Intelligent computing power scheduling system and method for mRNA (messenger ribonucleic acid) coding protein structure prediction

Abstract

The invention discloses an intelligent computing power scheduling system and method for mRNA coding protein structure prediction and electronic equipment, and the system comprises an input and task management layer which is used for obtaining and translating an mRNA sequence to obtain atomic subtasks; the intelligent sensing and scheduling core layer is connected with the input and task management layer and is used for generating a corresponding optimal scheduling decision according to the atomic subtasks; the heterogeneous computing power resource layer is connected with the intelligent sensing and scheduling core layer and is used for carrying out classical calculation and / or quantum calculation according to the optimized scheduling decision to obtain a corresponding classical calculation result and / or quantum calculation result; and the output and optimization layer is connected with the heterogeneous computing power resource layer and is used for generating a protein 3D structure model according to the classical computing result and / or the quantum computing result. According to the system, the accuracy and efficiency of protein structure prediction results can be improved, and the computing power resource utilization rate can also be improved.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to an intelligent computing power scheduling system for predicting the structure of mRNA-encoded proteins, an intelligent computing power scheduling method for predicting the structure of mRNA-encoded proteins, and an electronic device. Background Technology

[0002] With the deep application of artificial intelligence technology in the field of bioinformatics, deep learning-driven protein structure prediction methods, represented by AlphaFold2, have achieved a breakthrough in prediction accuracy, becoming the mainstream technical solution for mRNA protein structure prediction research in both industry and academia. This type of solution mainly relies on classic computing nodes such as CPUs (Central Processing Units) and GPUs (Graphics Processing Units) in High Performance Computing (HPC) clusters to complete core operations such as large-scale deep learning model inference and biological sequence database retrieval. Task scheduling typically employs static or simple rules (such as queue priority). The entire prediction process is treated as a whole or a few large tasks, uniformly allocating similar computing resources. However, this approach has the following shortcomings: 1. The task of predicting the three-dimensional structure of a protein from an mRNA sequence typically includes different core subtasks such as small molecule folding and long protein chain structure optimization. The computing power requirements of each subtask are significantly and dynamically different. However, the static scheduling mode of traditional HPC clusters cannot dynamically allocate computing power according to the real-time computing power requirements of subtasks. This can easily lead to uneven computing power distribution, such as redundant computing power for some subtasks and insufficient computing power for others. At the same time, it may also cause long queue waiting times for a large number of tasks, resulting in a serious waste of high-performance computing resources and a significant reduction in overall resource utilization.

[0003] 2. Existing systems cannot dynamically select appropriate computing resources based on task characteristics, resulting in limited overall efficiency. Regardless of whether the task is data-intensive multi-sequence alignment, computationally intensive neural network inference, or sidechain packaging belonging to combinatorial optimization problems, they tend to allocate the most powerful GPU computing power, which does not match the optimal computing paradigm and is neither economical nor efficient.

[0004] 3. Lack of task granularity identification and intelligent scheduling mechanism for mRNA protein prediction tasks, making it impossible to fully understand and allocate suitable neural network inference engines or quantum annealing nodes at the computing power scheduling layer. Summary of the Invention

[0005] The present invention is proposed to address at least one of the aforementioned problems. According to a first aspect of the present invention, an intelligent computing power scheduling system for predicting the structure of mRNA-encoded proteins is provided, comprising: an input and task management layer, an intelligent sensing and scheduling core layer, a heterogeneous computing power resource layer, and an output and optimization layer.

[0006] The input and task management layer is used to acquire and translate mRNA sequences to obtain atomic subtasks.

[0007] The intelligent sensing and scheduling core layer is connected to the input and task management layer, and is used to generate corresponding optimized scheduling decisions based on the atomic subtasks.

[0008] The heterogeneous computing power resource layer and the intelligent sensing and scheduling core layer are used to perform classical and / or quantum computing based on the optimized scheduling decision, and obtain the corresponding classical and / or quantum computing results.

[0009] The output and optimization layer is connected to the heterogeneous computing resource layer and is used to generate a protein 3D structure model based on the classical computing results and / or the quantum computing results.

[0010] In one embodiment of the present invention, the input and task management layer includes: a sequence input module and a task management and decomposition module.

[0011] The sequence input module is used to obtain the mRNA sequence input by the user.

[0012] The task management and decomposition module is connected to the sequence input module and is used to translate the mRNA sequence into an amino acid sequence and generate multiple atomic subtasks based on the amino acid sequence.

[0013] The atomic subtasks include: MSA generation, template search, backbone structure generation, sidechain conformation optimization, model confidence evaluation and refinement.

[0014] In one embodiment of the present invention, the intelligent sensing and scheduling core layer includes: a computational feature sensing module and a quantum-classical intelligent scheduler.

[0015] The computational feature perception module is connected to the input and task management layer and is used to extract the computational feature vectors of each of the atomic subtasks.

[0016] The quantum-classical intelligent scheduler is connected to the computational feature sensing module and is used to generate optimized scheduling decisions corresponding to each of the atomic subtasks based on the computational feature vector.

[0017] The computational feature vector includes at least three of the following: task type identifier, problem size, computational complexity category, data dependency, and preference label for computational paradigm.

[0018] In one embodiment of the present invention, the quantum-classical intelligent scheduler includes: a policy knowledge base, a resource status monitoring submodule, a decision engine, and a quantum adapter.

[0019] The strategy knowledge base is used to store predefined mapping rules between task characteristics and computing power schemes, as well as utility data of historical scheduling decisions.

[0020] The resource status monitoring submodule is used to monitor classical computing load and quantum device availability and queue information.

[0021] The decision engine is connected to the computational feature perception module, the strategy knowledge base, and the resource status monitoring submodule. It is used to query the strategy knowledge base based on the atomic subtasks and their computational feature vectors, and combine the classical computing load and the availability and queue information of the quantum device to make decisions through a multi-objective optimization function to obtain the optimized scheduling decision corresponding to each atomic subtask.

[0022] The quantum adapter is connected to the decision engine and is used to transform the quantum optimization problem in the optimization scheduling decision into a computational model adapted to the target quantum computing hardware when the optimization scheduling decision is a hybrid scheduling strategy of quantum and classical computing power.

[0023] The utility data includes execution time, operating cost, and accuracy gains; the optimized scheduling decision includes a pure classical computing power scheduling strategy and a hybrid quantum and classical computing power scheduling strategy.

[0024] In one embodiment of the present invention, the quantum adapter, when transforming the quantum optimization problem into a computational model adapted to the target quantum computing hardware, specifically includes: The decision variables in the quantum optimization problem are mapped to binary variables.

[0025] Based on the binary variables, construct a Hamiltonian containing the objective function and constraints.

[0026] The Hamiltonian is formatted as a standard QUBO matrix or a quantum circuit parameter optimization problem.

[0027] In one embodiment of the present invention, the heterogeneous computing power resource layer includes: a classical computing resource pool, a classical computing power proxy module, a quantum computing resource pool, and a quantum computing power proxy module.

[0028] The classic computing resource pool is used to provide classic computing resources.

[0029] The classic computing power proxy module is connected to the decision engine and the classic computing resource pool, and is used to schedule and monitor the classic computing resource pool according to the optimization scheduling decision, complete job submission, monitoring and result collection, and obtain classic computing results.

[0030] The quantum computing resource pool is used to provide quantum computing resources.

[0031] The quantum computing power proxy module is connected to the quantum adapter and the quantum computing resource pool, and is used to schedule and monitor the quantum computing resource pool according to the computing model, complete quantum computing, and obtain quantum computing results.

[0032] In one embodiment of the present invention, the classical computing resource pool includes: a CPU cluster, a GPU cluster, and a TPU cluster; the quantum computing resource pool includes: a quantum annealing machine and a noisy quantum computer.

[0033] The quantum annealing machine is used to handle combinatorial optimization problems, and the noisy quantum computer is used to handle variable quantum algorithms.

[0034] In one embodiment of the present invention, the output and optimization layer includes: a result fusion and feedback module and a result output module.

[0035] The result fusion and feedback module is connected to the heterogeneous computing resource layer and is used to integrate the classical computing results and the quantum computing results to obtain the protein 3D structure model.

[0036] The result output module is connected to the result fusion and feedback module and the strategy knowledge base. It is used to output the protein 3D structure model, record the time and cost of this scheduling and the confidence improvement value of the protein 3D structure model, and send it to the strategy knowledge base through the feedback learning path to update the strategy knowledge base.

[0037] According to a second aspect of the present invention, a smart computing power scheduling method for predicting the structure of mRNA-encoded proteins is provided, comprising: Obtain and translate the mRNA sequence to obtain the atomicity subtask.

[0038] Based on the atomic subtasks, corresponding optimized scheduling decisions are generated.

[0039] Based on the optimized scheduling decision, classical and / or quantum computations are performed to obtain the corresponding classical and / or quantum computation results.

[0040] Based on the classical calculation results and / or the quantum calculation results, a 3D structural model of the protein is generated.

[0041] According to a third aspect of the present invention, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory, wherein when the computer program is executed by the processor, it implements the above-described intelligent computing power scheduling method for predicting the structure of mRNA-encoded proteins.

[0042] The intelligent computing power scheduling system and method for predicting the structure of mRNA-encoded proteins of the present invention have the following beneficial effects: 1. The system of the present invention, through the cooperation between various working layers, can adapt to the task characteristics of predicting the three-dimensional structure of proteins from mRNA sequences. It constructs an intelligent scheduling system that deeply integrates quantum and classical heterogeneous computing power, and has the technical advantages of improved prediction accuracy, optimized computing efficiency, industrial application of quantum computing power and strong adaptability. It provides a full-stack solution for the integrated application of quantum and classical heterogeneous computing power in the field of bioinformatics.

[0043] 2. This invention enhances the processing of key optimization problems through quantum computing, enabling the discovery of lower-energy protein conformations. This effectively improves the accuracy of structural prediction for difficult-to-predict regions such as disordered regions and novel folds, expanding the application boundaries of protein structure prediction technology. Simultaneously, the hybrid quantum and classical computing model compensates for the shortcomings of classical computing in conformation sampling efficiency, improving the overall accuracy and robustness of protein structure prediction results.

[0044] 3. This invention breaks down the protein prediction process into multiple atomic subtasks through the input and task management layer, and then uses the intelligent perception and scheduling core layer to accurately match the computing power of different types of atomic subtasks. This avoids the monopolization of high-performance GPUs throughout the process, accurately assigns tasks to the most suitable computing units, reduces the overall computing time and cost, and maximizes the utilization of global resources.

[0045] 4. The heterogeneous computing resource layer, based on optimized scheduling decisions, schedules and computes both quantum and classical computing power. This achieves interface encapsulation and unified scheduling between quantum and classical computing power, breaking the current situation of isolated applications of quantum computing in protein structure prediction. Simultaneously, it promotes the practical application of quantum computing in the biomedical field. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1This is a schematic diagram of the intelligent computing power scheduling system for predicting the structure of mRNA-encoded proteins according to an embodiment of the present invention. Figure 2 A schematic diagram of the structure of an intelligent computing power scheduling system for predicting the structure of mRNA-encoded proteins provided in another embodiment of the present invention; Figure 3 This is a schematic diagram of the working process of a quantum adapter according to an embodiment of the present invention; Figure 4 This is a flowchart illustrating an intelligent computing power scheduling method for predicting the structure of mRNA-encoded proteins according to an embodiment of the present invention. Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of the present invention more apparent, exemplary embodiments according to the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are merely a part of the embodiments of the present invention, and not all of the embodiments of the present invention. It should be understood that the present invention is not limited to the exemplary embodiments described herein. Based on the embodiments of the present invention described herein, all other embodiments obtained by those skilled in the art without inventive effort should fall within the protection scope of the present invention.

[0049] In the following description, numerous specific details are set forth in order to provide a more thorough understanding of the invention. However, it will be apparent to those skilled in the art that the invention can be practiced without one or more of these details. In other instances, certain technical features well-known in the art have not been described in order to avoid obscuring the invention.

[0050] It should be understood that the invention can be embodied in various forms and should not be construed as being limited to the embodiments set forth herein. Rather, providing these embodiments will make the disclosure thorough and complete, and will fully convey the scope of the invention to those skilled in the art.

[0051] To fully understand this invention, a detailed structure will be presented in the following description to illustrate the technical solution proposed by this invention. Optional embodiments of the invention are described in detail below; however, in addition to these detailed descriptions, the invention may have other embodiments.

[0052] The following detailed description of some embodiments of the present invention is provided in conjunction with the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0053] To improve the accuracy, efficiency, and computational resource utilization of protein structure prediction results, the first aspect of this invention provides an intelligent computational scheduling system for mRNA-encoding protein structure prediction, such as... Figure 1 As shown, the intelligent computing power scheduling system for mRNA-encoded protein structure prediction includes: an input and task management layer 10, an intelligent sensing and scheduling core layer 20, a heterogeneous computing power resource layer 30, and an output and optimization layer 40.

[0054] The input and task management layer 10 is used to acquire and translate mRNA sequences to obtain atomic subtasks.

[0055] As an example, the input and task management layer 10 receives the mRNA sequence input by the user, translates it into an amino acid sequence, and initiates and decomposes the standard protein structure prediction process to generate a series of atomic subtasks. The prediction process can be an AlphaFold2 process.

[0056] The intelligent sensing and scheduling core layer 20 is connected to the input and task management layer 10, and is used to generate corresponding optimized scheduling decisions based on atomic subtasks.

[0057] It should be noted that the optimized scheduling decision includes executing instructions using pure classical computing power and executing instructions using a hybrid of quantum and classical computing power.

[0058] The heterogeneous computing power resource layer 30 and the intelligent sensing and scheduling core layer 20 are used to perform classical and / or quantum computing based on the optimized scheduling decision, and obtain the corresponding classical and / or quantum computing results.

[0059] As an example, the heterogeneous computing power resource layer 30 includes classical computing power agents and quantum computing power agents. The classical computing power agent encapsulates the access interface of relevant computing power clusters on the local or cloud, thereby receiving classical task instructions and completing job submission, monitoring and result collection. The quantum computing power agent encapsulates the API calls of mainstream quantum cloud service platforms at home and abroad (such as D-Wave Leap, IBM Quantum Cloud), thereby submitting the problem model output by the intelligent sensing and scheduling core layer 20 to the corresponding device and retrieving the results of quantum computing (such as ground state samples and optimal parameters).

[0060] The output and optimization layer 40 is connected to the heterogeneous computing power resource layer 30 to generate protein 3D structure models based on classical computing results and / or quantum computing results.

[0061] As an example, the output and optimization layer 40 is also used to collect the actual performance metrics of this task chain and feed them back to the policy knowledge base in the intelligent perception and scheduling core layer 20 for adaptive learning.

[0062] The intelligent computing power scheduling system for mRNA-encoded protein structure prediction of the present invention has the following beneficial effects: 1. This invention, through the cooperation between various working layers, can adapt to the task characteristics of predicting the three-dimensional structure of proteins from mRNA sequences. It constructs an intelligent scheduling system that deeply integrates quantum and classical heterogeneous computing power, and has the technical advantages of improved prediction accuracy, optimized computing efficiency, industrial application of quantum computing power, and strong adaptability. It provides a full-stack solution for the integrated application of quantum and classical heterogeneous computing power in the field of bioinformatics.

[0063] 2. This invention enhances the processing of key optimization problems through quantum computing, enabling the discovery of lower-energy protein conformations. This effectively improves the accuracy of structural prediction for difficult-to-predict regions such as disordered regions and novel folds, expanding the application boundaries of protein structure prediction technology. Simultaneously, the hybrid quantum and classical computing model compensates for the shortcomings of classical computing in conformation sampling efficiency, improving the overall accuracy and robustness of protein structure prediction results.

[0064] 3. This invention breaks down the protein prediction process into multiple atomic subtasks through the input and task management layer 10, and then accurately matches the computing power of different types of atomic subtasks through the intelligent perception and scheduling core layer 20. This avoids the monopoly of high-performance GPUs throughout the process, accurately assigns tasks to the most suitable computing units, reduces the overall computing time and cost, and maximizes the utilization of global resources.

[0065] 4. The heterogeneous computing power resource layer 30, based on optimized scheduling decisions, schedules and computes both quantum and classical computing power. This achieves interface encapsulation and unified scheduling between quantum and classical computing power, breaking the current situation of isolated applications of quantum computing in protein structure prediction. Simultaneously, it promotes the practical application of quantum computing in the biomedical field.

[0066] In some embodiments, such as Figure 2 As shown, the input and task management layer 10 includes a sequence input module 101 and a task management and decomposition module 102.

[0067] The sequence input module 101 is used to acquire the mRNA sequence input by the user.

[0068] For example, the sequence input module 101 can perform preliminary verification on all acquired mRNA sequences, remove invalid sequences according to bioinformatics rules, and after the verification is passed, perform unified standardization processing on the valid mRNA sequences and convert them into a data format that can be recognized by the task management and decomposition module 102.

[0069] The task management and decomposition module 102 is connected to the sequence input module 101 and is used to translate the mRNA sequence into an amino acid sequence and generate multiple atomic subtasks based on the amino acid sequence.

[0070] The atomicity subtasks include: MSA (Multiple Sequence Alignment) generation, template search, backbone structure generation, sidechain conformation optimization, model confidence evaluation and refinement.

[0071] As an example, the task management and decomposition module 102 translates the mRNA sequence into an amino acid sequence and drives a standard protein structure prediction workflow (such as the AlphaFold2 workflow), which decomposes the entire workflow into multiple atomic subtasks.

[0072] In this embodiment, through the synergistic effect between the sequence input module 101 and the task management and decomposition module 102, protein structure prediction can be decomposed into standardized and decoupled atomic subtasks, providing a task basis for subsequent intelligent computing power matching and improving the standardization and efficiency of the overall task processing.

[0073] In some embodiments, such as Figure 2 As shown, the intelligent sensing and scheduling core layer 20 includes: a computational feature sensing module 201 and a quantum-classical intelligent scheduler 202.

[0074] The computational feature perception module 201 is connected to the input and task management layer 10 and is used to extract the computational feature vectors of each atomic subtask.

[0075] The computational feature vector includes at least three of the following: task type identifier, problem size, computational complexity category, data dependencies, and preference label for computational paradigm.

[0076] It should be noted that the problem size may include the number of residues to be optimized and the number of conformational candidates; the computational complexity category includes P problems (Polynomial-time Problems), NP-Hard problems (Non-deterministic Polynomial-time Hard Problems), etc.; and the preference labels for computational paradigms include data parallelism, graph optimization, quantum-friendly optimization, etc.

[0077] It should be noted that the computational feature perception module 201 extracts the core dimensions of the computational feature vector for each atomic subtask according to preset rules, and quantifies or symbolically represents each dimension, transforming unstructured computing power requirements into structured feature vectors.

[0078] The quantum-classical intelligent scheduler 202 is connected to the computational feature perception module 201 and is used to generate optimized scheduling decisions for each atomic subtask based on the computational feature vector.

[0079] As an example, the quantum-classical intelligent scheduler 202 can match computing tasks with classical computing power and quantum computing power based on computational complexity, whether there is a preference for quantum computing, and other factors.

[0080] In this embodiment, the intelligent perception and scheduling core layer 20 first quantifies and extracts the core computational feature vectors of atomic subtasks to accurately represent the computing power requirements, and then generates appropriate computing power optimization scheduling decisions for each task accordingly, thereby realizing the intelligence and precision of computing power scheduling and improving the utilization efficiency of heterogeneous computing power resources.

[0081] In some embodiments, such as Figure 2 As shown, the quantum-classical intelligent scheduler 202 includes: a policy knowledge base 2021, a resource status monitoring submodule 2022, a decision engine 2023, and a quantum adapter 2024.

[0082] The Strategy Knowledge Base 2021 is used to store predefined mapping rules between task characteristics and computing power schemes, as well as utility data of historical scheduling decisions.

[0083] The utility data includes execution time, operating costs, and accuracy gains.

[0084] It should be noted that execution time represents the total time taken for the corresponding task from the start of calculation to the output result under the specified computing power scheme; operating cost represents the comprehensive cost of hardware resource occupation, energy consumption, computing power leasing, etc., incurred by the corresponding task under the specified computing power scheme; and accuracy gain represents the difference between the accuracy of the calculation result of the corresponding task under the specified computing power scheme and the preset accuracy threshold.

[0085] The Resource Status Monitoring Submodule 2022 is used to monitor classical computing load and quantum device availability and queue information.

[0086] As an example, the resource status monitoring submodule 2022 can collect core operational metrics of classical computing resources within the system in real time, accurately reflecting the load and availability of classical computing power. Among these metrics, the availability of quantum devices includes the number of available qubits, the success rate of quantum state preparation and measurement, etc., while task queue information can include the number of pending quantum computing tasks, the average waiting time for each task, and the progress of the quantum circuits currently executing the task.

[0087] The decision engine 2023 is connected to the computational feature perception module 201, the strategy knowledge base 2021, and the resource status monitoring submodule 2022. It is used to query the strategy knowledge base 2021 based on the atomic subtasks and their computational feature vectors, and combine classical computing load and quantum device availability and queue information to make decisions through a multi-objective optimization function to obtain the optimized scheduling decision corresponding to each atomic subtask.

[0088] Among them, the optimized scheduling decision includes a pure classical computing power scheduling strategy and a hybrid scheduling strategy of quantum and classical computing power.

[0089] Specifically, the multi-objective optimization function can be expressed as: min F = α * execution time + β * cost - γ * accuracy gain, where α, β and γ represent weighting coefficients.

[0090] As an example, optimizing scheduling decisions involves the specific execution instructions corresponding to each atomic subtask.

[0091] It's worth noting that Decision Engine 2023 can make decisions entirely based on a comprehensive set of "if-then-else" rules, pre-defined by domain experts (computational biology + quantum computing). Decision Engine 2023 can also be a pre-trained machine learning model (such as a reinforcement learning agent or gradient boosting tree). This model takes task feature vectors and real-time resource states as input and directly outputs the optimal computing power allocation scheme. The model is trained and iterated using historical data.

[0092] The quantum adapter 2024 connects to the decision engine 2023 and is used to transform the quantum optimization problem in the optimization scheduling decision into a computational model adapted to the target quantum computing hardware when the optimization scheduling decision is a hybrid scheduling strategy of quantum and classical computing power.

[0093] It should be noted that the quantum adapter 2024 is only activated when the decision engine 2023 determines that a certain atomic subtask requires a hybrid solution of quantum and classical computing power.

[0094] Specifically, the target quantum computing hardware includes a quantum annealing machine and a noisy quantum computer.

[0095] For example, the core optimization problem can be the optimal combination problem of a side-chain rotation network containing 10 residues.

[0096] In this embodiment, by accurately assigning computational tasks based on atomic subtasks and their computational feature vectors, combined with the actual situation of the system, computational efficiency is optimized and overall computation time and cost are reduced.

[0097] In some embodiments, such as Figure 3As shown, the Quantum Adapter 2024, when transforming a quantum optimization problem into a computational model adapted to the target quantum computing hardware, specifically includes: A1 maps decision variables in quantum optimization problems to binary variables.

[0098] Specifically, the abstract, non-binary decision variables in the quantum optimization problem of protein structure prediction are transformed into binary variables (0 / 1) that can be recognized by quantum computing hardware.

[0099] A2, based on binary variables, construct a Hamiltonian containing the objective function and constraints.

[0100] A3 formats the Hamiltonian as a standard QUBO (Quadratic Unconstrained Binary Optimization) matrix or quantum circuit parameter optimization problem.

[0101] Among them, the quantum annealing machine and the noisy quantum computer are adapted to the QUBO matrix and quantum circuit parameter optimization formats, respectively.

[0102] In this embodiment, the abstract quantum optimization problem is transformed into a format specific to different types of quantum hardware, enabling the quantum optimization problem to run correctly and output valid results on different hardware.

[0103] In some embodiments, the heterogeneous computing power resource layer 30 includes: a classical computing resource pool 301, a classical computing power proxy module 302, a quantum computing resource pool 303, and a quantum computing power proxy module 304.

[0104] Classic computing resource pool 301 is used to provide classic computing resources.

[0105] Specifically, the classic computing resource pool 301 includes: CPU clusters, GPU clusters, and TPU (Tensor Processing Unit) clusters.

[0106] The classic computing power agent module 302 is connected to the decision engine 2023 and the classic computing resource pool 301. It is used to schedule and monitor the classic computing resource pool 301 according to the optimization scheduling decision, complete job submission, monitoring and result collection, and obtain classic computing results.

[0107] Quantum computing resource pool 303 is used to provide quantum computing resources.

[0108] Specifically, the quantum computing resource pool 303 includes a quantum annealing machine and a noisy quantum computer. The quantum annealing machine is used to handle combinatorial optimization problems, while the noisy quantum computer is used to handle variational quantum algorithms.

[0109] The quantum computing agent module 304 is connected to the quantum adapter 2024 and the quantum computing resource pool 303. It is used to schedule and monitor the quantum computing resource pool 303 according to the computing model, complete the quantum computing, and obtain the quantum computing results.

[0110] In this embodiment, the heterogeneous computing power resource layer 30 provides classical computing power of CPU / GPU / TPU clusters and quantum computing power of quantum annealing machines and noisy quantum computers, respectively, to accurately adapt to the needs of different types of computing tasks, so as to accurately complete relevant computing tasks according to the instructions output by the intelligent sensing and scheduling core layer 20.

[0111] In some embodiments, such as Figure 2 As shown, the output and optimization layer 40 includes a result fusion and feedback module 401 and a result output module 402.

[0112] The result fusion and feedback module 401 is connected to the heterogeneous computing power resource layer 30 to integrate classical computing results and quantum computing results to obtain a protein 3D structure model.

[0113] Specifically, the result fusion and feedback module 401 is connected to the classical computing resource pool 301 and the quantum computing resource pool 303.

[0114] The result output module 402 is connected to the result fusion and feedback module 401 and the strategy knowledge base 2021. It is used to output the protein 3D structure model, record the time and cost of this scheduling and the confidence improvement value of the protein 3D structure model, and send it to the strategy knowledge base 2021 through the feedback learning path to update the strategy knowledge base 2021.

[0115] In this embodiment, the system continuously feeds back the time and cost of scheduling, as well as the confidence improvement value of the protein 3D structure model, enabling it to adapt to the prediction needs of different proteins, the performance evolution of quantum hardware, and changes in cloud computing prices, thus maintaining optimal long-term performance.

[0116] In addition, this invention also provides an intelligent computing power scheduling method for predicting the structure of mRNA-encoded proteins, such as... Figure 4 As shown, intelligent computing power scheduling methods for predicting the structure of mRNA-encoded proteins include: S1, obtain and translate the mRNA sequence to obtain the atomic subtask.

[0117] S2 generates corresponding optimized scheduling decisions based on atomic subtasks.

[0118] S3, based on the optimized scheduling decision, performs classical and / or quantum computations to obtain the corresponding classical and / or quantum computation results.

[0119] S4 generates a 3D structural model of the protein based on classical and / or quantum computation results.

[0120] Other specific implementations of the intelligent computing power scheduling method for mRNA-encoded protein structure prediction in the embodiments of the present invention can be found in the specific implementations of the intelligent computing power scheduling system for mRNA-encoded protein structure prediction in the above embodiments of the present invention.

[0121] Next, using a specific mRNA sequence as an example, this invention will illustrate its specific workflow: B1. System startup, input mRNA sequence, task management and decomposition module 102 starts working, generating prediction workflow and initial subtask queue. The initial subtask queue includes multiple atomic subtasks.

[0122] B2, the computational feature perception module 201 takes an atomic subtask (e.g., "side chain conformation optimization - sub-network A") from the queue, analyzes it and generates a computational feature vector (e.g., {type: discrete optimization, size: 12 residues, complexity: NP-Hard, label: quantum friendly}).

[0123] B3, Decision Engine 2023 received this feature vector. After querying the strategy knowledge base 2021, it found that historical records showed that for this type of "medium-scale, quantum-friendly" discrete optimization task, using quantum annealing to process the core part could achieve a significant improvement in accuracy. Considering the current short queue of quantum annealers, Decision Engine 2023 decided to adopt a hybrid approach: further decompose the 12-residue network, assigning the 8-residue core subnetwork with the strongest interactions to quantum computing, and processing the remaining 4 residues by classical computing.

[0124] B4, Quantum Adapter 2024 is invoked. For these 8 residues, each residue has 3 common rotamers. The adapter defines 8 * 3 = 24 binary variables x. i,j Construct the target Hamiltonian H = Σ(single-point energy term) + Σ(two-point interaction term) + k*(constraint term for selecting exactly one Rotamer for each residue), and output the corresponding 24x24 QUBO matrix, where k represents the preset parameter.

[0125] B5, the quantum computing agent module 304 submits the QUBO matrix to the cloud quantum annealing machine. Simultaneously, the classical computing agent module 302 submits the remaining 4-residue optimization task to the CPU cluster.

[0126] B6, two types of computations are executed in parallel. The quantum annealing machine returns a set of lowest-energy solution samples (i.e., x) in milliseconds. i,j (Assignment).

[0127] B7, the result fusion and feedback module 401 will decode the quantum code to obtain the optimal Rotamer ID of the 8 core residues, and combine it with the classical calculation result of 4 residues to generate the complete side chain conformation.

[0128] B8 records the time, cost, and confidence boost value of the protein 3D structure model used in this scheduling, and updates the strategy knowledge base in 2021.

[0129] B9. Repeat steps B2-B8 until all atomic subtasks in the workflow are completed, and output the final protein structure.

[0130] In addition, the present invention also provides an electronic device, such as Figure 5 As shown, the controller 500 includes a processor 501 and a memory 503. The processor 501 and the memory 503 are connected, for example, via a bus 502.

[0131] Optionally, the controller 500 may also include a transceiver 504. It should be noted that in practical applications, the transceiver 504 is not limited to one, and the structure of the controller 500 does not constitute a limitation on the embodiments of the present invention.

[0132] Processor 501 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this invention. Processor 501 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0133] Bus 502 may include a pathway for transmitting information between the aforementioned components. Bus 502 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. Bus 502 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 5 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0134] The memory 503 stores a computer program corresponding to the intelligent computing power scheduling method for predicting the structure of mRNA-encoded proteins according to the above embodiments of the present invention. This computer program is executed under the control of the processor 501. The processor 501 executes the computer program stored in the memory 503 to implement the content shown in the aforementioned method embodiments.

[0135] The controller 500 includes, but is not limited to, mobile terminals such as laptops, PDAs (personal digital assistants), and PADs (tablet computers), as well as fixed terminals such as desktop computers. Figure 5 The controller 500 shown is merely an example and should not be construed as limiting the functionality and scope of use of embodiments of the present invention.

[0136] Although exemplary embodiments have been described herein with reference to the accompanying drawings, it should be understood that the above exemplary embodiments are merely illustrative and are not intended to limit the scope of this application. Various changes and modifications can be made therein by those skilled in the art without departing from the scope and spirit of this application. All such changes and modifications are intended to be included within the scope of this application as claimed in the appended claims.

[0137] Those skilled in the art will recognize that the units and algorithm steps of the various examples 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 implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0138] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed.

[0139] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of this application may be practiced without these specific details. In some instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0140] Similarly, it should be understood that, in order to streamline this application and aid in understanding one or more of the various inventive aspects, features of this application may sometimes be grouped together in a single embodiment, figure, or description thereof in the description of exemplary embodiments of this application. However, this approach should not be construed as reflecting an intention that the claimed application requires more features than are expressly recited in each claim. Rather, as reflected in the corresponding claims, its inventive point lies in solving the corresponding technical problem with features fewer than all features of a single disclosed embodiment. Therefore, the claims following the detailed description are hereby expressly incorporated into that detailed description, wherein each claim itself is a separate embodiment of this application.

[0141] Those skilled in the art will understand that, apart from the mutual exclusion of features, all features disclosed in this specification (including the accompanying claims, abstract, and drawings) and all processes or elements of any method or apparatus so disclosed may be combined in any combination. Unless otherwise expressly stated, each feature disclosed in this specification (including the accompanying claims, abstract, and drawings) may be replaced by an alternative feature that serves the same, equivalent, or similar purpose.

[0142] Furthermore, those skilled in the art will understand that although some embodiments described herein include certain features but not others included in other embodiments, combinations of features from different embodiments are intended to be within the scope of this application and form different embodiments. For example, in the claims, any one of the claimed embodiments can be used in any combination.

[0143] The various component embodiments of this application can be implemented in hardware, or as software modules running on one or more processors, or a combination thereof. Those skilled in the art will understand that microprocessors or digital signal processors (DSPs) can be used in practice to implement some or all of the functions of some modules according to the embodiments of this application. This application can also be implemented as an apparatus program (e.g., a computer program and computer program product) for performing part or all of the methods described herein. Such an implementation of this application can be stored on a computer-readable medium, or can be in the form of one or more signals. Such signals can be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.

[0144] It should be noted that the above embodiments are illustrative of this application and not restrictive, and that those skilled in the art can devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses should not be construed as limiting the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. This application can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names.

[0145] The above description is merely a specific embodiment or illustration of the embodiments of this application. The scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. The scope of protection of this application shall be determined by the scope of the claims.

Claims

1. An intelligent computing power scheduling system for predicting the structure of mRNA-encoded proteins, characterized in that, The system includes: The input and task management layer is used to acquire and translate mRNA sequences to obtain atomic subtasks. The intelligent sensing and scheduling core layer is connected to the input and task management layer and is used to generate corresponding optimized scheduling decisions based on the atomic subtasks. The heterogeneous computing power resource layer, together with the intelligent sensing and scheduling core layer, is used to perform classical and / or quantum computing based on the optimized scheduling decision, and obtain the corresponding classical and / or quantum computing results. The output and optimization layer, connected to the heterogeneous computing resource layer, is used to generate a 3D protein structure model based on the classical computing results and / or the quantum computing results.

2. The intelligent computing power scheduling system for predicting the structure of mRNA-encoded proteins according to claim 1, characterized in that, The input and task management layer includes: A sequence input module is used to acquire the mRNA sequence input by the user; The task management and decomposition module, connected to the sequence input module, is used to translate the mRNA sequence into an amino acid sequence and generate multiple atomic subtasks based on the amino acid sequence. The atomic subtasks include: MSA generation, template search, backbone structure generation, sidechain conformation optimization, model confidence evaluation and refinement.

3. The intelligent computing power scheduling system for predicting the structure of mRNA-encoded proteins according to claim 1, characterized in that, The core layer of intelligent sensing and scheduling includes: The computational feature perception module, connected to the input and task management layer, is used to extract the computational feature vectors of each of the atomic subtasks. A quantum-classical intelligent scheduler, connected to the computational feature sensing module, is used to generate optimized scheduling decisions for each of the atomic subtasks based on the computational feature vector. The computational feature vector includes at least three of the following: task type identifier, problem size, computational complexity category, data dependency, and preference label for computational paradigm.

4. The intelligent computing power scheduling system for predicting the structure of mRNA-encoded proteins according to claim 3, characterized in that, The quantum-classical intelligent scheduler includes: The strategy knowledge base is used to store predefined mapping rules between task characteristics and computing power schemes, as well as utility data of historical scheduling decisions; The resource status monitoring submodule is used to monitor classical computing load and quantum device availability and queue information; The decision engine, connected to the computational feature perception module, the strategy knowledge base, and the resource status monitoring submodule, is used to query the strategy knowledge base based on the atomic subtasks and their computational feature vectors, and combine the classical computing load and the availability and queue information of the quantum device to make decisions through a multi-objective optimization function to obtain the optimized scheduling decisions corresponding to each atomic subtask. A quantum adapter, connected to the decision engine, is used to transform the quantum optimization problem in the optimization scheduling decision into a computational model adapted to the target quantum computing hardware when the optimization scheduling decision is a hybrid scheduling strategy of quantum and classical computing power. The utility data includes execution time, operating cost, and accuracy gains; the optimized scheduling decision includes a pure classical computing power scheduling strategy and a hybrid quantum and classical computing power scheduling strategy.

5. The intelligent computing power scheduling system for predicting the structure of mRNA-encoded proteins according to claim 4, characterized in that, The quantum adapter, in transforming the quantum optimization problem into a computational model adapted to the target quantum computing hardware, specifically includes: Map the decision variables in the quantum optimization problem to binary variables; Based on the binary variables, construct a Hamiltonian containing the objective function and constraints; The Hamiltonian is formatted as a standard QUBO matrix or a quantum circuit parameter optimization problem.

6. The intelligent computing power scheduling system for predicting the structure of mRNA-encoded proteins according to claim 4, characterized in that, The heterogeneous computing power resource layer includes: The Classic Computing Resource Pool is used to provide classic computing resources; The classic computing power proxy module is connected to the decision engine and the classic computing resource pool. It is used to schedule and monitor the classic computing resource pool according to the optimization scheduling decision, complete job submission, monitoring and result collection, and obtain classic computing results. A quantum computing resource pool is used to provide quantum computing resources; The quantum computing power proxy module is connected to the quantum adapter and the quantum computing resource pool. It is used to schedule and monitor the quantum computing resource pool according to the computing model, complete quantum computing, and obtain quantum computing results.

7. The intelligent computing power scheduling system for predicting the structure of mRNA-encoded proteins according to claim 6, characterized in that, The classical computing resource pool includes: CPU clusters, GPU clusters, and TPU clusters; the quantum computing resource pool includes: quantum annealing machines and noisy quantum computers. The quantum annealing machine is used to handle combinatorial optimization problems, and the noisy quantum computer is used to handle variable quantum algorithms.

8. The intelligent computing power scheduling system for predicting the structure of mRNA-encoded proteins according to claim 4, characterized in that, The output and optimization layer includes: The result fusion and feedback module is connected to the heterogeneous computing power resource layer and is used to integrate the classical computing results and the quantum computing results to obtain the protein 3D structure model. The result output module, connected to the result fusion and feedback module and the strategy knowledge base, is used to output the protein 3D structure model, record the time and cost of this scheduling and the confidence improvement value of the protein 3D structure model, and send it to the strategy knowledge base through the feedback learning path to update the strategy knowledge base.

9. A smart computing power scheduling method for predicting the structure of mRNA-encoded proteins, characterized in that, The method includes: Obtain and translate the mRNA sequence to obtain the atomicity subtask; Based on the atomic subtasks, generate corresponding optimized scheduling decisions; Based on the optimized scheduling decision, classical and / or quantum computations are performed to obtain the corresponding classical and / or quantum computation results. Based on the classical calculation results and / or the quantum calculation results, a 3D structural model of the protein is generated.

10. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory, wherein when the computer program is executed by the processor, it implements the intelligent computing power scheduling method for predicting the structure of mRNA-encoded proteins as described in claim 9.

Images