Recognition method and application of respiratory tract pathogen marker
By combining gas chromatography-mass spectrometry and Transformer model, the identification of respiratory pathogen markers is solved, and the problems of long detection time and insufficient sensitivity and specificity in the prior art are solved, and rapid and accurate pathogen detection is achieved.
Patent Information
- Application Number
- CN202510110592.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-23
- Publication Date
- 2025-05-13
AI Technical Summary
The existing respiratory pathogen detection technology has problems of long detection time and insufficient sensitivity and specificity, especially when the sample processing and analysis process of PCR technology is not suitable for rapid screening and emergency response, and the rapid antigen detection is insufficient in accuracy when the viral load is low.
The gas chromatography-mass spectrometry combined technology was used to combine with the Transformer model, and the GC-MS map data of the target gas was obtained for preprocessing and classification marking. The GC-MS map data was analyzed using the trained Transformer model to identify and label respiratory pathogen markers.
It realizes rapid and accurate identification of respiratory pathogen markers, can be effectively applied to the detection of respiratory pathogens in regional air, greatly reducing the judgment workload of professionals and able to replace professionals to a certain extent.
Smart Images
Figure CN119985819A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of deep learning and biological analysis, and more specifically, to a method for identifying respiratory pathogen markers and applications. Background Art
[0002] In the technical field of respiratory pathogen detection, the current situation mainly relies on polymerase chain reaction (PCR) technology and antigen rapid detection. PCR technology is widely used due to its high sensitivity and high specificity, and is suitable for detecting various viral and bacterial respiratory diseases. However, PCR testing usually requires professional laboratory equipment and well-trained technicians to operate, and the time from sample collection to result output is long (usually several hours to a day). Although antigen rapid testing can provide faster test results (usually within 15 to 30 minutes), its sensitivity and accuracy are usually lower than PCR testing.
[0003] Therefore, the main technical problems in the current field of respiratory pathogen detection include:
[0004] 1. Long detection time: Especially for PCR technology, its long sample processing and analysis process is not suitable for rapid screening and emergency response.
[0005] 2. Limitations in sensitivity and specificity: Rapid antigen testing is fast but lacks accuracy, especially at the stage of low viral load. Summary of the invention
[0006] The object of the present invention is to provide a method for identifying respiratory pathogen markers, which can quickly and accurately assist users in identifying respiratory pathogen markers and can be effectively applied to the detection of respiratory pathogens in regional air.
[0007] In order to solve the above technical problems, the technical solution adopted by the present invention is:
[0008] In a first aspect, the present invention provides a method for identifying respiratory pathogen markers, comprising the following steps:
[0009] S1. Obtaining GC-MS spectrum data of the target gas and preprocessing the GC-MS spectrum data;
[0010] S2. Input the GC-MS spectrum data into the trained Transformer model, and the Transformer model analyzes the GC-MS spectrum data to identify and mark the markers.
[0011] The Transformer model is a deep learning model based on the self-attention mechanism. The training method of the Transformer model described in the present invention comprises the following steps: collecting existing GC-MS spectrum data related to respiratory pathogen markers; preprocessing the GC-MS spectrum data; annotating the preprocessed GC-MS spectrum data to identify compounds that are respiratory pathogen markers; dividing the processed GC-MS spectrum data into a training set and a validation set, inputting the training set into the Transformer model for training, and using the validation set to evaluate the performance of the trained Transformer model for optimization and adjustment.
[0012] The method for obtaining GC-MS spectrum data of the target gas comprises the following steps: detecting the collected target gas by gas chromatography-mass spectrometry technology to obtain GC-MS spectrum data.
[0013] The target gas may originate from, including but not limited to, air collected from a target area.
[0014] The preprocessing includes peak extraction, peak retrieval and peak filtering.
[0015] The peak extraction method is as follows: peak detection is performed on the mass spectrum in the GC-MS spectrum data, and peak extraction is performed using maximum chromatographic peak extraction and signal-to-noise ratio threshold screening to obtain characteristic peaks. As one implementation method, mzR is used to perform peak detection on the mass spectrum in the GC-MS spectrum data.
[0016] The peak search method is as follows: the characteristic peaks extracted by the peaks are compared and matched with the database to obtain the mass, molecular weight, substance name, molecular formula and other information of the molecular ion peak.
[0017] The peak filtering method is as follows: the characteristic peaks of the peak-searched spectrum are compared with the characteristic peaks of the substance, and filtering is performed according to the mass of the molecular ion peak and the retention index of the peak to remove misjudgment.
[0018] Furthermore, the preprocessing also includes selecting the position of the characteristic peak of the characteristic ion, the mass, peak intensity, and mass-to-charge ratio information of the molecular ion peak, and establishing a data set; and performing normalization processing on the data set.
[0019] The Transformer model includes an input module, a multi-head attention module, a residual connection module, a feedforward neural network module and an output module.
[0020] The Transformer model has the following format requirements for input data: each row represents a compound, and each compound contains the following data units: compound name, retention time, mass of molecular ion peak, peak intensity, and mass-to-charge ratio.
[0021] The input module of the Transformer model uses a data normalization method to process the input values into training data with a mean of 0 and a standard deviation of 1, and uses Position Embedding to embed position information.
[0022] As one implementation method, in the input data, the values of the compound name, retention time, mass of the molecular ion peak, and peak intensity data are multiplied by different weights to obtain the final output value. The input module of the Transformer model performs correlation operations on the mass of the molecular ion peak and the peak intensity of the peak in each data and then adds them as the label input of the target task.
[0023] The data passing through the input module of the Transformer model enters the multi-head self-attention module for feature extraction and addition operations.
[0024] The multi-head self-attention module of the Transformer model divides single-head attention into multi-head spatial attention and multi-head contextual attention. Through the multi-head attention mechanism, the model extracts features at different positions.
[0025] Furthermore, the feature maps passed by the multi-head self-attention module are added and then passed through the residual connection module, and the data after passing the residual module is transmitted to the feedforward neural network module. After linear transformation and normalization, the features are finally output through the output module.
[0026] The Transformer model adds a residual module to each layer, adds the input signal and the output signal obtained by the self-attention module, and then transmits it to the next module.
[0027] More preferably, the output module of the Transformer model classifies each compound separately and outputs the probability that each compound belongs to a marker.
[0028] The recognition and labeling process of the input GC-MS spectrum data by the Transformer model of the present invention includes: obtaining a weight matrix through training the Transformer model, obtaining a feature matrix after data standardization processing of the input GC-MS spectrum data; sending the feature matrix to the Transformer model for processing to obtain the identification of the marker.
[0029] More specifically, the present invention provides a method for identifying respiratory pathogen markers comprising the following steps:
[0030] S1. Obtain GC-MS spectrum data of the target gas.
[0031] S2. Preprocessing the GC-MS spectrum data, including peak extraction, peak retrieval and filtering, to obtain high-quality GC-MS spectrum data;
[0032] S3. Classify and label the pre-processed GC-MS spectrum data, and divide them into training set and validation set, compare the characteristic peaks of the GC-MS spectrum data with the database, and label the corresponding characteristic spectra with categories to obtain training set and validation set;
[0033] S4. Input the training set data into the Transformer model for training, use the validation set data to determine whether the training results meet the requirements and make adjustments to obtain a Transformer model for identifying markers;
[0034] S5. Identify the GC-MS spectrum data that needs to be judged through the Transformer model and output the results.
[0035] The GC-MS spectrum data preprocessing method described in step S2 serves the purpose of improving the quality of the spectrum data and filtering noise.
[0036] In step S3, the training set and the validation set are generally divided in a ratio of 8:2. Optionally, the number of categories in the training set and the validation set is consistent, and the number of each category is ensured to be sufficient. After step S3, the data corresponding to each marker is obtained. Specifically, each marker is respectively annotated with the name of the corresponding compound, the mass of the molecular ion peak, and the peak intensity of the peak.
[0037] The present invention uses the Transformer model to assist in the identification of respiratory pathogen markers. The Transformer model has unique advantages in processing semi-structured or unstructured data. It usually only takes a few groups of data to mark a target compound, namely the compound name, the corresponding MS characteristic ion and its relative intensity, with fast detection speed, high sensitivity and strong specificity.
[0038] The respiratory pathogens include but are not limited to influenza A virus, respiratory syncytial virus, parainfluenza virus, Middle East respiratory syndrome coronavirus, SARS-related coronavirus, MERS-related coronavirus and new coronavirus.
[0039] In a second aspect, another object of the present invention is to provide a device for a method for identifying respiratory pathogen markers. The device comprises:
[0040] An input unit for acquiring GC-MS spectrum data;
[0041] A processing unit, used for processing the GC-MS spectrum data, wherein the processing includes preprocessing and Transformer model processing;
[0042] The output unit is used to output the identified pathogen markers based on the processed GC-MS spectrum data.
[0043] In a third aspect, the present invention further provides an electronic device. The electronic device includes at least one memory and at least one processor; the at least one memory is coupled to the at least one processor, the at least one memory is used to store a computer program, the at least one processor is used to call the computer program, and the computer program includes instructions, and when the instructions are executed by the at least one processor, the electronic device executes the method for identifying respiratory pathogen markers as described in the first aspect above.
[0044] In a fourth aspect, the present invention further provides a computer storage medium comprising computer instructions, which, when executed on an electronic device, enable the electronic device to execute the respiratory pathogen marker identification method described in the first aspect.
[0045] Compared with the prior art, the present invention has the following beneficial effects:
[0046] The identification method of the present invention can quickly and accurately assist users in identifying respiratory pathogen markers by combining gas chromatography-mass spectrometry technology with the Transformer model, and can be effectively applied to the detection of respiratory pathogens in regional air. The identification method of the present invention uses deep learning technology, which can greatly reduce the workload of professionals in identifying respiratory pathogen markers and can replace professionals to a certain extent. BRIEF DESCRIPTION OF THE DRAWINGS
[0047] Figure 1 Flowchart of the method for identifying respiratory pathogen markers.
[0048] Figure 2 The present invention is a functional module block diagram of a device for identifying respiratory pathogen markers.
[0049] Figure 3 This is a structural block diagram of an electronic device. DETAILED DESCRIPTION
[0050] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0051] The following describes a method for identifying respiratory pathogen markers in an embodiment of the present invention. The GC-MS spectrum data of respiratory pathogen markers in the air is established for training, and the trained Transformer model is used to identify respiratory pathogen markers. The advantage of this method is that it does not require complex clinical sample pretreatment, has fast detection speed, high sensitivity, and strong specificity.
[0052] Specifically, Figure 1 As shown, the method for identifying respiratory pathogen markers includes the following steps:
[0053] S1. Obtain GC-MS spectrum data of the target gas.
[0054] S2. Preprocessing the GC-MS spectrum data, including peak extraction, peak retrieval and filtering, to obtain high-quality GC-MS spectrum data;
[0055] S3. Classify and label the pre-processed GC-MS spectrum data, and divide them into training set and validation set, compare the characteristic peaks of the GC-MS spectrum data with the database, and label the corresponding characteristic spectra with categories to obtain training set and validation set;
[0056] S4. Input the training set data into the Transformer model for training, use the validation set data to determine whether the training results meet the requirements and make adjustments to obtain a Transformer model for identifying markers;
[0057] S5. Identify the GC-MS spectrum data that needs to be judged through the Transformer model, identify and mark the respiratory pathogen markers, and output the results.
[0058] The training method of the Transformer model comprises the following steps:
[0059] a) Collect existing GC-MS spectroscopic data related to respiratory pathogens;
[0060] b) preprocessing the GC-MS spectrum data;
[0061] c) annotating the preprocessed GC-MS spectral data to identify compounds that are markers of respiratory pathogens;
[0062] d) Divide the processed GC-MS spectrum data into a training set and a validation set, input the training set into the Transformer model for training, and use the validation set to evaluate the performance of the trained Transformer model for optimization and adjustment.
[0063] Preferably, the trained Transformer model can be deployed on a web server, and the GC-MS spectrum data of the target gas can be received through the web server and input into the adjusted Transformer model to obtain the recognition result.
[0064] In step a), the respiratory pathogens include but are not limited to influenza A virus, respiratory syncytial virus, parainfluenza virus, Middle East respiratory syndrome coronavirus, SARS-related coronavirus, MERS-related coronavirus and new coronavirus.
[0065] In step b), the GC-MS spectrum data is preprocessed comprising the following steps:
[0066] b-1) using mzR to perform peak detection on the mass spectrum in the GC-MS spectrum data, using maximum chromatographic peak extraction and signal-to-noise ratio threshold screening to extract the peak, and obtain the characteristic peak of the characteristic ion; comparing and matching the characteristic peak after the peak extraction with the database to obtain the mass, molecular weight, substance name, molecular formula and other information of the molecular ion peak; comparing the characteristic peak of the spectrum after the peak search with the characteristic peak of the substance, filtering according to the mass of the molecular ion peak and the retention index of the peak, and removing misjudgment;
[0067] b-2) selecting the position of the characteristic peak of the characteristic ion, the mass, peak intensity, and mass-to-charge ratio information of the molecular ion peak, and establishing a data set;
[0068] b-3) Normalize the data set;
[0069] Step b-2) comprises: selecting a retention time within a specific retention time range as position information of characteristic ions according to GC-MS spectrum data; selecting ions stably expressed within the specific retention time range as mass-to-charge ratio information of characteristic ions;
[0070] The mass-to-charge ratios for different ions may vary due to their different relative molecular masses, so the mass-to-charge ratios for different compounds within a specific retention time range may be different.
[0071] The peak intensity will be different for different mass-to-charge ratios, so for the mass-to-charge ratio within a specific retention time range, the stably expressed ion peak can be preferred, and 2 / 3 of the stably expressed ion peaks can be selected. When the intensity of the stably expressed ion peak within the selected specific retention time range is less than 5%, the characteristic ion is reselected and the data set is reestablished.
[0072] The step c) comprises:
[0073] c-1) Select some characteristic ions, annotate them manually, and mark them as respiratory pathogen markers;
[0074] c-2) Based on the data set containing the characteristic ions manually identified as respiratory pathogen markers, search the data set for stably expressed ions within the same retention time range, and automatically identify the stably expressed ions found as respiratory pathogen markers.
[0075] In the step d),
[0076] The Transformer model includes an input module, a multi-head attention module, a residual connection module, a feedforward neural network module and an output module.
[0077] The Transformer model requires the following format for input data: each row represents a compound, and each compound contains the following data units: compound name, retention time, mass of molecular ion peak, peak intensity, and mass-to-charge ratio.
[0078] The input module of the Transformer model uses a data standardization method to process the input values into training data with a mean of 0 and a standard deviation of 1, and uses Position Embedding to embed the position information. In the input data, the values of the compound name, retention time, mass of the molecular ion peak, and peak intensity data are multiplied by different weights to obtain the final output value. The input module of the Transformer model performs correlation operations on the mass of the molecular ion peak and the peak intensity of the peak in each data and adds them as the label input of the target task.
[0079] The data from the input module of the Transformer model enters the multi-head self-attention module for feature extraction and summation. The multi-head self-attention module of the Transformer model divides the single-head attention into multi-head spatial attention and multi-head contextual attention. Through the multi-head attention mechanism, the model extracts features at different positions.
[0080] The feature graphs passed by the multi-head self-attention module are summed up and passed through the residual connection module. The data after the residual module is transmitted to the feedforward neural network module. After linear transformation and normalization, the features are finally output through the output module. The Transformer model adds a residual module to each layer, adds the input signal and the output signal obtained by the self-attention module, and then transmits it to the next module.
[0081] The output module of the Transformer model is to classify each compound separately and output the probability that each compound belongs to a marker.
[0082] The recognition and labeling process of the input GC-MS spectrum data by the Transformer model of the present invention includes: obtaining a weight matrix through training the Transformer model, obtaining a feature matrix after data standardization processing of the input GC-MS spectrum data; sending the feature matrix to the Transformer model for processing to obtain the identification of the marker.
[0083] This embodiment also provides a device for the method of identifying respiratory pathogen markers.
[0084] The following describes an apparatus for executing the above-mentioned respiratory pathogen marker identification method.
[0085] Figure 2 The structural block diagram of the device provided in this embodiment is shown, and the device of the respiratory pathogen marker identification method includes:
[0086] An input unit 10 is used to obtain GC-MS spectrum data;
[0087] A processing unit 20, used for processing the GC-MS spectrum data, wherein the processing includes preprocessing and Transformer model processing;
[0088] The output unit 30 is used to output the identified pathogen markers according to the processed GC-MS spectrum data.
[0089] The device provided in this embodiment and the respiratory pathogen marker identification method provided in this application belong to the same concept. The specific implementation process is detailed in the full text of the specification and will not be repeated here.
[0090] This embodiment also provides an electronic device. The electronic device that can execute the above-mentioned respiratory pathogen marker identification method is described below.
[0091] Figure 3A structural block diagram of an electronic device 100 for implementing a method for identifying a respiratory pathogen marker is shown, wherein the electronic device 100 includes: at least one memory 101 and at least one processor 102; the at least one memory 101 is coupled to the at least one processor 102, the at least one memory 101 is used to store a computer program, and the at least one processor 102 is used to call the computer program, wherein the computer program includes instructions, and when the instructions are executed by the at least one processor, the electronic device executes the above-mentioned method for identifying a respiratory pathogen marker.
[0092] This embodiment also provides a computer storage medium. The following describes a computer storage medium containing the above-mentioned respiratory pathogen marker identification method.
[0093] A computer storage medium includes computer instructions, and when the computer instructions are executed on an electronic device, the electronic device executes the above-mentioned respiratory pathogen marker identification method.
[0094] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the embodiments here. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the claims of the present invention.
Claims
1. A method for identifying respiratory pathogen markers, characterized in that: The following steps are involved: S1. Obtaining GC-MS spectrum data of the target gas and preprocessing the GC-MS spectrum data; S2. Input the GC-MS spectrum data into the trained Transformer model, and the Transformer model analyzes the GC-MS spectrum data to identify and mark the markers.
2. The method for identifying respiratory pathogen markers according to claim 1, characterized in that: The training method of the Transformer model comprises the following steps: collecting existing GC-MS spectrum data related to respiratory pathogen markers; preprocessing the GC-MS spectrum data; annotating the preprocessed GC-MS spectrum data to identify compounds that are respiratory pathogen markers; dividing the processed GC-MS spectrum data into a training set and a validation set, inputting the training set into the Transformer model for training, and using the validation set to evaluate the performance of the trained Transformer model.
3. The method for identifying respiratory pathogen markers according to claim 1, characterized in that: The method for obtaining GC-MS spectrum data of the target gas comprises the following steps: detecting the collected target gas by gas chromatography-mass spectrometry technology to obtain GC-MS spectrum data.
4. The method for identifying respiratory pathogen markers according to claim 1, characterized in that: The preprocessing includes peak extraction, peak retrieval and peak filtering.
5. The method for identifying respiratory pathogen markers according to claim 1, characterized in that: The Transformer model includes an input module, a multi-head attention module, a residual connection module, a feedforward neural network module and an output module.
6. The method for identifying respiratory pathogen markers according to claim 1 or 5, characterized in that: The Transformer model has the following format requirements for input data: each row represents a compound, and each compound contains the following data units: compound name, retention time, mass and peak intensity of molecular ion peak.
7. The method for identifying respiratory pathogen markers according to claim 1, characterized in that: The respiratory pathogens include influenza A virus, respiratory syncytial virus, parainfluenza virus, Middle East respiratory syndrome coronavirus, SARS-related coronavirus, MERS-related coronavirus and new coronavirus.
8. A device for identifying a respiratory pathogen marker, characterized in that: include: An input unit for acquiring GC-MS spectrum data; A processing unit, used for processing the GC-MS spectrum data, wherein the processing includes preprocessing and Transformer model processing; The output unit is used to output the identified pathogen markers based on the processed GC-MS spectrum data.
9. An electronic device, characterized in that: comprising at least one memory and at least one processor; The at least one memory is coupled to the at least one processor, the at least one memory is used to store a computer program, the at least one processor is used to call the computer program, the computer program includes instructions, when the instructions are executed by the at least one processor, the electronic device executes the method for identifying respiratory pathogen markers as described in any one of claims 1 to 7.
10. A computer storage medium, characterized in that: The method comprises computer instructions, which, when executed on an electronic device, enable the electronic device to execute the method for identifying a respiratory pathogen marker as claimed in any one of claims 1 to 7.