A method and system for single cell metabolite qualification

Abstract

The application belongs to the technical field of single-cell metabolite detection, and specifically provides a single-cell metabolite qualitative method and system, the qualitative method comprising the following steps: A1. obtaining raw data of metabolites by using a detection module; A2. calculating similarity scores of the raw data and candidate data in a database by using a data analysis module; A3. arranging the candidate data in descending order from high to low according to the similarity scores, and cutting off the top 20% of candidate data with high confidence to form sequence 1, denoted as [candidate metabolites] i ; A4. re-arranging the candidate data in sequence 1 according to the biological rationality of metabolites by using a data arrangement module to form sequence 2, denoted as [candidate metabolites] j ; A5. qualitatively determining the metabolites as the candidate metabolites equal to min(i+j) . The application has the advantages of high metabolite qualitative accuracy, high efficiency, strong applicability and the like.

Description

Technical Field

[0001] This invention relates to single-cell metabolite detection technology, and particularly to a method and system for qualitative analysis of single-cell metabolites. Background Technology

[0002] The rapid development of qualitative detection technology for single-cell metabolites is fundamentally reshaping our understanding of life sciences. It overcomes the masking effect of the "average" in traditional population cell analysis, advancing research resolution to the microscopic dimension of individual cells. As a crucial bridge connecting cellular molecular maps and functional phenotypes, qualitative detection of single-cell metabolites is gradually moving from niche exploration to large-scale application. It not only allows scientists to glimpse the most subtle dynamic changes in life activities but also provides an unprecedentedly powerful tool for solving major medical challenges such as tumors and neurodegenerative diseases.

[0003] Because single-cell data is extremely large, manual detection and labeling are very slow. Therefore, how to use big data algorithms to conduct batch metabolite identification is a critical issue in the scientific community today. Summary of the Invention

[0004] To address the shortcomings of the existing technical solutions, this invention provides a method for qualitative analysis of single-cell metabolites.

[0005] The objective of this invention is achieved through the following technical solution: A method for qualitative analysis of single-cell metabolites, comprising the following steps: A1. Obtain raw data of metabolites using the detection module; A2. The data analysis module calculates a similarity score between the raw data and the candidate data in the database; A3. The data sorting module sorts the candidate data in descending order from high to low according to the similarity score, and extracts the top 20% of high-confidence candidate data to form sequence 1, denoted as [candidate metabolite]. i Wherein, candidate metabolites are the proprietary names of the metabolites corresponding to the candidate data. i The sequence number is the sequence number of the first sequence. i∈{1,2, ...,N} , N The number of the top 20% of the candidate data subsets; A4. The data sorting module performs a secondary sorting of the candidate data in Sequence 1 based on the biological plausibility of the metabolites, forming Sequence 2, denoted as [Candidate Metabolite]. j ， The candidate metabolite is the proprietary name of the metabolite corresponding to the candidate data. j The sequence number is the sequence number of the second sequence. j∈{1,2, ...,N} , NThe number of the top 20% of the candidate data subsets; A5. The metabolite qualitative module qualitatively identifies the metabolite as equal to... min(i+j) The candidate metabolites.

[0006] Compared with the prior art, the present invention has the following beneficial effects.

[0007] 1. High accuracy: Using raw metabolite data, the similarity score is calculated between the raw data and candidate data in the database. The candidate data are sorted in descending order from high to low according to the similarity score. The top 20% of high-confidence candidate data are selected to form Sequence 1. The candidate data in Sequence 1 are then sorted again according to the biological rationality of the metabolite to form Sequence 2. The correlation between the two sequences is used to comprehensively determine the metabolite. This method has high accuracy.

[0008] 2. High efficiency: It provides a method and device for the qualitative identification of metabolites, which can directly obtain the metabolite property category by data processing after detection by the detection device, without the need for time-consuming and laborious manual judgment and analysis based on mass spectra, spectra, chromatograms, etc. to obtain the metabolite property category.

[0009] 3. High applicability: The provided metabolite qualitative methods can be applied to the processing of spectral data from mass spectrometers, spectrometers, chromatographs, etc., as well as the qualitative analysis of metabolites. Attached Figure Description

[0010] The disclosure of this invention will become more readily understood with reference to the accompanying drawings. It will be readily understood by those skilled in the art that these drawings are merely illustrative of the technical solutions of this invention and are not intended to limit the scope of protection of this invention. In the drawings: Picture 1 This is a flowchart of a method for qualitative analysis of metabolites from single cells. Detailed Implementation

[0011] Picture 1 The following description illustrates optional embodiments of the invention to teach those skilled in the art how to implement and reproduce the invention. Some conventional aspects have been simplified or omitted to teach the technical solutions of the invention. Those skilled in the art should understand that variations or substitutions derived from these embodiments will be within the scope of the invention. Those skilled in the art should understand that the following features can be combined in various ways to form multiple variations of the invention. Therefore, the invention is not limited to the optional embodiments described below, but is defined only by the claims and their equivalents.

[0012] Example 1

[0013] The single-cell metabolite qualitative method of this embodiment, such as Picture 1 As shown, the steps include: A1. Obtain raw data of metabolites using the detection module; A2. The data analysis module calculates a similarity score between the raw data and the candidate data in the database; Currently, commonly used scoring metrics mainly include dot product, cosine similarity, matching factor, and weighted cosine similarity. Taking dot product as an example, this method represents the experimental spectrum and the reference spectrum as vectors composed of fragment ion intensities, and quantifies their consistency by calculating the inner product of the two vectors. To improve the robustness of matching, many algorithms also introduce preprocessing strategies such as peak alignment, mass error weighting, and neutral loss matching to reduce interference from instrument bias and fragment differences.

[0014] In this embodiment, the similarity score is based on an improved 679-bit molecular fingerprint. For each successfully extracted metabolite, three complementary molecular fingerprints are calculated using the RDKit toolkit: the Morgan fingerprint (256 bits, radius 2, corresponding to ECFP4), also known as the circular fingerprint, is a fingerprint type that captures local molecular environment information by iteratively expanding the radius of the atomic neighborhood. The generation process of this fingerprint follows the Morgan algorithm: starting from each non-hydrogen atom, it gradually collects atomic information in the neighborhood with increasing radius, and maps this information to a fixed-length bit vector through a hash function. When the radius parameter is set to 2, the Morgan fingerprint can capture the structural environment within a range of 2 chemical bonds around each atom, which is equivalent to the ECFP4 (Extended Connectivity Fingerprint of diameter 4) fingerprint widely used in the pharmaceutical industry in terms of information coverage. This study uses a 256-bit compressed representation to control the feature dimension while maintaining sufficient structural recognition power, avoiding the risk of overfitting caused by excessive dimensionality. The overall workflow of the molecular fingerprint is as follows: in the training phase, the preprocessed spectral data is trained through the model to learn the correspondence between the spectrum and the molecular fingerprint. In the prediction phase, biological samples are first obtained, and then experimental spectra are obtained through experiments. The spectra are then fed into a trained model to predict the molecular fingerprint of the metabolite. Candidate metabolites are then selected from the database based on the precise mass measured in the experiment, and the molecular fingerprints of these candidate metabolites are calculated. Finally, the similarity score between the predicted molecular fingerprint and the molecular fingerprints of all candidate metabolites is calculated.

[0015] A3. The data sorting module sorts the candidate data in descending order from high to low according to the similarity score, and extracts the top 20% of high-confidence candidate data to form sequence 1, denoted as [candidate metabolite]. i Wherein, candidate metabolites are the proprietary names of the metabolites corresponding to the candidate data. i The sequence number is the sequence number of the first sequence. i∈{1,2, ...,N} , NThe number of the top 20% of the candidate data subsets; A4. The data sorting module performs a secondary sorting of the candidate data in Sequence 1 based on the biological plausibility of the metabolites, forming Sequence 2, denoted as [Candidate Metabolite]. j ， The candidate metabolite is the proprietary name of the metabolite corresponding to the candidate data. j The sequence number is the sequence number of the second sequence. j∈{1,2, ...,N} , N The number of the top 20% of the candidate data subset.

[0016] Using the biological rationality index W Sort in descending order from highest to lowest: W=β 1 ⋅C 1 +β 2 ⋅C 2 +β 3 ⋅C 3 ,in: W Biological rationality index; β 1 , β 2 , β 3 To preset weights, β 1 =0.1~0.4，β 2 =0.3~0.7，β 3 =0.1~0.3，β 1 + β 2 + β 3 =1； C 1 This is the abundance stability coefficient. C 2 The pathway correlation coefficient. C 3 This is the chemical stability coefficient.

[0017] Abundance stability coefficient C 1 The abundance stability coefficient is defined as follows: if the deviation rate between the expression value of a metabolite in a single cell and the median of the cell population is greater than the threshold T, then the abundance stability coefficient is... C 1 =0.5, otherwise C 1 =1; where the threshold T > 1.5.

[0018] Pathway correlation coefficient C 2 If the metabolite belongs to the core pathway of the single cell and there is a clear expression record for that cell type, then C 2 =1; If the metabolite belongs to the relevant pathway of the single cell but is not the core, then C 2 =0.8; if there is no clear cell type-specific record of the metabolite, then C 2 =0.6.

[0019] Chemical stability coefficient C 3 Let: If the half-life t of the metabolite is 1 / 2 If it takes ≥1 min and is easily ionized, then C 3 =1; if the half-life of the metabolite is 1s ≤ t 1 / 2 <1min and can be ionized, then C 3 =0.6; if the half-life t of the metabolite is... 1 / 2 If the result is less than 1 second or extremely difficult to detect, then C 3 =0.1.

[0020] A5. The metabolite qualitative module qualitatively identifies the metabolite as equal to... min(i+j) The candidate metabolites.

[0021] If multiple identical values ​​exist min(i+j) Then filter i The top 10% of the candidate data, and among them j The candidate metabolite with the smallest value is selected as the final metabolite.

[0022] The database updates the original data and metabolite category of each metabolite in real time based on each qualitative result.

[0023] Example 2

[0024] A single-cell metabolite qualitative system, comprising: The detection module is used to detect raw data of metabolites; the detection module includes mass spectrometers, spectrometers, chromatographs, etc., as well as their coupled instruments, such as chromatography-mass spectrometry.

[0025] The database contains information such as known single-cell metabolite profiles and metabolite names, and updates the original data and metabolite categories of the current metabolite to the original database in real time based on the previous qualitative results.

[0026] The data analysis module acquires the raw data output by the detection module and performs a similarity score with the candidate data in the database.

[0027] The data sorting module receives candidate data and similarity scores from the data analysis module. Based on the similarity scores, it sorts the candidate data in descending order from highest to lowest confidence and extracts the top 20% of high-confidence candidate data to form sequence 1, denoted as [candidate metabolite]. i Wherein, candidate metabolites are the proprietary names of the metabolites corresponding to the candidate data. i The sequence number is the sequence number of the first sequence. i∈{1,2, ...,N} , N The number of the top 20% of the candidate data subsets; The data sorting module performs a secondary sorting of the candidate data in Sequence 1 based on the biological plausibility of the metabolites, forming Sequence 2, denoted as [Candidate Metabolite]. j ， The candidate metabolite is the proprietary name of the metabolite corresponding to the candidate data. j The sequence number is the sequence number of the second sequence. j∈{1,2, ...,N} , N The number of the top 20% of the candidate data subsets; The metabolite qualitative module receives Sequence 1 and Sequence 2 output from the data sorting module and processes them... i+j The smallest candidate metabolite is selected as the final metabolite; if multiple metabolites with the same value exist... min(i+j) Then filter i For the top 10% of candidate data, and among them... j The candidate metabolite with the smallest value is selected as the final metabolite.

[0028] Example 3

[0029] Unlike Examples 1 or 2, the detection module is a mass spectrometer or a liquid chromatography-mass spectrometry (LC-MS) system. This application example is for the detection of pyruvate, a metabolite found in mouse oocytes.

[0030] Raw mass spectra of mouse oocyte metabolites were obtained using mass spectrometry or liquid chromatography-mass spectrometry. The data analysis module calculated similarity scores between the raw data and candidate data from the database using a 679-bit molecular fingerprint algorithm. The molecular fingerprint algorithm model takes preprocessed mass spectrometry vectors and metadata as input and outputs a 679-bit binary vector, which is composed of three fingerprints: Morgan (256-bit), RDKit topology (256-bit), and MACCS (167-bit). The specific evaluation process is as follows: First, test sample preparation: a compound is selected from the test set, whose real information includes: precursor ion mass-to-charge ratio, ion mode, and corresponding molecular fingerprint (679-bit binary vector) and InChIKey identifier. Second, fingerprint prediction: the compound's mass spectrometry data and metadata are input into the trained model to obtain the predicted molecular fingerprint vector Fpred. Finally, a similarity score is calculated between the compound and candidate data from the database.

[0031] The database is constructed such that each candidate data point includes its true molecular fingerprint, precursor ion mass-to-charge ratio, ion mode, and InChIKey identifier. For candidate data filtering, the raw mass spectra of the test compounds are introduced to first obtain their precursor ion mass-to-charge ratio and a set mass error tolerance ∆m / z (expressed in parts per million, ppm). The database is then filtered, retaining only candidate data whose precursor ion mass-to-charge ratio error with the candidate data's precursor ion mass-to-charge ratio is within the set mass error tolerance ∆m / z.

[0032] The data sorting module sorts the candidate data in descending order based on similarity scores, and extracts the top 20% of high-confidence candidate data to form sequence 1, denoted as [candidate metabolite]. i Wherein, candidate metabolites are the proprietary names of the metabolites corresponding to the candidate data. i The sequence number is the sequence number of the first sequence. i∈{1,2, ...,N} , N This represents the number of the top 20% of the candidate data subset. The sequence 1 obtained in this experiment is: {[pyruvate]1, [α-ketoglutarate]2, [lactic acid]3}.

[0033] The data sorting module performs a secondary sorting of the candidate data in Sequence 1 based on the biological plausibility of the metabolites, forming Sequence 2, denoted as [Candidate Metabolite]. j ， The candidate metabolite is the proprietary name of the metabolite corresponding to the candidate data. j The sequence number is the sequence number of the second sequence. j∈{1,2, ...,N} , N The number of the top 20% of the candidate data subsets. Utilizing the biological plausibility index. W Sort them in descending order from highest to lowest.W=β 1 ⋅C 1 +β 2 ⋅C 2 +β 3 ⋅C 3 ,in: W Biological rationality index; β 1 , β 2 , β 3 To preset weights, β 1 =0.3，β 2 =0.5，β 3 =0.2，β 1 + β 2 + β 3 =1；C 1 This is the abundance stability coefficient. C 2 The pathway correlation coefficient. C 3 Chemical stability coefficient. Abundance stability coefficient. C 1 The abundance stability coefficient is defined as follows: if the deviation rate between the expression value of a metabolite in a single cell and the median of the cell population is greater than the threshold T, then the abundance stability coefficient is... C 1 =0.5, otherwise C 1 =1; where the threshold T > 1.5. Pathway correlation coefficient C 2 If the metabolite belongs to the core pathway of the single cell and there is a clear expression record for that cell type, then C 2 =1; If the metabolite belongs to the relevant pathway of the single cell but is not the core, then C 2 =0.8; if there is no clear cell type-specific record of the metabolite, then C 2 =0.6. Chemical stability coefficient. C 3 Let: If the half-life t of the metabolite is 1 / 2 If it takes ≥1 min and is easily ionized, then C 3 =1; if the half-life of the metabolite is 1s ≤ t 1 / 2<1min and can be ionized, then C 3 =0.6; if the half-life t of the metabolite is... 1 / 2 If the result is less than 1 second or extremely difficult to detect, then C 3 =0.1. The experimental sequence 2 obtained in this study is: {[pyruvate]1, [lactic acid]2, [α-ketoglutarate]3}.

[0034] The metabolite qualitative module calculated [pyruvate] i+j= 2 represents the minimum value, meaning the metabolite detected in this study was identified as pyruvate.

[0035] The database updates the original data and metabolite categories of pyruvate metabolites in real time based on the qualitative results.

Claims

1. A method for qualitative analysis of metabolites in single cells, characterized in that, The single-cell metabolite qualitative method includes the following steps: A1. Obtain raw data of metabolites using the detection module; A2. The data analysis module calculates a similarity score between the raw data and the candidate data in the database; A3. The data sorting module sorts the candidate data in descending order from high to low according to the similarity score, and extracts the top 20% of high-confidence candidate data to form sequence 1, denoted as [candidate metabolite]. i Wherein, candidate metabolites are the proprietary names of the metabolites corresponding to the candidate data. i The sequence number is the sequence number of the first sequence. i∈{1,2,…,N} , N The number of the top 20% of the candidate data subsets; A4. The data sorting module performs a secondary sorting of the candidate data in Sequence 1 based on the biological plausibility of the metabolites, forming Sequence 2, denoted as [Candidate Metabolite]. j ， The candidate metabolite is the proprietary name of the metabolite corresponding to the candidate data. j The sequence number is the sequence number of the second sequence. j∈{1,2,…,N} , N The number of the top 20% of the candidate data subsets; A5. The metabolite qualification module qualifies the metabolite as equal to min(i+j) the candidate metabolite.

2. The method for qualitative analysis of single-cell metabolites according to claim 1, characterized in that, In step A5: If multiple identical values ​​exist min(i+j) Then filter i The top 10% of the candidate data, and among them j The candidate metabolite with the smallest value is selected as the final metabolite.

3. The method for qualitative analysis of single-cell metabolites according to claim 1, characterized in that, The similarity score is based on weighted cosine similarity, or molecular fingerprint, and scores are given for peak-shaped features.

4. The method for qualitative analysis of single-cell metabolites according to claim 1, characterized in that, The secondary sorting is as follows: Using the biological rationality index W Sort them in descending order from highest to lowest; W=β 1 ⋅C 1 +β 2 ⋅C 2 +β 3 ⋅C 3 ,in: W For biological rationality index; β 1 , β 2 , β 3 To preset weights, β 1 = 0.1~0.4 ，β 2 = 0.3~0.7 ，β 3 = 0.1~0.3 ，β 1 + β 2 + β 3 = 1 ；C 1 This is the abundance stability coefficient. C 2 The pathway correlation coefficient. C 3 This is the chemical stability coefficient.

5. The method for qualitative analysis of single-cell metabolites according to claim 4, characterized in that, The abundance stability coefficient C 1 for: If the deviation rate between the expression value of the metabolite in a single cell and the median of the cell population is greater than the threshold T, then the abundance stability coefficient... C 1 =0.5, otherwise C 1 =1; where the threshold T > 1.

5.

6. The method for qualitative analysis of single-cell metabolites according to claim 4, characterized in that, The pathway correlation coefficient C 2 for: If the metabolite belongs to the core pathway of the single cell and there is a clear expression record for that cell type, then C 2 =1; If the metabolite belongs to a relevant pathway in the single cell but is not the core pathway, then C 2 =0.8; If the metabolite does not have a clearly defined cell type-specific record, then C 2 =0.

6.

7. The method for qualitative analysis of single-cell metabolites according to claim 4, characterized in that, The chemical stability coefficient C 3 for: If the half-life t of the metabolite 1 / 2 If it takes ≥1 min and is easily ionized, then C 3 =1; If the half-life of the metabolite is 1s≤t 1 / 2 <1min and can be ionized, then C 3 =0.6; If the half-life t of the metabolite 1 / 2 If the result is less than 1 second or extremely difficult to detect, then C 3 =0.

1.

8. The method for qualitative analysis of single-cell metabolites according to claim 1, characterized in that, The database is updated in real time based on each qualitative result.

9. A single-cell metabolite qualitative system, characterized in that, The single-cell metabolite qualitative system includes: A detection module, which is used to detect raw data of metabolites; A database containing known single-cell metabolite data as candidate data, which is updated in real time; The data analysis module acquires the raw data output by the detection module and performs a similarity score with the candidate data in the database. The data sorting module receives the candidate data and similarity scores from the data analysis module, sorts the candidate data in descending order based on the similarity scores, and extracts the top 20% of high-confidence candidate data to form Sequence 1, denoted as [Candidate Metabolite]. i Wherein, candidate metabolites are the proprietary names of the metabolites corresponding to the candidate data. i The sequence number is the sequence number of the first sequence. i∈{1,2,…,N} , N The number of the top 20% of the candidate data subsets; The data sorting module performs a secondary sorting of the candidate data in Sequence 1 based on the biological plausibility of the metabolites, forming Sequence 2, denoted as [Candidate Metabolite]. j ， The candidate metabolite is the proprietary name of the metabolite corresponding to the candidate data. j The sequence number is the sequence number of the second sequence. j∈{1,2,…,N} , N The number of the top 20% of the candidate data subsets; The metabolite identification module receives Sequence 1 and Sequence 2 output by the data sorting module, and identifies the metabolite as equal to... min(i+j) The candidate metabolites; If multiple identical values ​​exist min(i+j) Then filter i The top 10% of the candidate data, and among them j The candidate metabolite with the smallest value is selected as the final metabolite.

10. The single-cell metabolite qualitative system according to claim 9, characterized in that, The detection module includes a mass spectrometer, a spectrometer, a chromatograph, and a combination thereof.

Images