Drug molecule property prediction method and related device

Abstract

The invention discloses a drug molecule property prediction method and a related device, and belongs to the technical field of quantum computation.The method comprises the steps that expected values of various edge types in a molecular structure diagram are calculated through a quantum circuit, the expected values serve as first edge features of the molecular structure diagram, and edges in the molecular structure diagram are used for representing chemical bonds of drug molecules; performing noise reduction on the first edge feature through a clustering algorithm to obtain a second edge feature; constructing a new molecular structure diagram according to the second edge features and original node features in the molecular structure diagram; and performing chemical property prediction on the drug molecules according to the graph characteristics of the new molecular structure graph. According to the method, the QRGCN model is applied to drug molecule property prediction for the first time, a brand new possibility is opened for quantum acceleration processing of complex graph data, interference of quantum noise is effectively inhibited through application of the clustering method, and the robustness and calculation performance of the QRGCN model in a real quantum environment are greatly improved.

Description

Technical Field

[0001] The present invention belongs to the field of quantum computing technology, and in particular relates to a method for predicting the properties of drug molecules and related devices. Background Art

[0002] In the current drug development process, being able to accurately predict the various properties of candidate drug molecules, such as drug solubility and toxicity, is crucial to accelerating drug screening. As an advanced quantum neural network model, the Quantum Relational Graph Convolutional Network (QRGCN) uses the advantages of quantum computing to efficiently process molecular structure graphs, which is expected to significantly improve prediction performance. When the QRGCN model is applied to the prediction of drug molecular properties, due to the inherent noise in the quantum chip, the QRGCN model cannot be calculated according to the processing logic of the traditional relational graph convolutional network, which greatly reduces the performance of the model.

[0003] To sum up, the technical problem that needs to be solved urgently is to propose a method to reduce the impact of quantum chip noise to improve the performance advantage of the QRGCN model in complex data processing. Summary of the invention

[0004] The purpose of the present invention is to provide a method and related device for predicting the properties of drug molecules, aiming to suppress the interference of inherent noise in quantum chips and greatly improve the robustness and computational performance of the QRGCN model in a real quantum environment.

[0005] One embodiment of the present invention provides a method for predicting the properties of drug molecules, the method comprising: Calculating expected values ​​of various edge types in a molecular structure graph through quantum circuits, and using the expected values ​​as first edge features of the molecular structure graph, wherein the edges in the molecular structure graph are used to represent chemical bonds of drug molecules; De-noising the first edge feature by using a clustering algorithm to obtain a second edge feature; Constructing a new molecular structure graph according to the second edge feature and the original node feature in the molecular structure graph; The chemical properties of the drug molecule are predicted based on the graph features of the new molecular structure graph.

[0006] Optionally, the calculation of expected values ​​of various edge types in the molecular structure graph by quantum circuits includes: Acquiring original edge features of a molecular structure graph, and encoding the original edge features into parameters of a quantum logic gate; Configuring the quantum circuit according to the parameters, and running the quantum circuit to obtain an output quantum state; The expected values ​​of various edge types in the molecular structure graph are calculated according to the output quantum state.

[0007] Optionally, encoding the original edge features into parameters of a quantum logic gate includes: The original edge features are encoded into parameters of a quantum logic gate according to a mapping function, wherein the mapping function is:

[0008] Among them, the is the encoding weight, is the coding bias, the is the jth edge feature of the i-th molecular structure graph.

[0009] Optionally, there are multiple first edge features, and the step of reducing noise on the first edge features by using a clustering algorithm to obtain the second edge features includes: Clustering the plurality of first edge features by using a clustering algorithm to obtain a clustering result; According to the clustering result, the plurality of first edge features are classified into predefined edge categories, and an index is generated for each edge category to obtain a second edge feature.

[0010] Optionally, before predicting the chemical properties of the drug molecule according to the graph features of the new molecular structure graph, the method further includes: The graph feature is determined according to a graph convolution calculation formula, where the graph convolution calculation formula is:

[0011] Among them, 𝐴 is the adjacency matrix determined based on the second edge feature, 𝐷 is the degree matrix, is the node feature of the l-th layer graph convolutional network, is the weight of the l-th layer graph convolutional network, and 𝜎 is the activation function.

[0012] Another embodiment of the present invention provides a device for predicting the properties of drug molecules, the device comprising: A calculation unit is used to calculate the expected value of various edge types in the molecular structure graph through quantum circuits, and use the expected value as the first edge feature of the molecular structure graph. Edges are used to represent chemical bonds of drug molecules; A denoising unit, configured to denoise the first edge feature by using a clustering algorithm to obtain a second edge feature; A reconstruction unit, used for constructing a new molecular structure graph according to the second edge feature and the original node feature in the molecular structure graph; The prediction unit is used to predict the chemical properties of the drug molecule according to the graph features of the new molecular structure graph.

[0013] Optionally, in the aspect of calculating the expected values ​​of various edge types in the molecular structure graph by using quantum circuits, the calculation unit specifically includes: Acquiring original edge features of a molecular structure graph, and encoding the original edge features into parameters of a quantum logic gate; Configuring the quantum circuit according to the parameters, and running the quantum circuit to obtain an output quantum state; The expected values ​​of various edge types in the molecular structure graph are calculated according to the output quantum state.

[0014] Optionally, in encoding the original edge feature into a parameter of a quantum logic gate, the calculation unit specifically includes: The original edge features are encoded into parameters of a quantum logic gate according to a mapping function, wherein the mapping function is:

[0015] Among them, the is the encoding weight, is the coding bias, the is the jth edge feature of the i-th molecular structure graph.

[0016] Yet another embodiment of the present invention provides an electronic device, comprising: a processor and a memory; The processor is connected to a memory, wherein the memory is used to store a computer program, and the processor is used to call the computer program to execute the method described in any one of the above embodiments.

[0017] Yet another embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by a processor, the method described in any one of the above embodiments is executed.

[0018] Compared with the prior art, the present invention proposes a method for predicting the properties of drug molecules. First, the expected values ​​of various edge types in a molecular structure graph are calculated through quantum circuits, and the expected values ​​are used as the first edge features of the molecular structure graph. The edges in the molecular structure graph are used to represent the properties of drug molecules. chemical bond; then the first edge feature is denoised by a clustering algorithm to obtain a second edge feature; then a new molecular structure graph is constructed based on the second edge feature and the original node features in the molecular structure graph; finally, the chemical properties of the drug molecule are predicted based on the graph features of the new molecular structure graph.

[0019] The present invention first calculates the expected values ​​of various edge types in a molecular structure graph through quantum circuits, and uses the expected values ​​as the first edge features in the molecular structure graph. The chemical bond types in drug molecules are simulated through quantum circuits and the expected values ​​of the chemical bond types are calculated as the first edge features, thereby providing more accurate basic data for subsequent predictions. The first edge features are subjected to denoising processing through a clustering algorithm to obtain second edge features. Clustering effectively identifies and removes noise and redundant information in the first edge features, thereby improving the reliability of data. The denoised second edge features can more accurately reflect the key information of drug molecules, thereby facilitating the accuracy of subsequent analysis. A new molecular structure graph is constructed by combining the denoised second edge features and the original node features. The new molecular structure graph not only retains the original structural information, but also incorporates the accuracy of quantum calculation and the denoising effect of the clustering algorithm, thereby providing a more reliable basis for subsequent chemical property predictions. Finally, the chemical property predictions of the drug molecules are performed based on the graph features of the new molecular structure graph, thereby more accurately reflecting the real structure and properties of the drug molecules, improving the accuracy of predictions, and accelerating the drug screening and optimization process. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 A network block diagram of a drug molecule property prediction system provided in an embodiment of the present invention.

[0021] Figure 2 A flowchart of a method for predicting the properties of drug molecules provided in an embodiment of the present invention.

[0022] Figure 3 A flowchart of a method for calculating an edge type expected value provided by an embodiment of the present invention.

[0023] Figure 4 A flowchart of an edge feature clustering method provided by an embodiment of the present invention.

[0024] Figure 5 A schematic diagram of the structure of a drug molecule property prediction device provided by an embodiment of the present invention.

[0025] Figure 6 A schematic structural diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0026] The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, but should not be construed as limiting the present invention.

[0027] Figure 11 is a network block diagram of a drug molecule property prediction system provided by an embodiment of the present invention. The drug molecule property prediction system may include a network 110, a server 120, a wireless device 130, a client 140, a storage 150, a classical computing unit 160, a quantum computing unit 170, and may also include additional memories, classical processors, quantum processors, and other devices not shown.

[0028] The network 110 is a medium for providing communication links between various devices and computers connected together in the drug molecule property prediction system, including but not limited to the Internet, corporate intranet, local area network, mobile communication network and their combination. The connection method can be wired, wireless communication links or optical fiber cables.

[0029] The server 120, the wireless device 130 and the client 140 are conventional data processing systems, which may contain data and applications or software tools for performing conventional computing processes. The client 140 may be a personal computer or a network computer, so the data may also be provided by the server 120. The wireless device 130 may be a smart phone, a tablet, a laptop, a smart wearable device, etc. The storage unit 150 may include a database 151, which may be configured to store data such as quantum bit parameters, quantum logic gate parameters, quantum circuits, quantum programs, etc.

[0030] The classical computing unit 160 (quantum computing unit 170) may include a classical processor 161 (quantum processor 171) for processing classical data (quantum data) and a memory 162 (memory 172) for storing classical data (quantum data). The classical data (quantum data) may be a boot file, an operating system image, and an application 163 (application 173). The application 163 (application 173) may be used to implement a quantum algorithm compiled by a method for predicting the properties of drug molecules provided in an embodiment of the present invention.

[0031] Any data or information stored or generated in the classical computing unit 160 (quantum computing unit 170) may also be configured to be stored or generated in another classical (quantum) processing system in a similar manner, and any application program executed therein may also be configured to be executed in another classical (quantum) processing system in a similar manner.

[0032] It should be noted that a true quantum computer is a hybrid structure, which includes at least Figure 1 There are two major parts in it: the classical computing unit 160, which is responsible for performing classical computing and control; the quantum computing unit 170, which is responsible for running quantum programs and thus realizing quantum computing.

[0033] The classical computing unit 160 and the quantum computing unit 170 may be integrated into one device or distributed in two different devices. For example, the first device including the classical computing unit 160 runs a classical computer operating system, on which quantum application development tools and services are provided, as well as storage and network services required for quantum applications. Users develop quantum programs through the quantum application development tools and services thereon, and send quantum programs to the second device including the quantum computing unit 170 through the network services thereon. The second device runs a quantum computer operating system, and the code of the quantum program is parsed and compiled into instructions that can be recognized and executed by the quantum processor 170 through the quantum computer operating system. The quantum processor 170 implements the quantum algorithm corresponding to the quantum program according to the instructions.

[0034] The computing unit of the classic processor 161 in the classic computing unit 160 is a CMOS tube based on a silicon chip. This computing unit is not limited by time and coherence, that is, this computing unit is not limited by the length of use and can be used at any time. In addition, in the silicon chip, the number of such computing units is also sufficient. At present, the number of computing units in a classic processor 161 is tens of thousands. The number of computing units is sufficient and the computing logic that can be selected by the CMOS tube is fixed, such as: AND logic. When calculating with the help of CMOS tubes, a large number of CMOS tubes are combined with limited logic functions to achieve the calculation effect.

[0035] The basic computing unit of the quantum processor 171 in the quantum computing unit 170 is a quantum bit. The input of the quantum bit is limited by coherence and coherence time, that is, the quantum bit is limited by the duration of use and is not available at any time. Making full use of the quantum bit within the available duration of use of the quantum bit is a key problem in quantum computing. In addition, the number of quantum bits in a quantum computer is one of the representative indicators of the performance of the quantum computer. Each quantum bit realizes the computing function through the logical function configured on demand. In view of the limited number of quantum bits, the logical functions in the field of quantum computing are diverse, such as: Hadamard gate (Hadamard gate, H gate), Pauli-X gate (X gate), Pauli-Y gate (Y gate), Pauli-Z gate (Z gate), X gate, RY gate, RZ gate, CNOT gate, CR gate, iSWAP gate, Toffoli gate, etc. In quantum computing, it is necessary to use limited quantum bits combined with a variety of logical function combinations to achieve the computing effect.

[0036] Based on these differences, the design of CMOS tubes by classical logic functions and the design of quantum bits by quantum logic functions are significantly and essentially different. The design of CMOS tubes by classical logic functions does not need to consider the individuality of CMOS tubes. For example, the representation of CMOS tubes in silicon chips is the individual identification, position, and usable life of each CMOS tube. Therefore, the classical algorithm composed of classical logic functions only expresses the operational relationship of the algorithm, but does not express the algorithm's dependence on individual CMOS tubes.

[0037] When quantum logic functions act on qubits, the individuality of qubits needs to be considered, such as the individual identification of the qubit in the quantum chip, its position, its relationship with surrounding qubits, and the usable time of each qubit. Therefore, the quantum algorithm composed of quantum logic functions not only expresses the computational relationship of the algorithm, but also expresses the algorithm's dependence on individual qubits.

[0038] Quantum chips can include quantum bits and channels for regulating quantum bits. Quantum logic gates are implemented through analog signals. Different combinations of analog signals are applied to quantum bits through channels for regulating quantum bits, thereby realizing quantum circuits with different functions and completing data processing. Therefore, the design of quantum logic functions acting on quantum bits (including the design of whether quantum bits are used or not and the design of the efficiency of each quantum bit) is the key to improving the computing performance of quantum computers, and special designs are required. This is also the uniqueness of quantum algorithms based on quantum logic functions, which is essentially and significantly different from classical algorithms based on classical logic functions. The above-mentioned design for quantum bits is a technical issue that ordinary computing devices do not need to consider or face.

[0039] In the field of quantum computing, when the QRGCN model is run on a real quantum chip, the inherent noise of the actual quantum chip will disrupt the quantum state representation of the chemical bond type, causing the QRGCN model to fail to work as expected. It runs as expected; however, the QRGCN model theoretically requires a clear and fixed definition of edge types, so the type discretization caused by noise interference becomes a key obstacle to its application in a real quantum environment.

[0040] Therefore, in order to overcome the inherent noise interference of quantum chips and ensure the effective application of the QRGCN model in a real quantum environment, the present invention proposes a method, device, electronic device and computer-readable storage medium for predicting drug molecular properties.

[0041] See also Figure 2 , Figure 2 A method for predicting the properties of drug molecules provided in an embodiment of the present invention comprises the following steps: Step S201, calculating expected values ​​of various edge types in a molecular structure graph through quantum circuits, and using the expected values ​​as first edge features of the molecular structure graph, where the edges in the molecular structure graph are used to represent chemical bonds of drug molecules; Among them, the edge refers to the connection relationship between atoms in the molecular structure diagram. The atoms are usually connected by chemical bonds. The edge type is the type of chemical bond. For example, the edge type can include single chemical bond, double chemical bond, multiple chemical bond, etc.; the first edge feature refers to the expected value of these chemical bonds obtained by quantum circuit simulation calculation using quantum computing technology.

[0042] Specifically, quantum circuits are used to simulate the chemical bond types in drug molecules and calculate the expected values ​​of the chemical bond types. These expected values ​​reflect the average performance of chemical bonds in quantum systems, such as key properties such as strength and stability. The calculated expected values ​​are then used as the first edge features of the molecular structure diagram, providing important input information for subsequent predictions of drug molecular properties.

[0043] For example, suppose a new drug molecule "X-Drug" is being studied. In order to gain a deeper understanding of the chemical properties and potential biological activity of the drug molecule, quantum computing technology is used to predict its properties. First, the expected values ​​of different edges in the molecular structure diagram are calculated by constructing quantum circuits. The edges represent chemical bonds in the molecular structure diagram, including single bonds, double bonds, and aromatic bonds. The calculated expected values ​​represent the average performance of chemical bonds in the quantum system, which can reflect the key properties of chemical bonds such as strength and stability. For example, it may be found that the expected value of a double bond is higher, indicating that it exhibits stronger interaction and stability in the quantum system, while the expected value of a single bond may be lower, meaning that its strength is weaker; the calculated expected value is used as the first edge feature of the "X-Drug" molecular structure diagram. These features provide important input information for subsequent machine learning models, which are used to predict various properties of the "X-Drug" molecule, such as anticancer activity, drug metabolic kinetic parameters, etc.

[0044] Step S202, performing noise reduction on the first edge feature by using a clustering algorithm to obtain a second edge feature; Among them, the clustering algorithm refers to dividing the first edge feature data set into K clusters through the K-means algorithm, randomly selecting K data points as the initial centroids, and iterating the data based on the Euclidean distance until the centroid remains unchanged, the algorithm converges, and the clustering is completed.

[0045] Specifically, the K-means algorithm is used to effectively identify and remove noise and redundant information in the first edge features, thereby obtaining more accurate second edge features as an important input for subsequent data prediction and analysis.

[0046] Step S203, constructing a new molecular structure graph according to the second edge feature and the original node feature in the molecular structure graph; The original node features refer to the attribute values ​​inherent to the atomic nodes in each molecular structure diagram without any transformation.

[0047] Specifically, the second edge feature set after denoising is integrated with the original node features to form a complete molecular structure representation.

[0048] For example, taking the ethane molecule as an example, assuming that its original node features only include atom types, carbon or hydrogen, these features are represented by one-hot encoding, the feature vector of carbon atom (C) = [1, 0], the feature vector of hydrogen atom (H) = [0,1], and the expected values ​​of various edge types in the ethane molecule are obtained through quantum circuit calculation, assuming that the expected value of CC bond = 0.9, and the expected value of CH bond = 0.8. In this example, it is assumed that denoising has no significant effect on the expected value, so the first edge feature (expected value) is directly used as the second edge feature, but in actual applications, denoising may change these values; finally, a new molecular structure graph is constructed, integrating the second edge feature set with the original node feature set. For ease of explanation, this process is conceptualized as a "weighted graph". The new ethane molecular structure graph will have two carbon atom nodes (each node has a feature vector of [1,0]) and six hydrogen atom nodes (each node has a feature vector of [0, 1]), and the edges will include two CC bonds (each with a weight of 0.9) and six CH bonds (each with a weight of 0.8).

[0049] Step S204: predicting the chemical properties of the drug molecules according to the graph features of the new molecular structure graph.

[0050] Among them, graph features refer to the more accurate feature representation extracted by the classic convolutional neural network through the comprehensive graph representation of the new molecular structure graph.

[0051] Specifically, the comprehensive graph representation of the newly constructed molecular structure graph is input into the classical convolutional neural network to obtain graph features, which are then converted into label probabilities through an activation function, and the chemical property prediction results of the drug molecules are obtained through the label probabilities.

[0052] Exemplarily, the comprehensive graph vector representation of the constructed new molecular structure graph is input into a classical convolutional neural network, which can extract key features in the molecular structure graph, such as the type of node (different atom types), the type of edge (different types of chemical bonds), and the neighbor relationship of the node, and encode these features into a more precise feature vector representation, namely, graph features; the graph features are then sent to the output layer of the model and converted into label probabilities through an activation function (such as the softmax function). The label probability represents the possibility that the drug molecule has various chemical properties. For example, for the property of anti-tumor activity, the model may output a probability value between 0 and 1. The closer to 1, the more likely the drug molecule is to have anti-tumor activity.

[0053] In summary, the present invention first calculates the expected values ​​of various edge types in the molecular structure graph through quantum circuits, and uses the expected values ​​as the first edge features in the molecular structure graph. The chemical bond types in the drug molecules are simulated by quantum circuits and the expected values ​​of the chemical bond types are calculated as the first edge features, providing more accurate basic data for subsequent predictions; the first edge features are denoised by a clustering algorithm to obtain second edge features, and clustering effectively identifies and removes noise and redundant information in the first edge features, thereby improving the reliability of the data. The denoised second edge features can more accurately reflect the key information of the drug molecules, which is beneficial to the accuracy of subsequent analysis; a new molecular structure graph is constructed by combining the denoised second edge features and the original node features, and the new molecular structure graph not only retains the original structural information, but also incorporates the accuracy of quantum computing and the denoising effect of the clustering algorithm, thereby providing a more reliable basis for subsequent chemical property predictions; finally, the chemical properties of the drug molecules are predicted based on the graph features of the new molecular structure graph, which can more accurately reflect the real structure and properties of the drug molecules, improve the accuracy of the prediction, and accelerate the drug screening and optimization process.

[0054] Step S301, obtaining original edge features of a molecular structure graph, and encoding the original edge features into parameters of a quantum logic gate; Among them, the original edge feature refers to the molecular structure diagram of the drug molecule obtained through chemical software or database. The structure diagram describes the connection relationship between atoms in the drug molecule, that is, chemical bonds (edges). These edges have various attributes, such as the type of bond (single bond, double bond, triple bond, etc.). These attribute features are then converted into numerical vector representations. For example, a unique integer value can be assigned to each different bond type, or a multidimensional vector can be used to represent continuous features such as bond length and bond angle, that is, the original edge feature.

[0055] Specifically, the characteristic vector representation of the original edge in the molecular structure graph is first obtained, and then converted into the parameters of the quantum logic gate through a specific mapping function.

[0056] Exemplarily, the original edge features are encoded into the parameters of the quantum logic gate through a mapping function. The specific mapping function is ,in is the encoding weight, is the coding bias, is the jth edge feature of the ith molecular structure graph. Assume that there are two edge features in the molecular structure graph of compound A, and the bond length and bond angle , encoding these two features into the parameters of two quantum logic gates and , assuming the encoding weights = 0.5, = 1.0 and the encoding bias =0.1, =0.2, the bond length is 1.5, and the bond angle is 120 degrees; the corresponding quantum logic gate parameters can be calculated as: = 0.5 *1.5 + 0.1 = 0.85, = 1.0 *120 + 0.2 =120.2.

[0057] Step S302, configuring the quantum circuit according to the parameters, and running the quantum circuit to obtain an output quantum state; Among them, the output quantum state consists of a series of quantum states and their corresponding probabilities.

[0058] Specifically, the quantum circuit is configured according to the parameters after quantum encoding, and the output quantum state is obtained by running the quantum circuit. The output quantum state is usually represented by the state of one or more quantum bit sets and their corresponding probability distribution.

[0059] For example, suppose a quantum circuit has two qubits, and the output of this quantum circuit is a mixed state, which contains two possible pure states; Pure state A: The first quantum bit is 0, the second quantum bit is 0, that is, state |00 ; Pure state B: The first quantum bit is 1, the second quantum bit is 1, that is, state |11 ; The density matrix ρ of the mixed state can be represented by a linear combination of two projection operators, each of which corresponds to a pure state; The matrix corresponding to |00><00| is: ;|11><11|The corresponding matrix is: ; The mixed state ρ is composed of a linear combination of these two projection operators, where and is a probability that satisfies + =1; that is, the mixed state can be expressed as:

[0060] By running the quantum circuit for measurement, due to the randomness of the quantum state, each measurement may get a different result, so it is necessary to perform enough shots (for example, 1000 times) to count the number of times pure state A and pure state B appear respectively, and calculate their respective probabilities. and , assuming pure state A (|00 ) appears 550 times, pure state B (|11 ) appeared 450 times, which means that the probability of each pure state and its corresponding state in this mixed state can be estimated as: =0.55, =0.45.

[0061] Step S303, calculating expected values ​​of various edge types in the molecular structure graph according to the output quantum state.

[0062] Specifically, the expected value of each edge type is calculated based on the output of the quantum circuit and used as the first edge feature of the molecular structure graph.

[0063] In summary, by encoding the original edge features of the molecular structure graph into the parameters of the quantum logic gate and configuring and running the quantum circuit accordingly, the expected values ​​of various edge types in the molecular structure graph can be efficiently calculated, thus providing a new quantum computing-based molecular structure analysis method, which helps to more accurately understand the interactions and properties between molecules, and thus improve the accuracy and reliability of the prediction of drug molecular properties.

[0064] See also Figure 4 , Figure 4 A flow chart of an edge feature clustering method provided in an embodiment of the present invention, the method comprising the following steps: Step S401, clustering the plurality of first edge features by a clustering algorithm to obtain a clustering result; The clustering result refers to the second edge feature after the first edge feature is subjected to denoising.

[0065] Specifically, the first edge feature is subjected to noise reduction processing by using a K-means clustering algorithm to obtain the second edge feature.

[0066] Exemplarily, in this process, the K-means algorithm is used to effectively cluster similar edge features together. Assume that there are three types of bonds in the drug molecule structure diagram: single bond, double bond and triple bond. In an ideal noise-free environment, after these bond types are processed by variational quantum circuit transformation, their numerical features are clearly mapped to different intervals. Assume that the value of a single bond varies from 1 to 10, the value of a double bond varies from 2 to 20, and the value of a triple bond varies from 3 to 30; however, in actual applications, the existence of quantum intrinsic noise requires running quantum circuits in a noisy environment. Assume that the quantum circuit is run 10 times for each bond type. For a single bond, 5 different results are obtained. This Some results are centered around the standard value of 10, showing slight fluctuations, namely 10, 10.1, 9.8, 10.2... etc. (the standard value 10 may appear multiple times). The double bond is centered around the standard value 20, 20, 20.2, 19.8, 20... etc. (the standard value 20 may appear multiple times). The triple bond is centered around the standard value 30, 30, 30.2, 29.8, 30... etc. (the standard value 30 may appear multiple times). At this time, the interval disturbance range with the standard value of 10 is processed as the first cluster result, the interval disturbance range with the standard value of 20 is processed as the second cluster result, and the interval disturbance range with the standard value of 30 is processed as the third cluster result.

[0067] Step S402: classify the plurality of first edge features into predefined edge categories according to the clustering result, and generate an index for each edge category to obtain a second edge feature.

[0068] The predefined edge category refers to how many types of chemical bonds (edges) for a drug molecule are known in advance before clustering; the index can be a unique identifier (such as an integer, a string, etc.).

[0069] Specifically, according to the clustering result, similar edge features are classified into predefined edge categories, and an index is generated for each edge category for fast retrieval and identification of the edge category. The edge category with the index is called the second edge feature.

[0070] Exemplarily, from the above example, it can be seen that the first clustering result of the interval perturbation range with a standard value of 10 is divided into a single-key category, the second clustering result of the interval perturbation range with a standard value of 20 is divided into a double-key category, and the third clustering result of the interval perturbation range with a standard value of 30 is divided into a triple-key category; and an index 1 identifier is generated for the single-key category, an index 2 identifier is generated for the double-key category, and an index 3 identifier is generated for the triple-key category for quick retrieval and identification of the edge category.

[0071] In summary, clustering the first edge features helps to separate noise and redundant information from the main edge features to obtain the second edge features. The second edge features better retain the essential characteristics of the original data and enhance the generalization ability of the model. Each edge category corresponds to an index, so that subsequent data processing can quickly locate related edge features based on these indexes, greatly improving the efficiency of data processing.

[0072] In one embodiment of the present application, before predicting the chemical properties of the drug molecule according to the graph features of the new molecular structure graph, the method further includes: The graph feature is determined according to a graph convolution calculation formula, where the graph convolution calculation formula is:

[0073] Among them, 𝐴 is the adjacency matrix determined based on the second edge feature, 𝐷 is the degree matrix, is the node feature of the l-th layer graph convolutional network, is the weight of the l-th layer graph convolutional network, and 𝜎 is the activation function.

[0074] Specifically, the second edge features obtained through denoising processing are used to construct a new molecular structure graph based on the second edge features and the original node features in the molecular structure graph. The comprehensive graph representation of the constructed new molecular structure is input into the classical convolutional neural network to obtain graph features. The obtained graph features are converted into label probabilities through activation functions, and the prediction results are obtained through the label probabilities.

[0075] See also Figure 5 , Figure 5 A drug molecule property prediction device provided in an embodiment of the present invention may include: A calculation unit 501 is used to calculate the expected values ​​of various edge types in a molecular structure graph through quantum circuits, and use the expected values ​​as first edge features of the molecular structure graph, where the edges in the molecular structure graph are used to represent chemical bonds of drug molecules; Among them, edge refers to the connection relationship between atoms in the molecular structure diagram, and may also contain information about the properties of chemical bonds. Edge type refers to the type of connection relationship. The first edge feature refers to the use of quantum computing technology to simulate the chemical bond type in drug molecules through quantum circuits and calculate the expected value of the chemical bond type.

[0076] Specifically, in the aspect of calculating the expected values ​​of various edge types in the molecular structure graph by using quantum circuits, the calculation unit 501 specifically includes: Acquiring original edge features of a molecular structure graph, and encoding the original edge features into parameters of a quantum logic gate; Configuring the quantum circuit according to the parameters, and running the quantum circuit to obtain an output quantum state; The expected values ​​of various edge types in the molecular structure graph are calculated according to the output quantum state.

[0077] Specifically, in the aspect of encoding the original edge feature into a parameter of a quantum logic gate, the calculation unit specifically includes: The original edge features are encoded into parameters of a quantum logic gate according to a mapping function, wherein the mapping function is:

[0078] Among them, the is the encoding weight, is the coding bias, the is the jth edge feature of the i-th molecular structure graph.

[0079] A denoising unit 502, configured to denoise the first edge feature by using a clustering algorithm to obtain a second edge feature; A reconstruction unit 503, configured to construct a new molecular structure graph according to the second edge feature and the original node feature in the molecular structure graph; The prediction unit 504 is used to predict the chemical properties of the drug molecules according to the graph features of the new molecular structure graph.

[0080] See also Figure 6 , the present specification also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor implements a drug molecule property prediction method in any of the above embodiments when executing the computer program. Figure 6 The electronic device may be a computer device, the computer device may be a classical computer, or the computer device may be a quantum computer.

[0081] The present specification also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a computer, the computer executes a method for predicting the properties of drug molecules in any of the above-mentioned embodiments.

[0082] It should be understood that the specific examples in this specification are only intended to help those skilled in the art to better understand the implementation methods of this specification, rather than to limit the scope of the present invention.

[0083] It can be understood that in the various implementations of this specification, the size of the serial number of each process does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the implementation methods of this specification.

[0084] It can be understood that the various embodiments described in this specification can be implemented individually or in combination, and the embodiments of this specification are not limited to this.

[0085] Unless otherwise specified, all technical and scientific terms used in the embodiments of this specification have the same meaning as those generally understood by those skilled in the art of the technical field of this specification. The terms used in this specification are only for the purpose of describing specific embodiments and are not intended to limit the scope of this specification. The term "and / or" used in this specification includes any and all combinations of one or more related listed items. The singular forms of "a", "above", and "the" used in the embodiments of this specification and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings.

[0086] It can be understood that the processor of the embodiment of this specification can be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method implementation can be completed by the hardware integrated logic circuit or software instructions in the processor. The above processor can be a general processor, a digital signal processor (Digital Signal Processor, DSP), an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), a field programmable gate array (Field Programmable Gate Array, FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiment of this specification can be implemented or executed. The general processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in the embodiment of this specification can be directly embodied as a hardware decoding processor to perform, or the hardware and software modules in the decoding processor are combined and executed. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware.

[0087] It is understood that the memory in the embodiments of this specification may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memory. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (programmable ROM, PROM), an erasable programmable read-only memory (erasablePROM, EPROM), an electrically erasable programmable read-only memory (EEPROM) or a flash memory. The volatile memory may be a random access memory (RAM). It should be noted that the memory of the systems and methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0088] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this specification.

[0089] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method implementation methods and will not be repeated here.

[0090] In the several embodiments provided in this specification, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device implementation described above is only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0091] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0092] In addition, each functional unit in each embodiment of the present specification may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0093] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this specification, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of this specification. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc., various media that can store program codes.

[0094] The above is only a specific implementation of this specification, but the protection scope of the present invention is not limited thereto. Any person skilled in the art who is familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in this specification, which should be included in the protection scope of this specification. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for predicting the properties of drug molecules, characterized in that: The method comprises: Calculating expected values ​​of various edge types in a molecular structure graph through quantum circuits, and using the expected values ​​as first edge features of the molecular structure graph, wherein the edges in the molecular structure graph are used to represent chemical bonds of drug molecules; De-noising the first edge feature by using a clustering algorithm to obtain a second edge feature; Constructing a new molecular structure graph according to the second edge feature and the original node feature in the molecular structure graph; The chemical properties of the drug molecule are predicted based on the graph features of the new molecular structure graph.

2. The method according to claim 1, characterized in that The expected values ​​of various edge types in the molecular structure graph are calculated by quantum circuits, including: Acquiring original edge features of a molecular structure graph, and encoding the original edge features into parameters of a quantum logic gate; Configuring the quantum circuit according to the parameters, and running the quantum circuit to obtain an output quantum state; The expected values ​​of various edge types in the molecular structure graph are calculated according to the output quantum state.

3. The method according to claim 2, characterized in that The encoding of the original edge features into parameters of a quantum logic gate comprises: The original edge features are encoded into parameters of a quantum logic gate according to a mapping function, wherein the mapping function is: i ij =W*e ij +b Wherein, W is the encoding weight, b is the encoding bias, and e is ij is the jth edge feature of the i-th molecular structure graph.

4. The method according to claim 1, characterized in that: The number of the first edge features is multiple, and the denoising of the first edge features by using a clustering algorithm to obtain the second edge features includes: Clustering the plurality of first edge features by using a clustering algorithm to obtain a clustering result; According to the clustering result, the plurality of first edge features are classified into predefined edge categories, and an index is generated for each edge category to obtain a second edge feature.

5. The method according to claim 1, characterized in that Before predicting the chemical properties of the drug molecule according to the graph features of the new molecular structure graph, the method further comprises: The graph feature is determined according to a graph convolution calculation formula, where the graph convolution calculation formula is: Where A is the adjacency matrix determined based on the second edge feature, D is the degree matrix, and H l is the node feature of the l-th layer graph convolutional network, W l is the weight of the l-th layer graph convolutional network, and σ is the activation function.

6. A drug molecular property prediction device, characterized in that: The device comprises: a calculation unit, configured to calculate expected values ​​of various edge types in a molecular structure graph through quantum circuits, and to use the expected values ​​as first edge features of the molecular structure graph, wherein the edges in the molecular structure graph are used to represent chemical bonds of drug molecules; A denoising unit, configured to denoise the first edge feature by using a clustering algorithm to obtain a second edge feature; A reconstruction unit, used for constructing a new molecular structure graph according to the second edge feature and the original node feature in the molecular structure graph; The prediction unit is used to predict the chemical properties of the drug molecule according to the graph features of the new molecular structure graph.

7. The device according to claim 6, characterized in that In the aspect of calculating the expected values ​​of various edge types in the molecular structure graph by using quantum circuits, the calculation unit specifically includes: Acquiring original edge features of a molecular structure graph, and encoding the original edge features into parameters of a quantum logic gate; Configuring the quantum circuit according to the parameters, and running the quantum circuit to obtain an output quantum state; The expected values ​​of various edge types in the molecular structure graph are calculated according to the output quantum state.

8. The device according to claim 7, characterized in that In terms of encoding the original edge features into parameters of a quantum logic gate, the calculation unit specifically includes: The original edge features are encoded into parameters of a quantum logic gate according to a mapping function, wherein the mapping function is: i ij =W*e ij +b Wherein, W is the encoding weight, b is the encoding bias, and e is ij is the jth edge feature of the i-th molecular structure graph.

9. An electronic device, characterized in that: include: Processor and memory; The processor is connected to a memory, wherein the memory is used to store a computer program, and the processor is used to call the computer program to execute the method according to claims 1-5.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by a processor, the method according to claims 1 to 5 is executed.

Images