Chromatography and mass spectrometry automatic integration method, system, equipment and medium for multiple sample items

Abstract

The invention provides a chromatography-mass spectrometry automatic integration method, system, equipment and medium for multiple sample items, and the automatic integration method comprises the following steps: converting an original spectrogram of a chromatography-mass spectrometry to generate a standardized spectrogram picture; performing peak recognition on the standardized spectrogram picture by using a trained deep learning peak recognition algorithm to obtain a candidate peak classification result; screening the candidate peak classification results based on the spectrogram classification result to obtain a preliminary integration result; performing peak clustering on the multi-sample items; carrying out peak recall on samples without peak reservation; and carrying out consistency correction on all reserved peaks. According to the method, deep learning model prediction is carried out on the chromatography-mass spectrometry image, peak clustering, peak recall and peak boundary inconsistency correction are carried out in combination with multi-sample data, manual judgment on the peak position and the peak shape of the chromatography-mass spectrometry can be replaced, and the method is high in accuracy, high in specificity and high in accuracy. And the false positive of the conventional automatic integration method is effectively reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of proteomics and metabolomics analysis, and in particular to a chromatographic mass spectrometry automatic integration method, system, equipment and medium for a multi-sample project. Background Art

[0002] Currently, in the field of proteomics and metabolomics analysis, chromatography-tandem mass spectrometry is used as a combined technology that combines the efficient separation capability of chromatography with the precise molecular weight and intensity information of the molecules to be detected acquired by mass spectrometry. This technology has the advantages of high sensitivity, strong selectivity and high throughput, and can obtain rich qualitative and quantitative information on proteins and metabolites.

[0003] Although liquid chromatography-mass spectrometry (LC-MS) technology has excellent performance, it also makes the generated raw data extremely complex, which brings greater challenges to the processing of metabolomics data. LC-MS data usually contains thousands of unique raw spectra, and the sequence of these spectra increases with the retention time. The thousands of signals (features) contained in the raw data make manual processing cumbersome and time-consuming, so improving the processing efficiency of these huge data is receiving more and more attention.

[0004] Traditional chromatography mass spectrometry peak software integration algorithms mainly rely on known mathematical models and rules, such as peak shape, peak width, signal intensity, etc. Although these software have high computational efficiency when processing large amounts of data, they usually have the following defects. First, there are limitations in the processing of complex data. When the data contains a lot of noise or the signal is complex, traditional algorithms may have difficulty accurately identifying peaks and are susceptible to external interference; second, they are sensitive to parameter adjustments. These algorithms usually rely on many parameter settings, and manual adjustments may be time-consuming and require more professional knowledge of chromatography or analysis; in addition, they have poor adaptability to different data. Traditional algorithms are insensitive to changes in data and cannot automatically adjust according to new data patterns, especially in complex and non-standardized metabolite mixtures.

[0005] In order to solve the defects of traditional automatic integration algorithms or software, more and more automatic integration algorithms or software based on deep learning are being developed in recent years. Automatic integration tools based on deep learning have shown more accurate peak identification, adaptability to different chromatographic peaks, and lower parameter sensitivity than traditional integration algorithms in the laboratory, but it still faces some important shortcomings. For example: there is a problem of low true positive recall rate; there are files with high computing and resource requirements, and high requirements for computer hardware. Even after the model is trained, the deep learning model may require a lot of computing resources when performing inference, which may not be realistic in a laboratory environment with limited resources; in addition, although deep learning can automatically learn features, it sometimes has difficulty distinguishing between noise and real signals, especially in low signal / noise ratio data, which may lead to false positives. When background noise is recognized as a characteristic peak through the peak extraction algorithm standard, a false positive characteristic peak will be generated; excessive overestimation of the background baseline will also cause peak loss and lead to the generation of false negative peaks; these erroneous integrations often lead to abnormal presentation of some data, which in turn affects the biological judgment of researchers.

[0006] At the same time, the scientific problems solved by existing automatic integration methods are limited to how to reduce false negatives and false positives during peak extraction, and rarely consider the problem of integration trend consistency among multiple samples in a project. In practice, the problem of integration trend consistency plays a decisive role in studying the differences between substances under different sample conditions.

[0007] Based on this, the art urgently needs a chromatographic mass spectrometry automatic integration method, system, equipment and medium for multi-sample projects that can accurately perform peak shape identification and extraction, have good multi-sample integration consistency, and require little computing resources. Summary of the invention

[0008] Based on the above description, the present invention provides a method, system, device and medium for automatic integration of chromatographic mass spectrometry for multi-sample projects, which automatically identifies the peak shape of the chromatographic mass spectrometry using a deep learning model, and optimizes the peak selection for multi-sample projects through methods such as peak clustering, peak recall, and peak inconsistency correction, and can obtain qualitative and quantitative results of chromatographic peaks with high accuracy, high efficiency and low resource consumption.

[0009] The technical solution of the present invention to solve the above technical problems is as follows: In a first aspect, the present invention provides a method for automatic integration of chromatogram mass spectra for a multi-sample project, comprising the following steps: S1. Convert the original spectrum of the chromatogram mass spectrum to generate a standardized spectrum image; S2. Use the trained deep learning peak recognition algorithm to perform peak recognition on the standardized spectrum image to obtain candidate peaks and candidate peak classification results; S3, screening the candidate peak classification results based on the spectrum classification results, removing the peak results whose prediction results are noise in the candidate peak classification results, and obtaining a preliminary integration result; S4, performing peak clustering on multiple sample items based on the preliminary peak integration results; S5, performing peak recall on samples with unretained peaks after peak clustering; S6. Perform consistency correction on all retained peaks to make the integrated peak width and peak position of the same substance consistent in all samples.

[0010] Based on the above technical solution, the present invention can also be improved as follows.

[0011] Furthermore, step S1 specifically includes: For each mass spectrum channel of the original chromatographic mass spectrum, a spectrum is plotted with time as the horizontal axis and mass spectrum signal intensity as the vertical axis, and the spectra of all the mass spectrum channels constitute the standardized spectrum image.

[0012] Furthermore, the deep learning peak recognition algorithm in step S2 is an improved YOLOv8 algorithm, which is based on the standardized spectrum image and manually annotated peak positions as labels for training to obtain a trained deep learning model.

[0013] Furthermore, step S2 specifically includes the following steps: S201, obtaining a data set; the data set includes a plurality of standardized spectrum images for training and a label corresponding to each standardized spectrum image for training; the label is the start and end positions of all peaks in the standardized spectrum image for training, and the type of each peak shape; S202, building a deep learning model; S203, dividing the data set into a training set, a validation set and a test set; S204, using the training set as input, training the deep learning model to obtain a trained model; S205, using the validation set to evaluate the model performance, and adjusting the hyperparameters according to the results to obtain an optimized model; S206, determining whether the preset number of iterations has been reached; If yes, the test set is used as input to calculate the prediction effect of the optimized model, and the best performing model is selected as the final trained deep learning model; If not, the optimized model is used as the basis for the next round of iteration and the process returns to step S204.

[0014] Furthermore, step S202 specifically includes: Input image: resize the input image to a fixed size and perform normalization; Feature extraction: extract multi-level features of the image through the backbone network; Feature fusion: Multi-scale features are fused through feature pyramid network; Object detection: The detection head outputs bounding boxes, categories, and confidence scores; Post-processing: Remove redundant detection frames through non-maximum suppression to obtain the detection results.

[0015] Furthermore, step S5 specifically includes the following steps: S501, extracting the original signals of all substances without retaining the peak signal samples, finding all the local maximum values ​​in the original signals, and retaining the local maximum values ​​near the average RT retained in step S4; S502, starting from the local maximum point as the original left and right boundaries, extending to the points on both sides until the edge point signal strength no longer decreases, updating the new left and right boundaries, and calculating their S / N, area, and width information; S503, using the Savgol Filter function to fit a polynomial to a local window of the data to smooth the signal; and then analyzing the smoothed signal to obtain peak information; S504, filtering the peak signals obtained in step S502 and step S503 according to the peak height, peak width, and S / N value conditions to select qualified signals; if there are multiple qualified signals, the peak signal with the largest area is selected as the final result and added to the result of peak clustering in step S4.

[0016] Further, step S6 specifically includes the following steps: S601, cyclically extract all sample peak information corresponding to each substance, calculate the median of the left boundary and the median of the right boundary of all samples; filter out all difference peak information of the left and right boundaries of all samples whose difference with the median is greater than a set threshold; S602, cyclically extracting difference peak information, finding all local minimum values ​​in the original signal where the difference peak is located, and sorting the indexes corresponding to these values ​​from small to large according to the signal strength; S603, traverse all combinations of local minimum points in sequence, calculate the difference between the peak signal boundary formed by them and the median of the left boundary and the median of the right boundary of the peak in all samples, and select the local minimum point combination with the smallest difference as the new signal boundary.

[0017] In a second aspect, the present invention provides a chromatogram mass spectrometry automatic integration system for multiple sample projects, comprising: A conversion module is used to convert the original spectrum of the chromatogram and mass spectrum to generate a standardized spectrum image; A peak recognition module is used to perform preliminary peak recognition on the standardized spectrum image using a trained deep learning peak recognition algorithm to obtain candidate peaks and candidate peak classification results; A peak screening module, used to screen the candidate peak results based on the spectrum classification results, remove the peak results whose prediction results are noise in the candidate peak classification results, and obtain automatic integration results; A peak clustering module, for clustering peaks at the same position in the peak detection results based on the preliminary peak integration results, and selecting the optimal category based on the RT error, peak rate, and S / N value conditions; The peak recall module is used to perform peak recall on samples that are not retained in the peak clustering results; The consistency correction module is used to perform consistency correction on all retained peaks so that the integrated peak width and peak position of the same substance in all samples are consistent.

[0018] In a third aspect, the present invention provides a chromatographic mass spectrometry automatic integration system for multiple sample projects, comprising: Processor; and a memory in which computer-readable program instructions are stored; Wherein, when the computer-readable program instructions are executed by the processor, the method for automatic integration of chromatographic mass spectra for multiple sample projects as described in the first aspect is executed.

[0019] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the steps of the method for automatic integration of chromatographic mass spectrometry for a multi-sample project as described in the first aspect.

[0020] Compared with the prior art, the technical solution of this application has the following beneficial technical effects: The present application provides a method, system, device and medium for automatic integration of chromatogram mass spectrometry for multi-sample projects. The first step is to convert the original chromatogram mass spectrometry to obtain a standardized spectrum image; the second step is to use a deep learning peak recognition algorithm to perform preliminary peak recognition on the standardized spectrum image to obtain candidate peaks and candidate peak classification results; the third step is to screen the candidate peak results based on the spectrum classification results, remove the peak results whose predicted results are noise in the candidate peak results, and obtain preliminary automatic integration results; the fourth step is to cluster the peaks at the same position in the peak detection results, and select the optimal category based on the RT error, peak rate, S / N value and other conditions; the fifth step is to recall the peaks of the samples that are not retained in the peak clustering results, and supplement a part of the blank results based on the idea of ​​local maximum and two-side extended exploration detection; the last step is to perform consistency correction on all existing retained peaks, so that the integrated peak width and peak position of the same substance in all samples are consistent, thereby enhancing the accuracy of qualitative and quantitative analysis.

[0021] By performing deep learning model prediction on the acquired chromatographic mass spectrometry images and combining multi-sample data for peak clustering, peak recall, and peak boundary inconsistency correction, it is possible to replace manual discrimination of the peak position and peak shape of the chromatographic mass spectrometry. This method has high accuracy and specificity, effectively reduces the false positives of conventional automatic integration methods, and is fully capable of efficiently realizing automated replacement of manual integration.

[0022] In summary, the present invention introduces self-developed algorithms such as peak clustering, peak recall, and peak consistency correction on the basis of peak detection, which makes up for the dilemma that the existing automatic integration program only focuses on the false negatives and false positives of a single peak integration, but does not pay attention to the situation between samples in the project; moreover, the degree of automation is higher and the applicability is wider. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] FIG1 is a method flow chart of a method for automatic integration of chromatogram mass spectra for multiple sample projects provided by Example 1 of the present invention; Figure 2 A method principle diagram of the chromatographic mass spectrometry automatic integration method for multiple sample projects provided by Example 1 of the present invention; Figure 3 This is a system block diagram of the chromatographic mass spectrometry automatic integration method for multiple sample projects provided in Example 2 of the present invention. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0025] The following is combined with Figures 1 to 3 The embodiments of the present invention are further described in detail with examples. The following examples are used to illustrate the present invention but cannot be used to limit the scope of the present invention.

[0026] The purpose of the present invention is to provide a method, system, device and medium for automatic integration of chromatogram mass spectrometry for multi-sample projects, which uses a deep learning model to automatically identify the peak shape of the chromatogram mass spectrometry, and can obtain quantitative results of the chromatogram mass spectrometry peaks with high accuracy and efficiency.

[0027] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0028] Example 1 like Figure 1 and Figure 2 As shown, an embodiment of the present invention provides a method for automatic integration of chromatogram mass spectra for multiple sample projects, comprising: Step S1, converting the original spectrum of the chromatographic mass spectrum to generate a standardized spectrum image.

[0029] Specifically, for each mass spectrum channel of the original chromatographic mass spectrum, a spectrum is drawn with time as the horizontal coordinate and mass spectrum signal intensity as the vertical coordinate, and the spectra of all the mass spectrum channels constitute the standardized spectrum picture.

[0030] That is, after the original chromatographic mass spectrometer is converted into a standardized spectrum image in this embodiment, the obtained standardized spectrum image includes multiple spectra, each spectrum corresponds to the data of a mass spectrum channel, the horizontal axis is time, and the vertical axis is the mass spectrum signal intensity.

[0031] Among them, the original chromatographic mass spectra may include: gas chromatography-single quadrupole mass spectrometry, gas chromatography-triple quadrupole mass spectrometry, gas chromatography-time-of-flight mass spectrometry, liquid chromatography-triple quadrupole mass spectrometry, liquid chromatography-time-of-flight mass spectrometry and liquid chromatography-orbitrap mass spectrometry.

[0032] As an optional implementation, before step S1, this embodiment can also convert the data format of the chromatogram mass spectrum original spectrum into a universal mass spectrum data format, such as mzML, mzXML, mgf, etc., through the instrument's own software or MSConvert (version V3.0.19243), and use the converted spectrum with the universal mass spectrum data format as the chromatogram mass spectrum original spectrum to perform step S1. By converting the data format of the chromatogram mass spectrum original spectrum, the method of this embodiment can be applied to chromatogram mass spectrum original spectra in different data formats provided by different instrument manufacturers, and has a wide range of applications.

[0033] Before step S1, this embodiment can also read the original chromatographic mass spectrometer and extract the spectral information therein, which may include the parent ion mass number, the daughter ion mass number, the retention time, the signal intensity, etc., and execute step S1 based on the spectral information.

[0034] Step S2: Use the trained deep learning peak recognition algorithm to perform peak recognition on the standardized spectrum image to obtain candidate peaks and candidate peak classification results.

[0035] The spectrum classification results include the type of each peak in the standardized spectrum image; the types include noise and normal.

[0036] The peak recognition algorithm is an improved YOLOv8 algorithm. A self-made training set is used to obtain a trained deep learning model. This step may specifically include: Step S201: Obtain a data set.

[0037] The data set includes a plurality of standardized spectrum images for training and a label corresponding to each standardized spectrum image for training; the label is the start and end positions of all peaks in the standardized spectrum image for training, and the type of each peak shape; In this embodiment, a manual labeling method is used to label the peak start, end point and category of each training standardized spectrogram image to obtain the label of the training standardized spectrogram image. All training standardized spectrogram images and the label of each training standardized spectrogram image constitute a data set. Specifically, the data set used in the training process can be a manually labeled actual project standardized spectrogram image, the number of which is 400,000.

[0038] Step S202: construct a deep learning model.

[0039] The deep learning model in this embodiment is YOLOv8, which is the latest generation of target detection model in the YOLO series and is an efficient and real-time target detection method. It is further optimized on the basis of the YOLO series, inheriting the speed advantage of the YOLO series and improving the detection accuracy.

[0040] YOLOv8 is suitable for various computing resource scenarios, from edge devices to servers, and can provide excellent performance. The workflow is as follows: Input image: The input image is resized and normalized.

[0041] Feature extraction: Extract multi-level features of the image through the backbone network.

[0042] Feature fusion: Multi-scale features are fused through the feature pyramid network.

[0043] Object Detection: The detection head outputs bounding boxes, categories, and confidence scores.

[0044] Post-processing: Remove redundant detection boxes through non-maximum suppression (NMS) to obtain the final detection result.

[0045] It should be noted that the embodiment of the present invention introduces the YOLO model for the first time in the field of metabolomics, and gives full play to the advantages of YOLOv8's high precision, fast inference speed, and easy integration and deployment in the automated detection of chromatographic mass spectrometry peaks. Compared with other automatic integration software based on deep learning, the integration program developed by the present invention has a stronger peak shape judgment accuracy and consumes less time under the same CPU condition. For example, a project of peak_CF processing 12 samples requires about 30 hours in a CPU environment (peak detection only), while the automatic integration program in the present invention only takes about 20 minutes to complete the peak detection, and does not require the use of GPU resources, so the computing resource cost is significantly lower.

[0046] Step S203, divide the data set into a training set (70%), a validation set (15%) and a test set (15%); Step S204, use the training set as input to train the deep learning model to obtain a trained model.

[0047] For the standardized spectrogram image used for training, its essence is a two-dimensional matrix, which serves as the input matrix of the deep learning model.

[0048] During the training process, the model uses the SGD optimizer, the number of training epochs is 500, the initial training learning rate is 0.01, and the cosine learning rate scheduler is used to adjust the learning rate.

[0049] Step S205: Use the validation set to evaluate the model performance and adjust the hyperparameters based on the results to obtain an optimized model.

[0050] Step S206, determining whether the preset number of iterations has been reached; If so, the test set is used as input to calculate the prediction effect of the optimized model, and the best performing model is selected as the final trained deep learning model; If not, the optimized model is used as the basis for the next iteration, and the process returns to step S204. In this embodiment, the adjustable model hyperparameters mainly include learning rate, number of trainings, optimizer type, L2 regularization term, etc. The model is optimized by optimizing hyperparameters such as number of trainings, number of nodes, and batch size to obtain an optimized model.

[0051] The final model was obtained after evaluation using the test set, and the model with the best effect on the test set was selected as the trained deep learning model. During the experiment, the final accuracy of the model was: 93% for Class A, 95% for Class B, 86% for Class C, and 98% for Class D.

[0052] As an optional implementation, before performing subsequent analysis steps, the automatic integration method of this embodiment also includes: deduplicating and screening the initial candidate peak results obtained by the multi-parameter, multi-method integration method according to parameters such as peak height, signal-to-noise ratio and peak width, to obtain the peak results after preliminary screening, and use the peak results after preliminary screening as new candidate peak results, so as to perform preliminary screening on the candidate peak results first, thereby improving the screening efficiency.

[0053] Step S3, screening the candidate peak classification results based on the spectrum classification results, removing the peak results of the candidate peak classification results whose prediction results are noise (D), and obtaining a preliminary integration result.

[0054] Step S4: Based on the preliminary peak integration result, peak clustering is performed on the multi-sample items.

[0055] If there are multiple clusters, they are screened and judged according to the conditions to select the same peaks in different samples for subsequent qualitative and quantitative processes. The workflow is as follows; Step S401: First, the peak detection results of all samples corresponding to the same substance are screened out, and then the PAM (Partitioning Around Medoids) clustering algorithm is executed based on the pamk function (from the fpc package). This algorithm is an improvement of the K-means algorithm, which performs clustering based on the selection of "center points".

[0056] Step S402: If the clustering result contains multiple categories, a certain weight is taken according to the deviation of the average retention time (RT) of the peaks in the category from the theoretical RT, the number of peaks in all samples, and the average S / N value to obtain the retention score. The category with the lowest retention score and the average RT corresponding to the category are selected as the final retained result.

[0057] Step S5: perform peak recall on samples that have not retained peaks after peak clustering. This step is based on the fact that the deep learning peak recognition program in step S2 has a certain error rate. In order to ensure more accurate results, peak recall is performed. The specific process is as follows: Step S501: extract the original signals of all substances without retaining the peak signal samples, find out all the local maximum values ​​in the original signals, and retain the local maximum values ​​near the average RT retained in S4.

[0058] Step S502: Starting from the local maximum point as the original left and right boundaries, extend to the points on both sides until the edge point signal strength no longer decreases, update the new left and right boundaries, and calculate their S / N, area, width and other information.

[0059] Step S503: Use the Savgol Filter function to fit a polynomial to a local window of the data to smooth the signal, and then perform the operation of S5 on the smoothed signal to obtain peak information.

[0060] Step S504, filter the peak signals obtained in step S502 and step S503 according to conditions such as peak height, peak width, S / N value, and screen out qualified signals; if there are multiple qualified signals, select the peak signal with the largest area as the final result and add it to the result of peak clustering in step S4.

[0061] Step S6: Perform consistency correction on all existing retained peaks to make the integrated peak width and peak position of the same substance in all samples consistent, thereby enhancing the accuracy of qualitative and quantitative analysis. The specific process is as follows: Step S601, cyclically extract all sample peak information corresponding to each substance, calculate the median of the left boundary and the median of the right boundary of all samples, and then filter out all difference peak information whose difference between one or both of the left and right boundaries of all samples and the above median is greater than a set threshold.

[0062] Step S602: cyclically extract the difference peak information, find out all the local minimum values ​​in the original signal where the difference peak information is located, and sort the indexes corresponding to these values ​​from small to large according to the signal strength.

[0063] Step S603, traverse all combinations of local minimum points in sequence, calculate the difference between the peak signal boundary formed by them and the median of the left boundary and the median of the right boundary of the peak in all samples, and select the local minimum point combination with the smallest difference as the new signal boundary.

[0064] The automatic integration method provided in this embodiment can be divided into six parts. The first part converts the original chromatographic mass spectrometer to obtain a standardized spectrum image; the second part uses a deep learning peak recognition algorithm to perform preliminary peak recognition on the standardized spectrum image to obtain candidate peaks and candidate peak classification results; the third part screens the candidate peak results based on the spectrum classification results, removes the peaks predicted to be noise in the candidate peak results, and obtains preliminary automatic integration results; the fourth part clusters the peaks at the same position in the peak detection results, and selects the optimal category based on RT error, peak rate, S / N value and other conditions; the fifth part recalls the peaks that are not retained in the peak clustering results, and supplements a part of the blank results based on the idea of ​​local maximum and two-side extended exploration detection; the last part is used to perform consistency correction on all existing retained peaks, so that the integral peak width and peak position of the same substance in all samples are consistent, thereby enhancing the accuracy of qualitative and quantitative analysis.

[0065] The training set data in the chromatographic mass spectrometry automatic integration method for multiple sample projects based on the deep learning model creatively provided in this embodiment is complete and sufficient in quantity, suitable for metabolomics data generated under different samples and instruments, and a large number of training sets (40W) with different peak shapes are applied to the training of the model; in subsequent performance tests, the automatic integration program of the present invention is also perfectly suitable for different body fluids, tissues or organ samples from clinical and animals, as well as samples from different parts of various plant species, and can also process data with different peak shapes under different chromatographic conditions (reverse phase and hydrophilic chromatography, etc.), which shows that the automatic integration program in the present invention can process very complex metabolomics data. From the final results of a large number of validation data performance tests, the automatic integration tool of the present invention can achieve an accuracy of more than 95%, which is almost as good as that of human experts with professional skills, and is also better than other integration programs based on deep learning.

[0066] In summary, the present embodiment creatively provides a chromatographic mass spectrometry automatic integration method for multi-sample projects based on a deep learning model. By performing deep learning model prediction on the acquired chromatographic mass spectrometry images and combining multi-sample data for peak clustering, peak recall, and peak boundary inconsistency correction, it can replace manual discrimination of the peak position and peak shape of the chromatographic mass spectrometry. This method has high accuracy and specificity, effectively reduces the false positives of conventional automatic integration methods, and is fully capable of efficiently realizing automated replacement of manual integration.

[0067] Example 2 This embodiment is used to provide a chromatogram mass spectrometry automatic integration system for multiple sample projects, as shown in FIG3 , including: The conversion module M1 is used to convert the original chromatographic mass spectrometer to obtain a standardized spectrum image.

[0068] The peak recognition module M2 is used to perform preliminary peak recognition on the standardized spectrum image using a deep learning peak recognition algorithm to obtain candidate peaks and corresponding peak classification results.

[0069] The peak screening module M3 is used to screen the candidate peak results based on the spectrum classification results, remove the peak results whose predicted results are noise in the candidate peak results, and obtain the automatic integration results.

[0070] The peak clustering module M4 is used to cluster the peaks at the same position in the peak detection results, and select the optimal category based on the RT error, peak rate, S / N value and other conditions.

[0071] The peak recall module M5 is used to recall the samples that are not retained in the peak clustering results, and to supplement some blank results based on the idea of ​​local maximum and two-side extension exploration detection.

[0072] The consistency correction module M6 is used to perform consistency correction on all existing retention peaks, so that the integrated peak width and peak position of the same substance in all samples are consistent, thereby enhancing the accuracy of qualitative and quantitative analysis.

[0073] Embodiment 3: This embodiment is used to provide a chromatogram mass spectrum automatic integration device for multiple sample projects, including: a processor; and a memory storing computer readable program instructions, Wherein, when the computer-readable program instructions are executed by the processor, the chromatographic mass spectrometry automatic integration method for multiple sample projects as described in Example 1 is executed.

[0074] Embodiment 4: This embodiment is used to provide a computer-readable storage medium, on which a computer program is stored, characterized in that when the program is executed by a processor, the steps of the chromatographic mass spectrometry automatic integration method for a multi-sample project described in Example 1 are implemented.

[0075] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for automatic integration of chromatogram mass spectrometry for a multi-sample project, characterized in that: The steps include: S1. Convert the original spectrum of the chromatogram mass spectrum to generate a standardized spectrum image; S2. Use the trained deep learning peak recognition algorithm to perform peak recognition on the standardized spectrum image to obtain candidate peaks and candidate peak classification results; S3, screening the candidate peak classification results based on the spectrum classification results, removing the peak results whose prediction results are noise in the candidate peak classification results, and obtaining a preliminary integration result; S4, performing peak clustering on multiple sample items based on the preliminary peak integration results; S5, performing peak recall on samples with unretained peaks after peak clustering; S6. Perform consistency correction on all retained peaks to make the integrated peak width and peak position of the same substance consistent in all samples.

2. The method for automatic integration of chromatogram mass spectra for multiple sample projects according to claim 1, characterized in that: Step S1 specifically includes: For each mass spectrum channel of the original chromatographic mass spectrum, a spectrum is plotted with time as the horizontal axis and mass spectrum signal intensity as the vertical axis, and the spectra of all the mass spectrum channels constitute the standardized spectrum image.

3. The method for automatic integration of chromatogram mass spectra for multiple sample projects according to claim 2, characterized in that: The deep learning peak recognition algorithm in step S2 is an improved YOLOv8 algorithm, which is based on the standardized spectrum image and manually annotated peak positions as labels for training to obtain a trained deep learning model.

4. The method for automatic integration of chromatogram mass spectra for multiple sample projects according to claim 3, characterized in that: Step S2 specifically includes the following steps: S201, obtaining a data set; the data set includes a plurality of standardized spectrum images for training and a label corresponding to each standardized spectrum image for training; the label is the start and end positions of all peaks in the standardized spectrum image for training, and the type of each peak shape; S202, building a deep learning model; S203, dividing the data set into a training set, a validation set and a test set; S204, using the training set as input, training the deep learning model to obtain a trained model; S205. Use the validation set to evaluate the model performance and adjust the hyperparameters based on the results to obtain an optimized model. S206, determining whether a preset number of iterations has been reached; If so, the test set is used as input to calculate the prediction effect of the optimized model, and the best performing model is selected as the final trained deep learning model; If not, the optimized model is used as the basis for the next round of iteration and the process returns to step S204.

5. The method for automatic integration of chromatogram mass spectra for multiple sample projects according to claim 4, characterized in that: Step S202 specifically includes: Input image: resize the input image to a fixed size and perform normalization; Feature extraction: extract multi-level features of the image through the backbone network; Feature fusion: Multi-scale features are fused through feature pyramid network; Object detection: The detection head outputs bounding boxes, categories, and confidence scores; Post-processing: Remove redundant detection frames through non-maximum suppression to obtain the detection results.

6. The method for automatic integration of chromatogram mass spectra for multiple sample projects according to claim 1, characterized in that: Step S5 specifically includes the following steps: S501, extracting the original signals of all substances without retaining the peak signal samples, finding all the local maximum values ​​in the original signals, and retaining the local maximum values ​​near the average RT retained in step S4; S502, starting from the local maximum point as the original left and right boundaries, extending to the points on both sides until the edge point signal strength no longer decreases, updating the new left and right boundaries, and calculating their S / N, area, and width information; S503, using the Savgol Filter function to fit a polynomial to a local window of the data to smooth the signal; and then analyzing the smoothed signal to obtain peak information; S504, filtering the peak signals obtained in step S502 and step S503 according to the peak height, peak width, and S / N value conditions to select qualified signals; if there are multiple qualified signals, the peak signal with the largest area is selected as the final result and added to the result of peak clustering in step S4.

7. The method for automatic integration of chromatogram mass spectra for multiple sample projects according to claim 1, characterized in that: Step S6 specifically includes the following steps: S601, cyclically extracting all sample peak information corresponding to each substance, calculating the median of the left boundary and the median of the right boundary of all samples; screening out all difference peak information of the left and right boundaries of all samples whose difference with the median is greater than a set threshold; S602, cyclically extracting difference peak information, finding all local minimum values ​​in the original signal where the difference peak is located, and sorting the indexes corresponding to these values ​​from small to large according to the signal strength; S603, traverse all combinations of local minimum points in sequence, calculate the difference between the peak signal boundary formed by them and the median of the left boundary and the median of the right boundary of the peak in all samples, and select the local minimum point combination with the smallest difference as the new signal boundary.

8. A chromatographic mass spectrometry automatic integration system for a multi-sample project, characterized in that: include: A conversion module is used to convert the original spectrum of the chromatogram and mass spectrum to generate a standardized spectrum image; A peak recognition module is used to perform preliminary peak recognition on the standardized spectrum image using a trained deep learning peak recognition algorithm to obtain candidate peaks and candidate peak classification results; A peak screening module, used to screen the candidate peak results based on the spectrum classification results, remove the peak results whose prediction results are noise in the candidate peak classification results, and obtain automatic integration results; A peak clustering module, for clustering peaks at the same position in the peak detection results based on the preliminary peak integration results, and selecting the optimal category based on the RT error, peak rate, and S / N value conditions; The peak recall module is used to perform peak recall on samples that are not retained in the peak clustering results; The consistency correction module is used to perform consistency correction on all retained peaks so that the integrated peak width and peak position of the same substance in all samples are consistent.

9. A chromatographic mass spectrometry automatic integration device for a multi-sample project, characterized in that: include: processor; as well as a memory in which computer-readable program instructions are stored; Wherein, when the computer-readable program instructions are executed by the processor, the chromatographic mass spectrometry automatic integration method for a multi-sample project according to any one of claims 1 to 7 is executed.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the chromatographic mass spectrometry automatic integration method for a multi-sample project are implemented as claimed in any one of claims 1 to 7.

Images