post-processed expression is Where E is the expected value of the quantum system, v max is the maximum value of ρ, v min is the minimum value of ρ, ε is the floating value considering the over-limit factor, and ρ' is the final classical polar coordinate radial distance;

[0032] For the qubit φ: the initial state is |ψ>=|0>; after the Hadamard gate processing, the state is converted to the Bell ground state The obtained Bell ground state is converted through the RY gate and expressed as Calculate the expectation value of the quantum system on the Pauli Z operator, expressed as Where E is the expectation value of the quantum system, and ψ is the final state of the quantum system before measurement; finally, the expectation value of the φ quantum bit is post-processed, and the post-processed expression is φ'=π(E+1), where φ' is the final classical polar coordinate polar angle.

[0033] The step S4 of constructing the interaction strength quantization encoder quantum circuit and implementing the quantization encoding of the interaction strength between residues based on the data information obtained in step S2 specifically includes the following steps:

[0034] An arbitrary state RY rotation gate (ASRY gate) is used to transfer all amplitudes to the source state;

[0035] According to the amplitude of the source state and the obtained interaction strength parameter, the amplitude of the quantum state is distributed using several arbitrary state RY revolving gates ASRY gates. The distribution process is expressed as

[0036]

[0037] Where |ψ> is the quantum state of the entire quantum system; ← is the update of the quantum state of the system after performing a specific operation; To apply an arbitrary state RY rotation gate to the quantum states |0> and |k> to exchange the amplitudes corresponding to these two states; |k> is the kth partial state of the quantum system |ψ>; N is the number of all partial states of the system. For a system with n quantum bits, N = 2 n ; In order to transfer the amplitude between the quantum state |k> and the quantum state |i> using the arbitrary state rotation RY gate, the transfer strength is determined by the interaction strength parameter θ k,i decision; θ k,i is the interaction strength parameter to be solved; α i is the amplitude intensity of the i-th partial state; |i> is the i-th partial state of the quantum system |ψ>;

[0038] Ultimately, the quantized encoding of the interaction strength between residues is achieved.

[0039] The arbitrary state RY revolving door ASRY door specifically includes the following steps:

[0040] Implementation and decomposition of arbitrary state RY revolving door ASRY door:

[0041] Assume that all n qubits are in superposition Apply the RY gate to states i> and |j>; the unitary matrix of the RY rotation gate ASRY gate It is described by the following unitary matrix:

[0042]

[0043] Where i<j; at positions (i,i) and (j,j), the matrix elements are At position (i,j), the matrix element is At position (j,i), the matrix element is When i=j, degenerates into the identity matrix;

[0044] When i>j, the unitary matrix is ​​expressed as

[0045] According to the characteristics of trigonometric even functions and odd functions, the unitary matrix is ​​expressed as Finally, through quantum gate decomposition, the arbitrary state RY rotation gate ASRY gate is decomposed into quantum gates and control circuits;

[0046] Consider the source state S=|S0S1…S n-1 ,S i ∈{0,1}> to the target state T = |T0T1…T n-1 ,T i ∈{0,1}> ASRY operation; through and N δ The length of determines the difference between the two binary states; setting the final different qubit δ n-1 For the bits that perform the rotation operation in big-endian mode, the remaining qubits The multi-control X-gate MCS gate is used to exchange to ensure that all qubits except the first qubit are the same; for all qubits k except the first qubit, k∈{0,1,...,N δ -1}, use the following formula to perform difference exchange:

[0047]

[0048] Where |ψ> next is the quantum state of the quantum system at the next moment after the quantum gate operation is performed; |ψ> is the state of the superposition quantum system at the previous moment; MCX k (CtrlQubits, CtrlValues) is a multi-control Pauli X-gate, with CtrlQubits as the control bit. When the value of the control bit is CtrlValues, the Pauli X-gate is executed to flip the state of the target bit k; CtrlQubits is the control bit; δ i is the ith quantum bit with a difference; CtrlValues ​​is the value triggered by the control bit; ψ i is the binary value of the i-th quantum bit in the current system;

[0049] Arrange the index order to ensure that the conditions of the arbitrary state RY revolving door ASRY door are met and Where δ0 is the first bit that differs between quantum states S and T in binary representation. is the value of this difference in the quantum system ψ, is the value of this difference in the target quantum state T;

[0050] The sequential step operation is represented as

[0051] The source and target states differ only in qubit δ0; the adjoint operation then restores the altered states to their original positions and amplitudes.

[0052] The adjoint operation of the MCX gate is itself, and the total operation of all the exchange difference steps is expressed as U swap , the total operation of all rotation steps is expressed as U rot , then the adjoint matrix operation is expressed as

[0053]

[0054] Where MCX i is a multi-control Pauli X gate that executes the Pauli X gate to flip the state of the target bit i; MCX0 is a multi-control Pauli X gate that executes the Pauli X gate to flip the state of the target bit 0; χ1(i>j) is the indicator function; U adjoint is the total number of operations for the adjoint matrix; final is the final state of the quantum system; |ψ> init is the initial state of the quantum system;

[0055] Through the combination of MCX gate and MCRY gate, the decomposition of arbitrary state RY revolving gate ASRY gate is realized.

[0056] Step S5, in which the data information obtained in steps S3 and S4 is iteratively updated to optimize all residue coordinate information, specifically includes the following steps:

[0057] For all residues, iterative updates are performed:

[0058] When ρ is the residue r i When performing iterative updates: residue r j The corresponding state is α j |j>,α j is the amplitude corresponding to state |j>, |j> is the jth sub-state of the quantum system; set the line related to ρ to |ψ>, and use the following formula for conversion:

[0059]

[0060] where MCRY(θ,CtrlQubits,CtrlValues) is the multi-control RY rotation gate; θ is the RY gate rotation parameter; Direct[i,j] is the normalized orientation between the ith residue and the jth residue; k[|ji] is the scaling parameter for different relative distance jumps between the ith residue and the jth residue; binary(j) is the binary form of the natural number j;

[0061] The transformation is performed using each interaction term as a control bit and the coordinate circuit as an entanglement circuit;

[0062] The coordinate values ​​of the classical state are obtained by evaluating the expectation values ​​of the three qubits representing the coordinate circuit on the Pauli Z;

[0063] Finally, the optimal protein structure prediction is achieved by performing N-1 iterations for each residue.

[0064] Step S6, in which the data information obtained in step S5 is subjected to normalization processing to obtain a prediction result of the protein structure, specifically comprises the following steps:

[0065] In the output stage, the normalized data is restored;

[0066] Recover the coordinate data of the converted data;

[0067] For the set extreme value data, eliminate the top-ranked data and the bottom-ranked data, calculate the mean to obtain the final relative direction, and convert it into the Cartesian coordinates of the spatial structure;

[0068] To standardize the coordinate output, all points are transformed by translation and rotation to ensure that the first three residues are located at (0,0,0), (0,0,z1), and (0,y2,z2), respectively, to complete the step-by-step alignment of the first three residues. The specific steps include the following:

[0069] For the input three-dimensional Cartesian coordinate A, first translate so that point 0 is at the origin, denoted as A′ i =A i -A0, i=0, 1, ..., n-1; A′ i is the three-dimensional Cartesian coordinate of the i-th point after translation transformation, A i is the three-dimensional Cartesian coordinate of the i-th point before transformation, A0 is the coordinate of the 0-th point;

[0070] Then rotate the structure so that point 1 is on the Z axis, creating the rotation matrix:

[0071]

[0072] Where R x (θ x ) is the rotation θ around the X axis x degrees; θ x is the angle of rotation around the X axis; R z (θ z ) is the rotation θ around the Z axis z degrees; θ z is the angle of rotation around the Z axis; (A′1x ,A′ 1y ,A′ 1z ) is the coordinate of the first point of the structure A' after translation; R is the overall rotation matrix to achieve the predetermined rotation effect;

[0073] Get the coordinate A″ after the secondary transformation i =R·A′ i ; Finally, perform a rotation transformation so that point 2 is in the YOZ plane and establish the rotation matrix:

[0074]

[0075] Where R z (θ′ z ) is the second rotation around the Z axis θ′ z degree; θ′ z is the second rotation angle around the Z axis; (A″ 2x ,A″ 2y ,A″ 2z ) are the coordinates of the second point of structure A″ after the second rotation;

[0076] Get the final coordinate A″′ i =R z (θ′ z )·A″ i .

[0077] The present invention also provides a system for implementing the quantum computing method for protein structure prediction, comprising a data acquisition module, a pre-processing module, a coordinate encoding module, an intensity encoding module, an iterative update module and a standard processing module; the data acquisition module, the pre-processing module, the coordinate encoding module, the intensity encoding module, the iterative update module and the standard processing module are connected in series in sequence; the data acquisition module is used to acquire the protein amino acid sequence to be processed and upload the data information to the pre-processing module; the pre-processing module is used to generate coordinates and perform polar coordinate conversion on each residue in the acquired protein amino acid sequence according to the received data information, and calculate the interaction strength parameter to complete the sequence pre-processing, and upload the data information to the coordinate encoding module; the coordinate encoding ... Data information is used to construct a continuous coordinate quantization encoder quantum circuit, and based on the obtained data information, realize the quantum encoding of the continuous spatial coordinates of the protein, and upload the data information to the intensity encoding module; the intensity encoding module is used to construct an interaction intensity quantization encoder quantum circuit according to the received data information, and based on the obtained data information, realize the quantum encoding of the interaction intensity between residues, and upload the data information to the iterative update module; the iterative update module is used to iteratively update the obtained data information according to the received data information to realize the optimization of the coordinate information of all residues, and upload the data information to the standard processing module; the standard processing module is used to perform standardization processing on the received data information according to the obtained data information to obtain the prediction result of the protein structure.

[0078] The quantum computing method and system for protein structure prediction provided by the present invention not only achieves protein structure prediction, but also significantly improves the accuracy and reliability of protein structure prediction through the quantization encoding of protein continuous spatial coordinates, quantization encoding of the interaction strength between residues and an iterative update process, while occupying fewer computing resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 Schematic diagram of the process flow of the present invention.

[0080] Figure 2 Schematic diagram of the circuit for quantized encoding of continuous spatial coordinates of proteins according to the method of the present invention.

[0081] Figure 3 Schematic diagram of the circuit for quantized encoding of the interaction strength between residues in the method of the present invention.

[0082] Figure 4 The figure is a circuit diagram of the arbitrary state RY revolving door ASRY door of the method of the present invention.

[0083] Figure 5 Schematic diagram of the functional modules of the system of the present invention. DETAILED DESCRIPTION

[0084] like Figure 1 The figure shows a schematic flow chart of the method of the present invention: The quantum computing method for protein structure prediction disclosed in the present invention comprises the following steps:

[0085] S1. Obtain the amino acid sequence of the protein to be processed;

[0086] S2. For the protein amino acid sequence obtained in step S1, coordinate generation and polar coordinate conversion are performed for each residue therein, and interaction strength parameters are calculated to complete sequence pre-processing; specifically, the steps include:

[0087] Random coordinate generation and polar coordinate conversion:

[0088] For a protein of length N, the Cartesian coordinates of the corresponding structure are represented by R N×3 , Cartesian coordinates are used to represent the coordinates of each residue in three-dimensional space;

[0089] In order to describe the relative coordinates between two consecutive residues, polar coordinates are used for generation; for the i-th residue r i and the i+1th residue r i+1 , the relative polar coordinates are expressed as (ρ i ,θ i ,φ i ); where ρ i represents the radial distance, θ i represents the direction angle in spherical coordinates, φ i represents the polar angle in spherical coordinates;

[0090] In order to obtain the spatial coordinate distribution of each residue, the bond length distribution between each residue is randomly initialized using statistical prior knowledge; after normal distribution evaluation, it is ensured that the bond length above the set ratio meets the following is a normal distribution with parameters, where is the residue r j With residue r k The mean of the normal distribution of bond lengths between is the residue r j With residue r k The standard deviation of the normal distribution of bond lengths between ; μ and σ of all bond lengths are taken as the hypothesis distribution of test failure, where μ is the mean of the overall distribution of protein bond lengths and σ is the standard deviation of the overall distribution of protein bond lengths; Randomly generate ρ from the normal distribution of the characteristic i The value of θ is randomly generated from the interval [0,π]. i The value of φ is randomly generated from the interval [0,2π] i The value of

[0091] Use the following formula to convert polar coordinates:

[0092]

[0093]

[0094] Where (x, y, z) are Cartesian coordinates in three-dimensional space; (ρ, θ, φ) are spherical coordinates in three-dimensional space;

[0095] Calculate the interaction strength parameter:

[0096] In order to minimize the number of parameters, set Represents the properties of the residue itself, where is the i-th amino acid, A is a set of 22 amino acid types; the initial parameters of the quantum gate are calculated using the following formula:

[0097]

[0098] Where θ k,i is the interaction strength parameter between the amino acid at position k and the amino acid at position i in the protein; is the property corresponding to the kth amino acid in the protein sequence; is the property corresponding to the i-th amino acid in the protein sequence; σ is the adaptive scaling parameter; d k,i is the spatial Euclidean distance between two amino acids; ρ k-i is the adjacency relationship between residues; with d k,i The result of the division reflects that the interaction strength decreases with increasing distance. This design follows the actual laws of biology. The final result is the interaction strength that varies with distance, fully exploiting the long-range and short-range interactions.

[0099] Normalize the initial parameters of the obtained quantum gate to complete the sequence pre-processing;

[0100] This pre-processing step provides reasonable coordinate input for the quantum model, enabling subsequent quantization encoding and optimization to achieve more accurate prediction results with lower computational complexity. By constructing initial inputs that are closer to the actual structure, it helps to improve the convergence speed of the model and reduce redundant calculations in the optimization step.

[0101] S3. Constructing a continuous coordinate quantization encoder quantum circuit, and based on the data information obtained in step S2, realizing the quantization encoding of the continuous spatial coordinates of the protein; specifically comprising the following steps:

[0102] In order to i ,θ i ,φi ) is encoded into three different qubits, and the coordinates are normalized to the interval [-1,1] to ensure that the Pauli Z operator σ z The values ​​on the projection can represent scaled polar coordinates; ρ i ∈[v min -ε,v max +ε], where v min For all ρ i The minimum value of v max For all ρ i The maximum value of , ε is the set floating value (preferably 0.02); θ∈[0,π], φ∈[0,2π]; the following formula is used to convert the starting parameters:

[0103]

[0104] Where (ω p ,ω θ ,ω φ ) are the parameters of the quantum gate corresponding to the relative coordinate line (ρ, θ, φ) when it is initialized to the quantum state;

[0105] Continuous coordinate quantization encoding circuit such as Figure 2 As shown;

[0106] For quantum bit θ: the initial state is |ψ>=|0>; after the Hadamard gate processing, the state is converted to the Bell ground state The obtained Bell ground state is converted through the RY gate and expressed as Calculate the expectation value of the quantum system on the Pauli Z operator, expressed as Where E is the expected value of the quantum system, ψ is the final state of the quantum system before measurement; finally, the expected value of the θ quantum bit is post-processed, and the post-processed expression is θ' is the final classical polar coordinate azimuth;

[0107] For the quantum bit ρ: the initial state is |ψ>=|0>; after the Hadamard gate processing, the state is converted to the Bell ground state The obtained Bell ground state is converted through the RY gate and expressed as Calculate the expectation value of the quantum system on the Pauli Z operator, expressed as Where E is the expected value of the quantum system, ψ is the final state of the quantum system before measurement; finally, the expected value of the ρ quantum bit is post-processed, and the post-processed expression is Where E is the expected value of the quantum system, v max is the maximum value of ρ, v min is the minimum value of ρ, ε is the floating value considering the over-limit factor, and ρ' is the final classical polar coordinate radial distance;

[0108] For the qubit φ: the initial state is |ψ>=|0>; after the Hadamard gate processing, the state is converted to the Bell ground state The obtained Bell ground state is converted through the RY gate and expressed as Calculate the expectation value of the quantum system on the Pauli Z operator, expressed as Where E is the expectation value of the quantum system, ψ is the final state of the quantum system before measurement; finally, the expectation value of the φ qubit is post-processed, and the post-processed expression is φ'=π(E+1), where φ' is the final polar angle in classical polar coordinates;

[0109] The coordinates encoded by the continuous coordinate quantization encoder can be accurately reproduced without additional transformation, which shows that the continuous coordinate quantization encoder can fully encode coordinate information and has strong interpretability;

[0110] Through quantum coding, the continuous coordinates of protein residues can be represented in quantum states using fewer qubit resources, significantly reducing the time complexity of subsequent calculations. Furthermore, this encoding method discretizes the originally continuous spatial information into quantum states, providing efficient data representation for subsequent quantum circuit optimization and state measurement, thereby improving the computational efficiency and accuracy of the prediction model.

[0111] S4. Constructing an interaction strength quantization encoder quantum circuit and implementing quantization encoding of the interaction strength between residues based on the data information obtained in step S2; specifically comprising the following steps:

[0112] Interaction intensity quantization encoder quantum circuit Figure 3 As shown;

[0113] An arbitrary state RY rotation gate (ASRY gate) is used to transfer all amplitudes to the source state;

[0114] According to the amplitude of the source state and the obtained interaction strength parameter, the amplitude of the quantum state is distributed using several arbitrary state RY revolving gates ASRY gates. The distribution process is expressed as

[0115]

[0116] Where |ψ> is the quantum state of the entire quantum system; ← is the update of the quantum state of the system after performing a specific operation; To apply an arbitrary state RY rotation gate to the quantum states |0> and |k> to exchange the amplitudes corresponding to these two states; |k> is the kth partial state of the quantum system |ψ>; N is the number of all partial states of the system. For a system with n quantum bits, N = 2 n ; In order to transfer the amplitude between the quantum state |k> and the quantum state |i> using the arbitrary state rotation RY gate, the transfer strength is determined by the interaction strength parameter θ k,i decision; θ k,i is the interaction strength parameter to be solved; α i is the amplitude intensity of the i-th partial state; |i> is the i-th partial state of the quantum system |ψ>;

[0117] Ultimately, the quantized encoding of the interaction strength between residues is achieved;

[0118] The larger α is, the more positive the interaction of the residue pair in the protein is, and the smaller α is, the more negative the interaction is. The above method constitutes an interaction strength quantization encoder. For N interaction strength terms that need to be encoded, the interaction strength quantization encoder can be The superposition state is encoded using qubits, rather than the N qubits used in the past, which saves exponential storage space.

[0119] This quantized superposition encoding effectively compresses the complex interaction strength information between residues, enabling the model to efficiently store the interactions between protein residues in a quantum state. This encoding not only enables the model to quantize the interaction strength using low-bit resources, but also provides efficient input for the adaptive optimization of quantum circuits, improving overall prediction accuracy and the effectiveness of spatial interaction modeling.

[0120] The arbitrary state RY revolving door ASRY door specifically includes the following steps:

[0121] The circuit of any state RY revolving door ASRY door is as follows Figure 4 As shown;

[0122] Implementation and decomposition of arbitrary state RY revolving door ASRY door:

[0123] Assume that all n qubits are in superposition Apply the RY gate to states i> and |j>; the unitary matrix of the RY rotation gate ASRY gate It is described by the following unitary matrix:

[0124]

[0125] Where i<j; at positions (i,i) and (j,j), the matrix elements are At position (i,j), the matrix element is At position (j,i), the matrix element is When i=j, degenerates into the identity matrix;

[0126] When i>j, the unitary matrix is ​​expressed as

[0127] According to the characteristics of trigonometric even functions and odd functions, the unitary matrix is ​​expressed as Finally, through quantum gate decomposition, the arbitrary state RY rotation gate ASRY gate is decomposed into quantum gates and control circuits;

[0128] Consider the source state S=|S0S1…S n-1 ,S i ∈{0,1}> to the target state T = |T0T1…T n-1 ,T i ∈{0,1} ASRY operation; by and N δ The length of determines the difference between the two binary states; setting the final different qubit δ n-1 For the bits that perform the rotation operation in big-endian mode, the remaining qubits The multi-control X-gate MCS gate is used to exchange to ensure that all qubits except the first qubit are the same; for all qubits k except the first qubit, k∈{0,1,...,N δ -1}, use the following formula to perform difference exchange:

[0129]

[0130] Where |ψ> next is the quantum state of the quantum system at the next moment after the quantum gate operation is performed; |ψ> is the state of the superposition quantum system at the previous moment; MCX k (CtrlQubits, CtrlValues) is a multi-control Pauli X-gate, with CtrlQubits as the control bit. When the value of the control bit is CtrlValues, the Pauli X-gate is executed to flip the state of the target bit k; CtrlQubits is the control bit; δ i is the ith quantum bit with a difference; CtrlValues ​​is the value triggered by the control bit; ψ i is the binary value of the i-th quantum bit in the current system;

[0131] Arrange the index order to ensure that the conditions of the arbitrary state RY revolving door ASRY door are met and Where δ0 is the first bit that differs between quantum states S and T in binary representation. is the value of this difference in the quantum system ψ, is the value of this difference in the target quantum state T;

[0132] The sequential step operation is represented as

[0133] The source and target states differ only in qubit δ0; the adjoint operation then restores the altered states to their original positions and amplitudes.

[0134] The adjoint operation of the MCX gate is itself, and the total operation of all the exchange difference steps is expressed as U swap , the total operation of all rotation steps is expressed as U rot , then the adjoint matrix operation is expressed as

[0135]

[0136] Where MCX i is a multi-control Pauli X gate that executes the Pauli X gate to flip the state of the target bit i; MCX0 is a multi-control Pauli X gate that executes the Pauli X gate to flip the state of the target bit 0; χ1(i>j) is the indicator function; U adjoint is the total operation of the adjoint matrix operation steps; final is the final state of the quantum system; |ψ> init is the initial state of the quantum system;

[0137] Through the combination of MCX gate and MCRY gate, the decomposition of arbitrary state RY revolving gate ASRY gate is realized;

[0138] S5. Iteratively update the data information obtained in step S3 and step S4 to optimize the coordinate information of all residues; specifically, the steps include:

[0139] For all residues, iterative updates are performed:

[0140] When ρ is the residue r i When performing iterative updates: residue r j The corresponding state is α j |j>,α j is the amplitude corresponding to state |j>, |j> is the jth sub-state of the quantum system; set the line related to ρ to |ψ>, and use the following formula for conversion:

[0141]

[0142] where MCRY(θ,CtrlQubits,CtrlValues) is the multi-control RY rotation gate; θ is the RY gate rotation parameter; Direct[i,j] is the normalized orientation between the ith residue and the jth residue; k[ji] is the scaling parameter for different relative distance jumps between the ith residue and the jth residue; binary(j) is the binary form of the natural number j;

[0143] The transformation is performed using each interaction term as a control bit and the coordinate circuit as an entanglement circuit;

[0144] The coordinate values ​​of the classical state are obtained by evaluating the expectation values ​​of the three qubits representing the coordinate circuit on the Pauli Z;

[0145] Finally, the optimal protein structure prediction is achieved by performing N-1 iterations for each residue;

[0146] This iterative mechanism not only improves computational efficiency but also ensures that interactions between residues are fully considered in spatial structure prediction, thereby significantly improving the accuracy and reliability of protein structure prediction. By reducing the spatial complexity from O(n) to O(log2 n), the quantum bit resource consumption of each iteration is significantly reduced.

[0147] S6. Based on the data information obtained in step S5, normalization processing is performed to obtain a prediction result of the protein structure; specifically, the steps include:

[0148] In the output stage, the normalized data is restored;

[0149] Recover the coordinate data of the converted data;

[0150] For the set extreme value data, eliminate the top-ranked data and the bottom-ranked data, calculate the mean to obtain the final relative direction, and convert it into the Cartesian coordinates of the spatial structure;

[0151] To standardize the coordinate output, all points are transformed by translation and rotation to ensure that the first three residues are located at (0,0,0), (0,0,z1), and (0,y2,z2), respectively, to complete the step-by-step alignment of the first three residues. The specific steps include the following:

[0152] For the input three-dimensional Cartesian coordinate A, first translate so that point 0 is at the origin, denoted as A′ i =A i -A0, i=0, 1, ..., n-1; A′ i is the three-dimensional Cartesian coordinate of the i-th point after translation transformation, A i is the three-dimensional Cartesian coordinate of the i-th point before transformation, A0 is the coordinate of the 0-th point;

[0153] Then rotate the structure so that point 1 is on the Z axis, creating the rotation matrix:

[0154]

[0155] Where R x (θ x ) is the rotation θ around the X axis xdegrees; θ x is the angle of rotation around the X axis; R z (θ z ) is the rotation θ around the Z axis z degrees; θ z is the angle of rotation around the Z axis; (A′ 1x ,A′ 1y ,A′ 1z ) is the coordinate of the first point of the structure A' after translation; R is the overall rotation matrix to achieve the predetermined rotation effect;

[0156] Get the coordinate A″ after the secondary transformation i =R·A′ i ; Finally, perform a rotation transformation so that point 2 is in the YOZ plane and establish the rotation matrix:

[0157]

[0158] Where R z (θ′ z ) is the second rotation around the Z axis θ′ z degree; θ′ z is the second rotation angle around the Z axis; (A″ 2x ,A″ 2y ,A″ 2z ) are the coordinates of the second point of structure A″ after the second rotation;

[0159] Get the final coordinate A″′ i =R z (θ′ z )·A″ i ;

[0160] After post-processing, the final output protein structure is highly reproducible and standardized, providing clear and stable structural data for downstream analysis.

[0161] like Figure 5The figure shows a schematic diagram of the functional modules of the system of the present invention: the system of the quantum computing method for realizing the protein structure prediction disclosed by the present invention comprises a data acquisition module, a pre-processing module, a coordinate encoding module, an intensity encoding module, an iterative update module and a standard processing module; the data acquisition module, the pre-processing module, the coordinate encoding module, the intensity encoding module, the iterative update module and the standard processing module are connected in series in sequence; the data acquisition module is used to acquire the protein amino acid sequence to be processed and upload the data information to the pre-processing module; the pre-processing module is used to generate coordinates and perform polar coordinate conversion for each residue in the acquired protein amino acid sequence according to the received data information, and calculate the interaction strength parameter to complete the sequence pre-processing, and upload the data information to the coordinate encoding module; the coordinate encoding module It is used to construct a continuous coordinate quantization encoder quantum circuit based on the received data information, and based on the obtained data information, realize the quantum encoding of the continuous spatial coordinates of the protein, and upload the data information to the intensity encoding module; the intensity encoding module is used to construct an interaction intensity quantization encoder quantum circuit based on the received data information, and based on the obtained data information, realize the quantum encoding of the interaction intensity between residues, and upload the data information to the iterative update module; the iterative update module is used to iteratively update the obtained data information based on the received data information to achieve the optimization of all residue coordinate information, and upload the data information to the standard processing module; the standard processing module is used to perform standardization processing on the obtained data information based on the received data information to obtain the prediction result of the protein structure.

Claims

1. A quantum computing method for protein structure prediction, comprising the following steps: S1. Obtain the amino acid sequence of the protein to be processed; S2. For the protein amino acid sequence obtained in step S1, coordinate generation and polar coordinate conversion are performed for each residue therein, and interaction strength parameters are calculated to complete sequence pre-processing; S3. Construct a continuous coordinate quantization encoder quantum circuit, and based on the data information obtained in step S2, realize the quantization encoding of the continuous spatial coordinates of the protein; S4. Constructing a quantum circuit for the interaction strength quantization encoder and realizing the quantization encoding of the interaction strength between residues based on the data information obtained in step S2; S5. Iteratively updating the data information obtained in step S3 and step S4 to optimize the coordinate information of all residues; S6. Perform normalization processing based on the data information obtained in step S5 to obtain a prediction result of the protein structure.

2. The quantum computing method for protein structure prediction according to claim 1, characterized in that Step S2 is to generate coordinates and perform polar coordinate conversion on each residue in the protein sequence obtained in step S1, and calculate the interaction strength parameter to complete sequence pre-processing, which specifically includes the following steps: Random coordinate generation and polar coordinate conversion: For a protein of length N, the Cartesian coordinates of the corresponding structure are represented by R N×3 , Cartesian coordinates are used to represent the coordinates of each residue in three-dimensional space; In order to describe the relative coordinates between two consecutive residues, polar coordinates are used for generation; for the i-th residue r i and the i+1th residue r i+1 , the relative polar coordinates are expressed as (ρ i ,θ i ,φ i ); where ρ i represents the radial distance, θ i represents the direction angle in spherical coordinates, φ i represents the polar angle in spherical coordinates; In order to obtain the spatial coordinate distribution of each residue, the bond length distribution between each residue is randomly initialized using statistical prior knowledge; after normal distribution evaluation, it is ensured that the bond length above the set ratio meets the following is a normal distribution with parameters, where is the residue r j With residue r k The mean of the normal distribution of bond lengths between is the residue r j With residue r k The standard deviation of the normal distribution of bond lengths between ; μ and σ of all bond lengths are taken as the hypothesis distribution of test failure, where μ is the mean of the overall distribution of protein bond lengths and σ is the standard deviation of the overall distribution of protein bond lengths; Randomly generate ρ from the normal distribution of the characteristic i The value of θ is randomly generated from the interval [0,π]. i The value of φ is randomly generated from the interval [0,2π] i The value of Use the following formula to convert polar coordinates: Where (x, y, z) are Cartesian coordinates in three-dimensional space; (ρ, θ, φ) are spherical coordinates in three-dimensional space; Calculate the interaction strength parameter: set up Represents the properties of the residue itself, where is the i-th amino acid, A is a set of 22 amino acid types; the initial parameters of the quantum gate are calculated using the following formula: Where θ k,i is the interaction strength parameter between the amino acid at position k and the amino acid at position i in the protein; is the property corresponding to the kth amino acid in the protein sequence; is the property corresponding to the i-th amino acid in the protein sequence; σ is the adaptive scaling parameter; d k,i is the spatial Euclidean distance between two amino acids; ρ |k-i| is the adjacency relationship between residues; The initial parameters of the obtained quantum gate are normalized to complete the sequence pre-processing.

3. The quantum computing method for protein structure prediction according to claim 2, characterized in that The step S3 of constructing a continuous coordinate quantized encoder quantum circuit and realizing the quantized encoding of the continuous spatial coordinates of the protein based on the data information obtained in step S2 specifically includes the following steps: In order to i ,θ i ,φ i ) is encoded into three different qubits, and the coordinates are normalized to the interval [-1,1] to ensure that the Pauli Z operator σ z The values ​​on the projection can represent scaled polar coordinates; ρ i ∈[v min -ε,v max +ε], where v min For all ρ i The minimum value of v max For all ρ i The maximum value of , ε is the set floating value; θ∈[0,π], φ∈[0,2π]; the following formula is used to convert the starting parameters: Where (ω p ,ω θ ,ω φ ) are the parameters of the quantum gate corresponding to the relative coordinate line (ρ, θ, φ) when it is initialized to the quantum state; For quantum bit θ: the initial state is |ψ>=|0>; after the Hadamard gate processing, the state is converted to the Bell ground state The obtained Bell ground state is converted through the RY gate and expressed as Calculate the expectation value of the quantum system on the Pauli Z operator, expressed as Where E is the expected value of the quantum system, ψ is the final state of the quantum system before measurement; finally, the expected value of the θ quantum bit is post-processed, and the post-processed expression is θ' is the final classical polar coordinate azimuth; For the quantum bit ρ: the initial state is |ψ>=|0>; after the Hadamard gate processing, the state is converted to the Bell ground state The obtained Bell ground state is converted through the RY gate and expressed as Calculate the expectation value of the quantum system on the Pauli Z operator, expressed as Where E is the expected value of the quantum system, ψ is the final state of the quantum system before measurement; finally, the expected value of the ρ quantum bit is post-processed, and the post-processed expression is Where E is the expected value of the quantum system, v max is the maximum value of ρ, v min is the minimum value of ρ, ε is the floating value considering the over-limit factor, and ρ' is the final classical polar coordinate radial distance; For the qubit φ: the initial state is |ψ>=|0>; after the Hadamard gate processing, the state is converted to the Bell ground state The obtained Bell ground state is converted through the RY gate and expressed as Calculate the expectation value of the quantum system on the Pauli Z operator, expressed as Where E is the expectation value of the quantum system, and ψ is the final state of the quantum system before measurement; finally, the expectation value of the φ quantum bit is post-processed, and the post-processed expression is φ'=π(E+1), where φ' is the final classical polar coordinate polar angle.

4. The quantum computing method for protein structure prediction according to claim 3, characterized in that The step S4 of constructing the interaction strength quantization encoder quantum circuit and implementing the quantization encoding of the interaction strength between residues based on the data information obtained in step S2 specifically includes the following steps: An arbitrary state RY rotation gate (ASRY gate) is used to transfer all amplitudes to the source state; According to the amplitude of the source state and the obtained interaction strength parameter, the amplitude of the quantum state is distributed using several arbitrary state RY revolving gates ASRY gates. The distribution process is expressed as Where |ψ> is the quantum state of the entire quantum system; ← is the update of the quantum state of the system after performing a specific operation; To apply an arbitrary state RY rotation gate to the quantum states |0> and |k> to exchange the amplitudes corresponding to these two states; |k> is the kth partial state of the quantum system |ψ>; N is the number of all partial states of the system. For a system with n quantum bits, N = 2 n ; In order to transfer the amplitude between the quantum state |k> and the quantum state |i> using the arbitrary state rotation RY gate, the transfer strength is determined by the interaction strength parameter θ k,i Decide; θ k,i is the interaction strength parameter to be solved; α i is the amplitude intensity of the i-th partial state; |i> is the i-th partial state of the quantum system |ψ>; Ultimately, the quantized encoding of the interaction strength between residues is achieved.

5. The quantum computing method for protein structure prediction according to claim 4, characterized in that The arbitrary state RY revolving door ASRY door specifically includes the following steps: Implementation and decomposition of arbitrary state RY revolving door ASRY door: Assume that all n qubits are in superposition Apply the RY gate to states |i> and |j>; the unitary matrix of the RY rotation gate ASRY gate It is described by the following unitary matrix: Where i<j; at positions (i,i) and (j,j), the matrix elements are At position (i,j), the matrix element is At position (j,i), the matrix element is When i=j, degenerates into the identity matrix; When i>j, the unitary matrix is ​​expressed as According to the characteristics of trigonometric even functions and odd functions, the unitary matrix is ​​expressed as Finally, through quantum gate decomposition, the arbitrary state RY rotation gate ASRY gate is decomposed into quantum gates and control circuits; Consider the source state S=|S0S1…S n-1 ,S i ∈{0,1}> to the target state T = |T0T1…T n-1 ,T i ∈{0,1}> ASRY operation; through and N δ The length of determines the difference between the two binary states; setting the final different qubit δ n-1 For the bits that perform the rotation operation in big-endian mode, the remaining qubits The multi-control X-gate MCS gate is used to exchange to ensure that all qubits except the first qubit are the same; for all qubits k except the first qubit, k∈{0,1,...,N δ -1}, use the following formula to perform difference exchange: Where |ψ> next is the quantum state of the quantum system at the next moment after the quantum gate operation is performed; |ψ> is the state of the superposition quantum system at the previous moment; MCX k (CtrlQubits, CtrlValues) is a multi-control Pauli X-gate, with CtrlQubits as the control bit. When the value of the control bit is CtrlValues, the Pauli X-gate is executed to flip the state of the target bit k; CtrlQubits is the control bit; δ i is the ith quantum bit with a difference; CtrlValues ​​is the value triggered by the control bit; ψ i is the binary value of the i-th quantum bit in the current system; Arrange the index order to ensure that the conditions of the arbitrary state RY revolving door ASRY door are met and Where δ0 is the first bit that differs between quantum states S and T in binary representation. is the value of this difference in the quantum system ψ, is the value of this difference in the target quantum state T; The sequential step operation is represented as The source and target states differ only in qubit δ0; the adjoint operation then restores the altered states to their original positions and amplitudes. The adjoint operation of the MCX gate is itself, and the total operation of all the exchange difference steps is expressed as U swap , the total operation of all rotation steps is expressed as U rot , then the adjoint matrix operation is expressed as Where MCX i is a multi-control Pauli X gate that executes the Pauli X gate to flip the state of the target bit i; MCX0 is a multi-control Pauli X gate that executes the Pauli X gate to flip the state of the target bit 0; χ1(i>j) is the indicator function; U adjoint is the total number of operations for the adjoint matrix; final is the final state of the quantum system; |ψ> init is the initial state of the quantum system; Through the combination of MCX gate and MCRY gate, the decomposition of arbitrary state RY revolving gate ASRY gate is realized.

6. The quantum computing method for protein structure prediction according to claim 5, characterized in that Step S5, in which the data information obtained in steps S3 and S4 is iteratively updated to optimize all residue coordinate information, specifically includes the following steps: For all residues, iterative updates are performed: When ρ is the residue r i When performing iterative updates: residue r j The corresponding state is α j |j>,α j is the amplitude corresponding to state |j>, |j> is the jth sub-state of the quantum system; set the line related to ρ to |ψ>, and use the following formula for conversion: where MCRY(θ,CtrlQubits,CtrlValues) is the multi-control RY rotation gate; θ is the RY gate rotation parameter; Direct[i,j] is the normalized orientation between the ith residue and the jth residue; k[|ji|] is the scaling parameter for different relative distance jumps between the ith residue and the jth residue; binary(j) is the binary form of the natural number j; The transformation is performed using each interaction term as a control bit and the coordinate circuit as an entanglement circuit; By evaluating the expectation values ​​of the three qubits representing the coordinate circuit on the Pauli Z, the coordinate values ​​of the classical state are obtained; Finally, the optimal protein structure prediction is achieved by performing N-1 iterations for each residue.

7. The quantum computing method for protein structure prediction according to claim 6, characterized in that Step S6, in which the data information obtained in step S5 is subjected to normalization processing to obtain a prediction result of the protein structure, specifically comprises the following steps: In the output stage, the normalized data is restored; Recover the coordinate data of the converted data; For the set extreme value data, eliminate the top-ranked data and the bottom-ranked data, calculate the mean to obtain the final relative direction, and convert it into the Cartesian coordinates of the spatial structure; To standardize the coordinate output, all points are transformed by translation and rotation to ensure that the first three residues are located at (0,0,0), (0,0,z1), and (0,y2,z2), respectively, to complete the step-by-step alignment of the first three residues. The specific steps include the following: For the input 3D Cartesian coordinate A, first translate so that point 0 is at the origin, represented by A' i =A i -A0, i=0,1,...,n-1;A' i is the three-dimensional Cartesian coordinate of the i-th point after translation transformation, A i is the three-dimensional Cartesian coordinate of the i-th point before transformation, A0 is the coordinate of the 0-th point; Then rotate the structure so that point 1 is on the Z axis, creating the rotation matrix: Where R x (θ x ) is the rotation θ around the X axis x degrees; θ x is the angle of rotation around the X axis; R z (θ z ) is the rotation θ around the Z axis z degrees; θ z is the angle of rotation around the Z axis; (A' 1x ,A' 1y ,A' 1z ) is the coordinate of the first point of the structure A' after translation; R is the overall rotation matrix to achieve the predetermined rotation effect; Get the coordinate A after the secondary transformation" i =R·A' i ; Finally, perform a rotation transformation so that point 2 is in the YOZ plane and establish the rotation matrix: Where R z (θ' z ) is the second rotation around the Z axis θ' z degrees; θ' z is the second rotation angle around the Z axis; (A” 2x ,A” 2y ,A” 2z ) are the coordinates of the second point of structure A″ after the second rotation; Get the final coordinate A'' i =R z (θ' z )·A” i .

8. A system for implementing the quantum computing method for protein structure prediction according to any one of claims 1 to 7, characterized in that The system comprises a data acquisition module, a pre-processing module, a coordinate encoding module, an intensity encoding module, an iterative update module and a standard processing module; the data acquisition module, the pre-processing module, the coordinate encoding module, the intensity encoding module, the iterative update module and the standard processing module are connected in series in sequence; the data acquisition module is used to acquire the protein amino acid sequence to be processed and upload the data information to the pre-processing module; the pre-processing module is used to generate coordinates and perform polar coordinate conversion for each residue in the acquired protein amino acid sequence based on the received data information, and calculate the interaction intensity parameter to complete the sequence pre-processing, and upload the data information to the coordinate encoding module; The coordinate encoding module is used to construct a continuous coordinate quantization encoder quantum circuit based on the received data information, and based on the obtained data information, realize the quantization encoding of the continuous spatial coordinates of the protein, and upload the data information to the intensity encoding module; The intensity encoding module is used to construct the interaction intensity quantization encoder quantum circuit based on the received data information, and based on the obtained data information, realize the quantization encoding of the interaction intensity between residues, and upload the data information to the iterative update module; The iterative update module is used to iteratively update the obtained data information based on the received data information to optimize the coordinate information of all residues and upload the data information to the standard processing module; The standard processing module is used to perform standardization processing based on the received data information to obtain the prediction result of the protein structure.

Images