A fast molecular conformation search method based on quantum computing

Abstract

The present invention relates to molecular conformation search, and in particular to a molecular conformation search method based on quantum computing. A molecular conformation fast search method based on quantum computing, based on the use of molecular force field method to model the interaction between atoms, uses quantum coding method to map the molecular conformation search problem with the minimum energy or scoring function to the Ising coupled spin glass model that can be effectively solved by quantum computer, and uses Hamiltonian quantum algorithm to solve the calculation problem. The present invention can improve the probability of accurately estimating ligand binding affinity or finding conformations close to cocrystallization for docking software, improve the accuracy and efficiency of virtual screening, and reflect its superiority over classical calculations.

Description

Technical Field

[0001] The present invention relates to molecular conformation search, and in particular to a molecular conformation search method based on quantum computing. Background Art

[0002] In the microscopic world, understanding the structure and dynamics of molecules is crucial to understanding their functions. The molecular conformational pathway describes how molecules transform from one conformation to another due to changes in conditions. The conformational space of molecules is complex and high-dimensional, which makes its analysis a very challenging task.

[0003] Because molecules change rapidly, it is often difficult to experimentally characterize intermediate structures and conformational trajectories. Computationally, this is also a challenging problem because physics-based simulations are very time-consuming and often fail to span a sufficient time scale to capture the complete path. For example, in the most common ligand-protein binding problem, when a ligand molecule binds to a protein through induced fit, it is usually accompanied by conformational changes in the protein, which is computationally difficult to achieve.

[0004] In order to save computational effort, molecular conformation search is often used as a method to predict intermolecular binding conformations. Molecular conformation search is common in research and applications in biology and drug research, including pharmacophore modeling, transition state search, protein folding, and de novo protein design. For example, when studying protein and ligand docking, the search space consists of all possible conformations of the protein and the ligand. With the current computational resources of classical computers, it is impossible to exhaust these search spaces. There are many strategies that attempt to sample the search space with optimal efficiency.

[0005] The mainstream implementation of conformational search is that in most molecular docking simulation programs, one of the molecules will be set as a rigid body during the molecular conformational search process. Even so, conformational search of all conformers of a ligand is a very computationally intensive process, requiring exhaustive combinations of rotatable chemical bonds and rings that rotate and flip at different angles. Currently, this virtual screening based on rigid molecular docking has become a standard tool for drug discovery to identify new lead compounds that target target biological protein molecules. In this process, most docking programs treat the target biological protein molecule as rigid and only consider the flexibility of the ligand molecule. This approach has achieved a certain degree of success in its results.

[0006] Molecular conformation search involves comprehensive sampling of conformations and a combination of structural parameters. There are mainly random search, global search and other methods. Random search is performed randomly through two main methods: Monte Carlo (MC) or genetic algorithm (GA), both of which form different conformations based on bond rotation as a degree of freedom. The molecular conformation search methods of mainstream software include: AutoDock4 applies a genetic algorithm (GA), and AutoDockFR further improves the effectiveness of GA. AutoDock Vina uses a Monte Carlo (MC) algorithm to improve runtime performance. Conformational space annealing (CSA) is also implemented in GalaxyDock as a global optimization method. GWOVina uses an improved Grey Wolf Optimizer (GWO) to replace the global search algorithm of MC used in AutoDock Vina to handle docking problems involving more flexible molecules. LeDock uses evolutionary algorithms and simulated annealing searches to find the best conformation of the ligand.

[0007] Limitations of these conformational search methods: In order to reduce the dimensionality of the search space, current molecular conformational search methods usually artificially set one molecule to a rigid structure and the other molecule as a ligand to a flexible structure. This modeling method sacrifices accuracy. Even so, the molecular conformational search space is already quite large, requiring a large amount of computing resources to carry out the search. If both the receptor molecule and the ligand molecule are considered flexible, the systematic search may produce a combinatorial explosion.

[0008] For example, innovative drug development, an important application area of ​​molecular docking, is a costly and inefficient task. Generally, more than 100,000 compounds need to be screened to obtain a new drug. At present, the number of known compound molecules has reached hundreds of millions, which provides a big data foundation for drug development, but also requires improving the efficiency of the conformation search and screening process of compound molecules.

[0009] Effective search algorithms can usually improve the docking software's ability to accurately estimate ligand binding affinity or find conformations close to cocrystallization, and improve the accuracy and efficiency of virtual screening. Quantum computing is more suitable for such problems and demonstrates its superiority over classical computing. Therefore, molecular conformation searches based on quantum computing will have the potential to fundamentally solve these problems. Summary of the invention

[0010] The present invention aims to solve the defects of the prior art and provide a method for rapid molecular conformation search based on quantum computing. The present invention can improve the accuracy and efficiency of virtual screening for docking software to estimate ligand binding affinity or find the probability of close cocrystallization conformation, and reflect its superiority in speed over classical computing.

[0011] The technical solution of the present invention is achieved as follows: a method for rapid molecular conformation search based on quantum computing, based on modeling the interaction between atoms using a molecular force field method, uses a quantum coding method to map the molecular conformation search problem with the minimum energy or scoring function to an Ising coupled spin glass model that can be effectively solved by a quantum computer, and uses a Hamiltonian quantum algorithm to solve the computational problem.

[0012] The quantum coding method for rapid molecular conformation search based on quantum computing refers to expressing the possible bond angles between adjacent chemical bonds in a molecule in the form of a binary number.

[0013] The method for rapid molecular conformation search based on quantum computing adopts the Hamiltonian quantum algorithm to solve the computational problem, and the required Hamiltonian is quantitatively obtained based on three principles: interaction given by the molecular force field, non-overlapping atomic positions, and symmetry.

[0014] The method for rapid molecular conformation search based on quantum computing adopts Hamiltonian quantum algorithm to solve the computational problem, which is a Hamiltonian ground state problem. The problem is solved on a quantum computer through one of quantum amplitude amplification, quantum adiabatic quenching, variational quantum computing, QAOA and quantum Lanczos algorithm.

[0015] The method for rapid molecular conformation search based on quantum computing includes decomposing a target molecule into m basic group units. The global conformation of the molecule is determined by the rotation angle of the chemical bonds of the adjacent groups. Therefore, the conformation search needs to calculate m-1 rotation angles, perform conformation sampling at a certain angle α, and convert each chemical bond 2 n Different rotation angles are encoded into n qubits, where 2 n = 180° / α, combined with the spatial position of the group corresponding to the quantum bit information, the interaction between the groups will be expressed as the coupling between the quantum bits H int In quantum computing, it is still necessary to consider constraints such as the groups cannot overlap in space and the removal of unreasonable rotation angles. These constraints are achieved by adding penalty terms in the quantum Hamiltonian, which are H olap and H constraint ; The overall Hamiltonian is expressed as follows:

[0016] H=H int +H olap +H constraint；

[0017] Based on the constructed quantum Hamiltonian, the corresponding quantum algorithm is used to obtain the global optimal molecular conformation.

[0018] The search for the global optimal molecular conformation is essentially an NP-hard computational problem, and the size of the search space and the corresponding computational complexity grow exponentially with the number of atoms that make up the molecule. Classical computing uses molecular dynamics simulation to reduce computational complexity, but its results cannot guarantee global optimality. Quantum computing has natural computational parallelism due to the principle of quantum superposition, and can perform holistic parallel operations in the entire molecular conformation space. Compared with existing classical algorithms, it can achieve exponential quantum acceleration, especially for large-scale molecular conformation searches, and has significant computational superiority.

[0019] The present invention converts the molecular conformation search problem into a Hamiltonian ground state problem, which can be solved on a quantum computer through algorithms such as quantum amplitude amplification, quantum adiabatic quenching, variational quantum computing, QAOA, and quantum Lanczos.

[0020] The search algorithm of the present invention can generally improve the probability of accurately estimating ligand binding affinity or finding a conformation close to cocrystallization for docking software, and improve the accuracy and efficiency of virtual screening. Quantum computing is more suitable for such problems and demonstrates its superiority over classical computing. Therefore, molecular conformation search based on quantum computing will have the potential to fundamentally solve these problems, especially in terms of improving efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The present invention will be further described below in conjunction with the accompanying drawings:

[0022] Figure 1 It is a flow chart of the present invention.

[0023] Figure 2 and Figure 3 It is a molecular structure diagram and molecular schematic diagram of the proteolysis targeting chimera (PROTAC) described in the examples. DETAILED DESCRIPTION

[0024] Below, the present invention is further described in conjunction with embodiments.

[0025] like Figure 1 As shown, the method for rapid molecular conformation search based on quantum computing described in the present invention is to determine the configuration space modeling and analysis of the bond angles of adjacent chemical bonds according to the chain structure of the molecule. The size of the configuration space is bounded by 1EB. If it is less than or equal to 1EB, a classical computer can be used to determine the search; if it is larger than 1EB, quantum computing is used. Conventionally, the classical Monte Carlo calculation method is used. The present invention uses the encoding of the molecular bond angle configuration, maps the molecular conformation search problem with the minimum energy or scoring function to the quantum computer, and uses the quantum Hamilton algorithm to calculate the molecular configuration.

[0026] The encoding of the molecular bond angle configuration is to decompose the molecule into m group units and discretize the chemical bond angles of adjacent groups into 2 n Different possibilities are represented by n quantum bits, and all the interactions between groups are defined. The quantum coupling between quantum bits is used to construct the quantum algorithm to solve the Hamiltonian quantum coding required for molecular conformation search. In this way, the code will represent the possible bond angles between adjacent chemical bonds in the molecule in the form of a binary number. The Hamiltonian quantum algorithm is further used to solve the calculation problem. The Hamiltonian required for this step can be quantitatively obtained based on the three principles of interaction given by the molecular force field, non-overlapping atomic positions, and symmetry.

[0027] Example

[0028] The protein hydrolysis targeting chimera PROTAC was selected as the calculation object, and its molecular structure and molecular schematic diagram are shown as follows: Figure 2 , Figure 3 shown.

[0029] First, the molecule is decomposed into 11 basic group units. The global conformation of the molecule is determined by the rotation angle of the chemical bonds of adjacent groups, so the conformation search needs to calculate 10 rotation angles. Conformation sampling is performed every 5.625° (this parameter is adjustable), and the configuration space dimension is 32 10 Approximately 10 15 , as the number of groups in the chain molecule increases exponentially, classical calculations are difficult to accurately handle. We use quantum computing methods to encode the 32 different rotation angles of each chemical bond into 5 quantum bits. Figure 2 The molecules in the equation require 50 qubits to encode. Combined with the spatial position of the groups corresponding to the qubit information, the interaction between the groups will be represented as the coupling H between the qubits. int In quantum computing, it is still necessary to consider constraints such as the groups cannot overlap in space and the removal of unreasonable rotation angles. These constraints are achieved by adding penalty terms in the quantum Hamiltonian, which are H olap , H constraint The overall Hamiltonian is expressed as follows:

[0030] H=H int +H olap +H constraint .

[0031] Quantum algorithms such as quantum amplitude amplification, adiabatic quantum annealing, and variational quantum circuits are used to perform quantum Hamiltonian ground state operations to obtain the global optimal conformation of the target molecule.

[0032] Based on the constructed quantum Hamiltonian, the corresponding quantum algorithm is used to obtain the global optimal molecular conformation. Tests show that the conformation search results can be obtained within 1ms using a quantum annealer with a coupling strength of 10GHz.

Claims

1. A method for rapid molecular conformation search based on quantum computing, characterized in that: Based on the molecular force field method to model the interaction between atoms, the quantum coding method is used to map the molecular conformation search problem with the minimum energy or scoring function to the Ising coupled spin glass model that can be effectively solved by quantum computers, and the Hamiltonian quantum algorithm is used to solve the calculation problem. The quantum coding method refers to representing the possible bond angles between adjacent chemical bonds in a molecule in the form of a binary number. The molecular conformation rapid search method includes decomposing a target molecule into m basic group units. The global conformation of the molecule is determined by the rotation angle of the chemical bonds of the adjacent groups. Therefore, the conformation search needs to calculate m-1 rotation angles, perform conformation sampling at a certain angle α, and convert each chemical bond 2 n Different rotation angles are encoded into n qubits, where 2 n = 180° / α, combined with the spatial position of the group corresponding to the quantum bit information, the interaction between the groups will be expressed as the coupling between the quantum bits H int In quantum computing, it is still necessary to consider that the groups cannot overlap in space and remove the unreasonable rotation angle constraints. These constraints are realized by adding penalty terms in the quantum Hamiltonian. olap and H constraint ; The overall Hamiltonian is expressed as follows: H=H int +H olap +H constraint ; Based on the constructed quantum Hamiltonian, the corresponding quantum algorithm is used to obtain the global optimal molecular conformation.

2. A method for rapid molecular conformation search based on quantum computing according to claim 1, characterized in that: The computational problem solved using the Hamiltonian quantum algorithm is a Hamiltonian ground state problem, which is solved on a quantum computer by one of the following methods: quantum amplitude amplification, quantum adiabatic quenching, variational quantum computing, QAOA, or quantum Lanczos algorithm.

Images