Method and device for realizing mass spectrum data intelligent processing and molecular prediction workflow, processor and computer readable storage medium thereof

Abstract

The invention relates to a method for realizing mass spectrum data intelligent processing and molecular prediction workflow, which comprises the following steps of: carrying out standardization and noise reduction processing on mass spectrum data, and constructing molecular fingerprints of a standard mass spectrum set; inputting the molecular fingerprints into a mass spectrum core operator neural network for structure prediction; carrying out similarity calculation on an output result of the mass spectrum core operator neural network and a chemical vector library, and constructing a candidate molecular structure set; retrieving chemical information of the candidate molecules, and integrating the information with a prediction result; and multi-dimensional output is generated. According to the method, the device, the processor and the computer readable storage medium for realizing the mass spectrum data intelligent processing and molecular prediction workflow, background noise interference in the mass spectrum data is eliminated through a standardized noise reduction algorithm, a pure data basis is provided, a molecular structure intelligent prediction model based on a large model is established, and the method and the device are used for realizing the mass spectrum data intelligent processing and molecular prediction workflow. A standard workflow is constructed, a unified processing normal form is formed, and automation of molecular prediction based on mass spectrum data is realized.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence-driven cheminformatics, and more particularly to the field of mass spectrometry analysis. Specifically, it relates to a method, apparatus, processor, and computer-readable storage medium for realizing intelligent processing of mass spectrometry data and molecular prediction workflow. Background Technology

[0002] With the widespread application of mass spectrometry in chemical analysis, drug screening, metabolomics, and other fields, how to quickly and accurately infer potential molecular structures from mass spectrometry data has become an important direction in cheminformatics research. However, traditional structure inference methods mostly rely on rule bases, expert systems, or database searches, which cannot effectively handle unknown compounds; the interpretation of structural fragments depends on human experience, resulting in low inference efficiency; and the ability to identify complex or non-standard structures (such as derivatives and isomers) is limited.

[0003] In recent years, deep learning-based molecular characterization methods have shown great potential and broad prospects in the field of mass spectrometry prediction. However, the scarcity of labeled data hinders the full realization of these advantages. Furthermore, different types of molecular data vary significantly in their sources, formats, quality, and processing methods, making multidimensional information fusion challenging. In addition, the fragmented mass spectrometry prediction workflow requires the use of multiple independent tools for similarity calculations and information retrieval, failing to achieve fully automated integration and severely limiting the efficiency of high-throughput analysis. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method, apparatus, processor and computer-readable storage medium for realizing intelligent processing of mass spectrometry data and molecular prediction workflow that meets the requirements of automation, high analysis efficiency and wide applicability.

[0005] To achieve the above objectives, the present invention provides a method, apparatus, processor, and computer-readable storage medium for intelligent processing of mass spectrometry data and molecular prediction workflow, as well as the following: The method for realizing intelligent processing and molecular prediction workflow of mass spectrometry data is characterized by the following steps: (1) Standardize and denoise the mass spectrometry data, and construct the molecular fingerprint of the standard mass spectrometry set; (2) Input the molecular fingerprint into the mass spectrometry core operator neural network for structure prediction; (3) The output of the mass spectrometry core operator neural network is compared with the chemical vector library to calculate the similarity and select the top n most similar candidate molecules to construct a candidate molecule structure set; (4) Retrieve the chemical information of candidate molecules through the model context protocol service and integrate the information with the prediction results; (5) Input the integrated data into the large model base and generate multi-dimensional output.

[0006] Preferably, step (1) specifically includes the following steps: (1.1) Extract raw mass spectrometry data of the target compound from a multi-source mass spectrometry database; (1.2) Standardization processing; (1.3) Perform feature conversion to convert the standardized mass spectrum into a molecular fingerprint.

[0007] Preferably, step (1.2) specifically includes the following steps: (1.2.1) The mass spectrometry data is preprocessed using an adaptive noise reduction algorithm to perform baseline correction, smoothing and noise reduction, and peak detection and integration, providing a list of reliable peaks; (1.2.2) Standardize the original intensity in the list of reliable peaks and output the purified standard mass spectrometry data.

[0008] Preferably, step (2) specifically includes the following steps: (2.1) Construct a neural network for the core operator of mass spectrometry; (2.2) Input the constructed molecular fingerprint vector into the neural network for structure prediction, output the potential structure representation of the target molecule, and further embed the prediction result into the feature vector.

[0009] Preferably, step (3) specifically includes the following steps: (3.1) The results are passed to the computing node, and the chemical vector library is automatically called. The mass spectrometry prediction output vector is used as the retrieval vector to perform a similarity search in the high-dimensional space. (3.2) Calculate the similarity between vectors and select the top n structures that are most similar to the current prediction result as potential candidate molecules; (3.3) Pack the above n structures into a candidate molecular structure set.

[0010] Preferably, step (4) specifically includes the following steps: (4.1) The candidate molecule set is passed to the chemical information retrieval node, and the API is automatically called to connect to the open source chemical information database to perform reverse query and automatically obtain the basic attribute information of the corresponding molecules; (4.2) Integrate the structure prediction results with the queried chemical information to form an integrated information package.

[0011] Preferably, step (8) specifically includes the following steps: (5.1) Optimize prompt words by adjusting and optimizing them to generate multi-dimensional output; (5.2) Input the optimized prompt words along with the integrated information into the base model. Through joint encoding of structural content and task prompts, multi-level semantic modeling and knowledge reasoning are completed. (5.3) Generate multi-dimensional output results based on optimized input information.

[0012] The device for realizing intelligent processing of mass spectrometry data and molecular prediction workflow is characterized in that the device comprises: A processor is configured to execute computer-executable instructions; The memory stores one or more computer-executable instructions, which, when executed by the processor, implement the various steps of the method described above for implementing intelligent processing of mass spectrometry data and molecular prediction workflow.

[0013] The processor for realizing intelligent processing of mass spectrometry data and molecular prediction workflow is characterized in that the processor is configured to execute computer-executable instructions, and when the computer-executable instructions are executed by the processor, the various steps of the above-described method for realizing intelligent processing of mass spectrometry data and molecular prediction workflow are implemented.

[0014] The computer-readable storage medium is characterized in that it stores a computer program thereon, which can be executed by a processor to implement the various steps of the above-described method for implementing intelligent processing of mass spectrometry data and molecular prediction workflow.

[0015] This invention employs a method, apparatus, processor, and computer-readable storage medium for intelligent processing and molecular prediction of mass spectrometry data. Through a standardized noise reduction algorithm, it effectively eliminates background noise interference in mass spectrometry data, providing a clean data foundation for the analysis of complex systems. Addressing the needs of rapidly iterating novel compound development, it establishes an intelligent molecular structure prediction model based on a large model, overcoming the limitations of traditional static databases' update lag. This invention innovatively constructs a standard workflow encompassing data preprocessing, fingerprint recognition, structure prediction, similarity calculation, chemical information retrieval, and result output. This unifies previously fragmented analytical steps into a unified processing paradigm, automating molecular prediction based on mass spectrometry data and driving chemical research towards high automation and intelligence. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the multi-source mass spectrometry standardization and noise reduction process for the method of realizing intelligent processing of mass spectrometry data and molecular prediction workflow of the present invention.

[0017] Figure 2 This diagram illustrates the molecular structure prediction process based on a large model, which is part of the method for intelligent processing of mass spectrometry data and molecular prediction workflow of the present invention.

[0018] Figure 3 This is a schematic diagram of the molecular structure prediction workflow engine operation process of the method for realizing intelligent processing of mass spectrometry data and molecular prediction workflow of the present invention. Detailed Implementation

[0019] To more clearly describe the technical content of the present invention, the following description is provided in conjunction with specific embodiments.

[0020] The method for intelligent processing of mass spectrometry data and molecular prediction workflow of the present invention includes the following steps: (1) Standardize and denoise the mass spectrometry data, and construct the molecular fingerprint of the standard mass spectrometry set; (2) Input the molecular fingerprint into the mass spectrometry core operator neural network for structure prediction; (3) The output of the mass spectrometry core operator neural network is compared with the chemical vector library to calculate the similarity and select the top n most similar candidate molecules to construct a candidate molecule structure set; (4) Retrieve the chemical information of candidate molecules through the model context protocol service and integrate the information with the prediction results; (5) Input the integrated data into the large model base and generate multi-dimensional output.

[0021] In a preferred embodiment of the present invention, step (1) specifically includes the following steps: (1.1) Extract raw mass spectrometry data of the target compound from a multi-source mass spectrometry database; (1.2) Standardization processing; (1.3) Perform feature conversion to convert the standardized mass spectrum into a molecular fingerprint.

[0022] In a preferred embodiment of the present invention, step (1.2) specifically includes the following steps: (1.2.1) The mass spectrometry data is preprocessed using an adaptive noise reduction algorithm to perform baseline correction, smoothing and noise reduction, and peak detection and integration, providing a list of reliable peaks; (1.2.2) Standardize the original intensity in the list of reliable peaks and output the purified standard mass spectrometry data.

[0023] In a preferred embodiment of the present invention, step (2) specifically includes the following steps: (2.1) Construct a neural network for the core operator of mass spectrometry; (2.2) Input the constructed molecular fingerprint vector into the neural network for structure prediction, output the potential structure representation of the target molecule, and further embed the prediction result into the feature vector.

[0024] In a preferred embodiment of the present invention, step (3) specifically includes the following steps: (3.1) The results are passed to the computing node, and the chemical vector library is automatically called. The mass spectrometry prediction output vector is used as the retrieval vector to perform a similarity search in the high-dimensional space. (3.2) Calculate the similarity between vectors and select the top n structures that are most similar to the current prediction result as potential candidate molecules; (3.3) Pack the above n structures into a candidate molecular structure set.

[0025] In a preferred embodiment of the present invention, step (4) specifically includes the following steps: (4.1) The candidate molecule set is passed to the chemical information retrieval node, and the API is automatically called to connect to the open source chemical information database to perform reverse query and automatically obtain the basic attribute information of the corresponding molecules; (4.2) Integrate the structure prediction results with the queried chemical information to form an integrated information package.

[0026] In a preferred embodiment of the present invention, step (8) specifically includes the following steps: (5.1) Optimize prompt words by adjusting and optimizing them to generate multi-dimensional output; (5.2) Input the optimized prompt words along with the integrated information into the base model. Through joint encoding of structural content and task prompts, multi-level semantic modeling and knowledge reasoning are completed. (5.3) Generate multi-dimensional output results based on optimized input information.

[0027] The apparatus of the present invention for realizing intelligent processing of mass spectrometry data and molecular prediction workflow is characterized in that the apparatus comprises: A processor is configured to execute computer-executable instructions; The memory stores one or more computer-executable instructions, which, when executed by the processor, implement the various steps of the method described above for implementing intelligent processing of mass spectrometry data and molecular prediction workflow.

[0028] The processor of the present invention for realizing intelligent processing of mass spectrometry data and molecular prediction workflow is characterized in that the processor is configured to execute computer-executable instructions, and when the computer-executable instructions are executed by the processor, the various steps of the above-described method for realizing intelligent processing of mass spectrometry data and molecular prediction workflow are implemented.

[0029] The computer-readable storage medium of the present invention is characterized in that it stores a computer program thereon, the computer program being executable by a processor to implement the various steps of the above-described method for implementing intelligent processing of mass spectrometry data and molecular prediction workflow.

[0030] This invention belongs to the field of AI-driven cheminformatics, specifically relating to a method for mass spectrometry data processing and molecular structure prediction based on workflow engine and large model collaboration. It aims to address the problems of low mass spectrometry resolution efficiency and delayed molecular structure generation in existing technologies. This method, based on model context protocol services and using a large model as the computational foundation, achieves end-to-end intelligent prediction from mass spectrometry to molecular structure, significantly improving the speed and accuracy of unknown compound identification and providing an efficient analytical tool for fields such as drug development and environmental monitoring.

[0031] This method achieves intelligent analysis from mass spectrometry to molecular structure through a phased workflow architecture. The workflow first receives the mass spectrometry data submitted by the user and immediately initiates a standardization processing node to denoise and unify the format of the raw signal, ensuring the input quality for subsequent modules. The workflow then automatically routes the processed data to a feature transformation node, where the mass spectrometry peak sequence is converted into a molecular fingerprint using the RDKit algorithm. Next, the workflow enters a structure prediction node, transmitting the features to a specially designed mass spectrometry core operator neural network to generate preliminary molecular structure predictions (SMILES representation) and convert them into vectorized embeddings. Then, it enters a computation node where the model context protocol service automatically calls a similarity calculation service to efficiently match these vectors with a chemical vector library, selecting the n candidate structures with the highest similarity to form a molecular structure candidate set. Next, it enters a retrieval node that automatically calls a chemical information retrieval service to obtain detailed chemical information of these candidate structures from open-source databases such as PubChem. Finally, the workflow aggregates the outputs of each module and enters an information integration node, where it is processed by a large model platform to output the final potential molecular structure and its related chemical properties. This phased workflow design not only overcomes the limitations of a single model in processing complex mass spectrometry data, but also significantly improves the accuracy and reliability of molecular structure prediction through the collaborative cooperation of various professional modules, while ensuring the interpretability and traceability of the entire analysis process.

[0032] In a specific embodiment of the present invention, an intelligent processing and molecular prediction workflow method for mass spectrometry data is proposed, covering a complete process from raw mass spectrometry data processing, molecular structure prediction, candidate structure screening, chemical information retrieval to multidimensional intelligent output, specifically including the following steps: S1: Mass spectrometry data standardization and noise reduction processing, and construction of molecular fingerprints for standard mass spectrometry sets.

[0033] S2: Input the molecular fingerprint into the mass spectrometry core operator neural network for structure prediction.

[0034] S3: Perform similarity calculations between the output of the mass spectrometry core operator neural network and the chemical vector library, select the top n most similar candidate molecules, and construct a candidate molecule structure set.

[0035] S4: Retrieve chemical information of candidate molecules, including molecular formula and chemical structure, through the model context protocol service, and integrate the information with the prediction results.

[0036] S5: Input the integrated data into the large model base to generate multi-dimensional output.

[0037] In this embodiment, step S1 specifically includes the following steps: S11: Extract raw mass spectrometry data of the target compound from a multi-source mass spectrometry database, including m / z (mass-to-charge ratio) values ​​and their corresponding intensity information.

[0038] Specifically, to facilitate data storage, cross-platform transmission, and subsequent model processing, the mass spectrometry data can be output as a file in a unified format after acquisition. The output format includes, but is not limited to: MGF format, which is suitable for traditional mass spectrometry recognition engines and database integration; and JSON format, which is used to structure m / z intensity pairs and related metadata, making it convenient to use as standard input for parsing in the core mass spectrometry operators.

[0039] S12: The workflow enters the standardization processing node, where the mass spectrometry data is preprocessed using an adaptive noise reduction algorithm. A reliable peak list is provided through methods such as baseline correction, smoothing and noise reduction, and peak detection and integration. Then, the original intensity in the peak list is standardized through methods such as logarithmic transformation, quantile normalization, and Gaussian normalization, and the purified standard mass spectrometry data is output as the basis for downstream analysis.

[0040] During baseline correction, the moving window minimum method is used. For each data point, the minimum value within a range of several points before and after it (the window size can be customized) is taken as a local limit estimate. Then, the baseline is subtracted from the original signal to eliminate instrument baseline drift and low-frequency noise, highlighting the true peak signal.

[0041] During the smoothing and denoising process, a Savitzky-Golay filter is used. Within several sliding windows (the window size can be customized), an nth-order polynomial (the number of polynomial terms can be customized) is used to fit the signal, preserving the peak characteristics while suppressing high-frequency noise.

[0042] During peak detection and integration, only peaks with intensity exceeding the set quantile are retained (filtering out small noise peaks), and peak prominence is set (retaining significantly convex peaks); adjacent peaks with m / z differences less than a threshold are merged into the same cluster (the threshold can be customized), the average m / z value is taken as the new peak position, and the cumulative intensity is taken as the total intensity.

[0043] During the standardization process, logarithmic transformation is used to compress the dynamic range of the data; the data is scaled based on the 25th and 75th percentiles to reduce the impact of extreme values; and finally, it is converted into a standard normal distribution with a mean of 0 and a standard deviation of 1 to improve the stability of subsequent modeling.

[0044] S13: The workflow enters the feature transformation node, where chemical toolkits such as RDKit are used to convert the standardized mass spectrum into a molecular fingerprint representation. Specifically, a topological fingerprint with a radius of 2 and a length of 4096 is constructed, and it is embedded into a high-dimensional vector using a hash mapping method to build a molecular feature description with structural discrimination capabilities.

[0045] Once the data is ready, the workflow enters the structure prediction node, automatically invoking the mass spectrometry core operator neural network to predict molecules. In this embodiment, step S2 specifically includes the following steps: S21: Construct a mass spectrometry core operator neural network.

[0046] S22: Input the molecular fingerprint vector constructed in S1 into the neural network for structure prediction, output the potential structure representation of the target molecule, and further embed the prediction result into a feature vector for downstream matching tasks.

[0047] After the molecular fingerprint and mass spectrometry prediction results are generated, the workflow is executed to the computing node. The model context protocol service automatically triggers the similarity calculation service and starts the candidate molecule generation task. In this embodiment, step S3 specifically includes the following steps: S31: The workflow passes the results to the computing node. The similarity calculation service automatically calls the chemical vector library and uses the mass spectrometry prediction output vector as the retrieval vector to perform a similarity search in the high-dimensional space.

[0048] In this embodiment, the system performs vector embedding processing on the structure prediction results generated by the mass spectrometry core neural network to obtain molecular characterization vectors with consistent dimensions. This vector is then used as a query vector input to the similarity calculation service. The service internally maintains a high-dimensional sparse vector index structure to complete an efficient structure similarity search operation.

[0049] S32: Calculate the similarity between vectors based on Euclidean distance, and select the top n structures that are most similar to the current prediction result as potential candidate molecules.

[0050] Preferably, the similarity assessment uses Euclidean distance as the metric, but alternative metrics such as cosine similarity, Manhattan distance (L1 distance), or Mahalanobis distance can be selected according to specific business needs.

[0051] Specifically, let the predicted structure vector be... The structure vector in the database is Then, by calculation: All candidate structures are sorted by distance, and the SMILES expressions corresponding to the top n structures are selected as the high similarity candidate structures of the current input molecule.

[0052] In particular, in this embodiment, the value of n can be set according to the task requirements (e.g., n=5 or n=10) to balance the diversity and credibility of candidate structures while ensuring search efficiency.

[0053] S33: Package the above n structures into a candidate molecular structure set for subsequent information integration and reasoning.

[0054] In this step, the first n structures are encapsulated into a candidate molecular structure set object in a unified format. The object structure includes index values, SMILES expressions, and other related information.

[0055] The candidate molecular structure set will be output in JSON format to ensure compatibility with the retrieval nodes in the model context protocol service chain, and can be directly input into large model systems for multidimensional structural analysis.

[0056] Then, the workflow executes to the retrieval node, where the Model Context Protocol service automatically loads the required features and automatically triggers the Chemical Information Retrieval Service to obtain basic descriptive information of candidate molecules, providing structural knowledge support for large model inference tasks. In this embodiment, step S4 specifically includes the following steps: S41: The workflow passes the candidate molecule set to the chemical information retrieval node. The chemical information retrieval service will automatically call the API to connect to the open-source chemical information database, perform a reverse query based on SMILES, and automatically obtain the basic attribute information of the corresponding molecules.

[0057] Specifically, the system uses the candidate SMILES generated in step S33 as query keywords to request the open API of the Chemistry Information Repository. The basic molecular attribute information returned by the database includes, but is not limited to, the following fields: molecular formula, molecular weight, standard structure identifier (InChIKey), IUPAC standard nomenclature, chemical description, etc.

[0058] S42: Integrate the structure prediction results with the retrieved chemical information to form an integrated information package with a unified format and consistent semantics, which is used to drive large model inference.

[0059] In particular, the information package is stored in JSON format, which is highly structured and scalable. It can be directly used as input prompt words or embedded feature vectors for the base model to drive subsequent multi-dimensional analysis tasks.

[0060] Finally, the workflow aggregates the outputs of each module, inputting the optimized input prompts and integrated information into the base model to further generate multi-dimensional outputs such as substructure fragments, physicochemical property predictions, and molecular SMILES for each candidate molecule. In this embodiment, step S5 specifically includes the following steps: S51: Optimize prompt words. By adjusting and optimizing prompt words, ensure that the large model can accurately understand task requirements and generate more efficient and relevant multi-dimensional outputs.

[0061] In this embodiment, the system constructs task-oriented prompt word templates and, combined with the characteristics of the molecular prediction task, performs semantically guided encoding of candidate molecules and their ensemble information. This step aims to improve the accuracy, responsiveness, and generation quality of the large-scale base model when processing structural description-type inputs.

[0062] Specifically, keyword optimization strategies include, but are not limited to, the following measures: Streamline non-critical content and highlight task verbs and objective descriptions; Input standardization and fault tolerance processing prevent output interruption due to missing data; An automated extension of substructure analysis, automatically linking its functional group characteristics with its physicochemical properties; The format of input data can be flexibly designed according to task requirements.

[0063] S52: The optimized prompts, along with integrated information, are input into the base model. The model performs multi-level semantic modeling and knowledge reasoning by jointly encoding the structural content and task prompts.

[0064] Specifically, the optimized model in this invention has a temperature value of 0.5 (customizable) and a maximum number of tokens of 500 (customizable). Optionally, a top-n sampling strategy (e.g., n=10) can be set to improve the rationality of the generated content.

[0065] S53: The large base model generates multi-dimensional output results based on optimized input information.

[0066] Preferably, the output of the large model includes, but is not limited to, the following dimensions: Predicting the physicochemical properties of candidate molecules Basic description of molecular structure The generated results are ultimately encapsulated into a JSON response object, which can be used for subsequent visualization.

[0067] The technical solution of this invention uses mass spectrometry data itself as input. After standardization, noise reduction, and fingerprinting, it is used as input to a neural network. The goal is to infer molecular structure, and the task is "to predict potential chemical structures through mass spectrometry of unknown substances".

[0068] The technical solution of this invention focuses on automated molecular structure analysis, aiming to quickly and accurately infer the structure of unknown compounds from experimental mass spectrometry data, and solves the problem of using large models to predict potential chemical structures.

[0069] In this invention, the neural network serves only as a core operator, responsible for extracting features from mass spectrometry data. Relying on a model context protocol, it integrates all working modules into a unified workflow to predict potential chemical molecular structures. This invention employs a workflow-driven, multi-module architecture, with the neural network being just one module within it.

[0070] The technical solution of this invention generates potential molecular structures by inputting the target mass spectrometer into a mass spectrometry operator neural network, and then connects it to an external database for similarity calculation and information integration. A large model is used to achieve diversified results generation of the target mass spectrometer, potential molecular sets, and their information. The key to this invention's technical solution lies in building a workflow, achieving automated and standardized processing through modular design (operator network, similarity calculation, chemical information database, external database, large model inference, etc.).

[0071] The technical solution of this invention meets diverse needs such as standardization, automation, and scalability by constructing a workflow, thereby adapting to multi-source databases and different tasks, and proposes a standard workflow scheme for intelligent processing of mass spectrometry data and molecular prediction.

[0072] For the specific implementation scheme of this embodiment, please refer to the relevant descriptions in the above embodiments, which will not be repeated here.

[0073] It is understood that the same or similar parts in the above embodiments can be referred to each other, and the contents not described in detail in some embodiments can be referred to the same or similar contents in other embodiments.

[0074] It should be noted that in the description of this invention, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance. Furthermore, in the description of this invention, unless otherwise stated, "a plurality of" means at least two.

[0075] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process, and the scope of the preferred embodiments of the invention includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as will be understood by those skilled in the art to which embodiments of the invention pertain.

[0076] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution device.

[0077] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The corresponding program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiments.

[0078] Furthermore, the functional units in the various embodiments of the present invention can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0079] The storage media mentioned above can be read-only memory, disk, or optical disk, etc.

[0080] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0081] This invention employs a method, apparatus, processor, and computer-readable storage medium for intelligent processing and molecular prediction of mass spectrometry data. Through a standardized noise reduction algorithm, it effectively eliminates background noise interference in mass spectrometry data, providing a clean data foundation for the analysis of complex systems. Addressing the needs of rapidly iterating novel compound development, it establishes an intelligent molecular structure prediction model based on a large model, overcoming the limitations of traditional static databases' update lag. This invention innovatively constructs a standard workflow encompassing data preprocessing, fingerprint recognition, structure prediction, similarity calculation, chemical information retrieval, and result output. This unifies previously fragmented analytical steps into a unified processing paradigm, automating molecular prediction based on mass spectrometry data and driving chemical research towards high automation and intelligence.

[0082] In this specification, the invention has been described with reference to specific embodiments thereof. However, it will be apparent that various modifications and variations can be made without departing from the spirit and scope of the invention. Therefore, the specification and drawings should be considered illustrative rather than restrictive.

Claims

1. A method for implementing an intelligent processing and molecular prediction workflow of mass spectrometry data, characterized in that, The method comprises the following steps: (1) standardizing and denoising the mass spectrometry data, and constructing a molecular fingerprint of the standard mass spectrometry set; (2) inputting the molecular fingerprint into a mass spectrometry core operator neural network for structure prediction; (3) performing similarity calculation on the output result of the mass spectrometry core operator neural network and the chemical vector library to screen the first n candidate molecules most similar to the output result, and constructing a candidate molecule structure set; (4) retrieving the chemical information of the candidate molecules through a model context protocol service, and integrating the information with the prediction result; (5) inputting the integrated data into a large model base to generate multi-dimensional output.

2. The method for implementing a mass spectrometry data intelligent processing and molecule prediction workflow of claim 1, wherein, The step (1) specifically comprises the following steps: (1.1) extracting the original mass spectrometry data of the target compound from a multi-source mass spectrometry database; (1.2) performing standardization processing; (1.3) performing feature conversion to convert the standardized mass spectrometry graph into a molecular fingerprint.

3. The method for implementing mass spectrometry data intelligent processing and molecular prediction workflow of claim 2, wherein, The step (1.2) specifically comprises the following steps: (1.2.1) preprocessing the mass spectrometry data through an adaptive denoising algorithm, performing baseline correction, smoothing denoising, and peak detection and integration to provide a reliable peak list; (1.2.2) standardizing the original intensity in the reliable peak list to output the purified standard mass spectrometry data.

4. The method for implementing mass spectrometry data intelligent processing and molecular prediction workflow of claim 1, wherein, The step (2) specifically comprises the following steps: (2.1) constructing a mass spectrometry core operator neural network; (2.2) inputting the constructed molecular fingerprint vector into the neural network for structure prediction, outputting the potential structure representation of the target molecule, and further embedding the prediction result into a feature vector.

5. The method for implementing mass spectrometry data intelligent processing and molecular prediction workflow of claim 1, wherein, The step (3) specifically comprises the following steps: (3.1) passing the result to a calculation node to automatically call the chemical vector library, taking the mass spectrometry prediction output vector as a retrieval vector, and performing similarity search in a high-dimensional space; (3.2) calculating the similarity between the vectors to match and select the first n structures most similar to the current prediction result as potential candidate molecules; (3.3) packaging the above n structures to form a candidate molecule structure set.

6. The method for implementing mass spectrometry data intelligent processing and molecular prediction workflow of claim 1, wherein, The step (4) specifically comprises the following steps: (4.1) inputting the candidate molecule set into a chemical information retrieval node to automatically call an API to interface an open-source chemical information library, performing reverse query, and automatically obtaining the basic attribute information of the corresponding molecule; (4.2) integrating the structure prediction result with the queried chemical information to form an integrated information package.

7. The method for implementing mass spectrometry data intelligent processing and molecular prediction workflow of claim 1, wherein, The step (8) specifically comprises the following steps: (5.1) performing prompt word optimization, generating multi-dimensional output by adjusting and optimizing the prompt words; (5.2) inputting the optimized prompt words together with the integrated information into the base large model, completing multi-level semantic modeling and knowledge reasoning through joint coding of structure content and task prompts; (5.3) generating multi-dimensional output results based on the optimized input information.

8. An apparatus for implementing an intelligent processing and molecular prediction workflow of mass spectrometry data, characterized in that, The device comprises: a processor configured to execute computer executable instructions; a memory storing one or more computer executable instructions, which, when executed by the processor, implement each step of the method for implementing the intelligent processing of mass spectrometry data and molecular prediction workflow according to any one of claims 1 to 7.

9. A processor for implementing an intelligent processing and molecular prediction workflow of mass spectrometry data, characterized in that, The processor is configured to execute computer-executable instructions, which, when executed by the processor, implement the steps of the method for implementing an intelligent processing and molecular prediction workflow of mass spectrometry data according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, A computer program is stored thereon, which can be executed by a processor to implement the steps of the method for implementing an intelligent processing and molecular prediction workflow of mass spectrometry data according to any one of claims 1 to 7.

Images