Method for constructing and qualitatively detecting chromatographic retention time prediction model of oxysterol

Abstract

The invention provides a construction and qualitative detection method of an oxysterol chromatographic retention time prediction model, and belongs to the technical field of analysis and detection. The method comprises the following steps: (1) constructing a theoretical database of oxysterol; (2) calculating a descriptor of each oxidized sterol molecule in the theoretical database by using E-Dragon and RDKit; (3) determining the retention time of the oxidized sterol according to the descriptor, correcting the retention time of the oxidized sterol by using an interior label, and taking an ln value of the relative retention time to obtain a data set; (4) combining the obtained data set with a regression algorithm by adopting a random division mode of 80: 20 to jointly construct a model; and (5) obtaining an optimal hyper-parameter combination by adopting a grid search method and five-fold cross validation, and optimizing the jointly constructed model to obtain an oxysterol chromatographic retention time prediction model.

Description

Technical Field

[0001] The invention belongs to the technical field of analysis and detection, and in particular relates to a method for constructing and qualitatively analyzing an oxysterol chromatographic retention time prediction model. Background Art

[0002] Cholesterol is a typical sterol. As an important component of the cell membrane of the entire animal kingdom, it is composed of 27 carbon atoms. Its molecular skeleton includes a ring system and a side chain. Its ring system is composed of a "cyclopentaphenanthrene" parent nucleus, which is composed of three six-carbon cyclohexanes (A, B, C) and a five-carbon ring (D). It is a fused four-ring compound. The side chain is a carbon chain composed of 8 carbon atoms connected by carbon atom No. 17. The structural diagram is shown in the figure below. Figure 1 As shown. The metabolism of cholesterol is first converted into oxysterols through oxidation, and its precursors can also be oxidized into oxysterols. Therefore, oxysterols constitute a large family of molecules, and their structural characteristics are the introduction of one or more additional oxygen-containing functional groups on the ring system or side chain of cholesterol or its precursors. These oxygen-containing functional groups include hydroxyl, keto, epoxy, etc. The naming of oxysterols follows the IUPAC nomenclature, and the schematic diagram of their carbon atom numbering is shown in Figure 2 (Take 4β-Hydroxycholesterol as an example). When a substituent such as hydroxyl is substituted on carbon atom No. 4, the compound is named 4β-Hydroxycholesterol. Oxysterols can be divided into hydroxyl oxysterols, keto oxysterols, epoxy oxysterols, and unsubstituted oxysterols according to the type of oxygen-containing functional groups and the substitution conditions.

[0003] Due to the differences in the number of oxygen-containing functional groups and substitution positions in the structure of oxysterols, there are a large number of isomers in the oxysterol group. In conventional qualitative methods, the experiment mainly relies on the retention time of the standard, supplemented by MS / MS information comparison for qualitative analysis. For oxysterols that lack standards, if the retention time prediction can be achieved, it can provide an additional dimension of information for the qualitative analysis of oxysterols, thereby reducing false positive results and improving the accuracy of identification. Therefore, it is crucial to separate oxysterols at the chromatographic level and determine the retention time of each oxysterol at the chromatographic level.

[0004] Oxysterols, as important products of cholesterol metabolism, are derived from cholesterol and its precursors through different enzymatic or non-enzymatic reactions. Oxysterols have important and abundant biological functions in organisms and play an important role in maintaining metabolic balance, regulating immune response and neural function. In addition, oxysterols are believed to play a role in various diseases and have the potential to serve as disease biomarkers, including atherosclerosis, neurodegenerative diseases, immune system diseases and cancer.

[0005] At present, the detection methods of oxysterols have many problems such as low coverage and poor sensitivity, which limit the further research. In view of the limited commercial standards and the unclear distribution of oxysterols in organisms, it is urgent to establish a high-coverage, stable and reliable qualitative analysis method for oxysterols, so as to provide technical support for the research of oxysterols. Summary of the invention

[0006] In view of this, the purpose of the present invention is to provide a method for constructing and qualitatively predicting the chromatographic retention time of oxidized sterols. Based on the correlation between the structure of oxidized sterols and the chromatographic retention behavior, the present invention has developed an accurate metabolite determination method based on the LC-MS / MS system and based on some oxidized sterol standards. Based on the chemical structure of the existing standards of oxidized sterols and their retention time in the chromatographic system, a QSRR retention prediction model is constructed to predict the chromatographic retention time; based on the cleavage fragments of the existing standards under mass spectrometry conditions, the possible mass spectrometry cleavage characteristic fragments of different oxidized sterols are inferred. Finally, based on the predicted retention time, supplemented by the mass spectrometry cleavage characteristic fragments, the qualitative analysis of oxidized sterols in the absence of standards is achieved.

[0007] In order to achieve the above-mentioned object of the invention, the present invention provides the following technical solutions:

[0008] The present invention provides a method for constructing an oxysterol chromatographic retention time prediction model, comprising the following steps:

[0009] (1) Construct a theoretical database of oxysterols;

[0010] (2) using E-Dragon and RDKit to calculate the descriptors of each oxysterol molecule in the theoretical database;

[0011] (3) determining the retention time of oxysterols according to the descriptors, using an internal standard to calibrate the retention time of oxysterols, and taking the ln value of the relative retention time to obtain a data set;

[0012] (4) For the obtained data set, a random partitioning method of 80:20 is used in combination with the regression algorithm to jointly construct the model;

[0013] (5) A grid search method and five-fold cross-validation were used to obtain the best hyperparameter combination, and the jointly constructed model was optimized to obtain a prediction model for the chromatographic retention time of oxysterols.

[0014] Preferably, the theoretical database of oxysterols is obtained by collecting all oxysterols reported in the database; the database includes lipidmaps, pubchem, Pubmed, and sci-finder.

[0015] Preferably, the descriptors include lipid solubility, number of hydrogen bonds, distance between oxygen atoms, topological polar surface area, number of double bonds, number of oxygen atoms, strength of intramolecular hydrogen bonds and molecular thermodynamic properties.

[0016] Preferably, the regression algorithm is an ElasticNet regression algorithm.

[0017] The present invention provides a method for qualitative detection of oxysterols, comprising the following steps:

[0018] (5.1) Constructing MS database of oxysterol molecules;

[0019] (5.2) Use Ful MS mode and PRM mode to obtain the molecular ion peaks and secondary mass spectrometry fragmentation information of oxysterol standards, and summarize the additive form and secondary mass spectrometry fragmentation rules of oxysterols;

[0020] (5.3) using full scan mode to obtain the molecular ion peak of the oxysterol in the sample, and using PRM mode to obtain the secondary mass spectrometry fragmentation information of the oxysterol in the sample, and combining the additive form law and the secondary mass spectrometry mass spectrometry fragmentation law obtained in step (5.2) to obtain the possible mass spectrometry fragmentation characteristic fragments of the oxysterol in the sample;

[0021] (5.4) Based on the chromatographic retention time prediction model of oxysterols obtained in claim 1, the mass spectrometry fragmentation characteristic fragments obtained in step (5.3) are compared with the MS database of oxysterol molecules constructed in step (5.1) to achieve qualitative detection of oxysterols in the sample.

[0022] Preferably, the MS database comprises a theoretical oxysterol database of MS information and MS / MS information.

[0023] Preferably, the sample is plasma.

[0024] Preferably, the method based on the obtained oxidized sterol chromatographic retention time prediction model, assisted by the mass spectrometry fragmentation characteristic fragments obtained in step (5.3) and compared with the oxidized sterol molecule MS database constructed in step (5.1) is: based on the characteristic fragment ions and precursor ions, the candidate structure is deduced in the database, and the retention time of the candidate structure is predicted according to the oxidized sterol chromatographic retention time prediction model obtained in claim 1, and the predicted retention time is compared with the experimental retention time to achieve qualitative detection.

[0025] The present invention also provides application of the method in qualitative detection of oxidized sterols lacking standard substances.

[0026] Compared with the prior art, the present invention has the following beneficial effects:

[0027] The present invention firstly constructs a theoretical database of oxidized sterols, and can more intuitively analyze the influence of the types of oxidized sterols and their structural characteristics on the chromatographic retention time, thereby improving the coverage of the analytical method; then based on the LC-MS / MS system and some available oxidized sterol standards, by summarizing the structural characteristics of oxidized sterols under specific analytical conditions, it is inferred that the characteristic fragment ions that may appear in the mass spectrum of oxidized sterols, and their potential relationship with the chromatographic retention time (RT), and is applied to the prediction of the retention time of other oxidized sterol molecules lacking commercial standards, thereby combining the characteristic fragment ion information of oxidized sterols, constructing a QSRR retention prediction model, and realizing the accurate characterization of oxidized sterols; finally, based on the cleavage characteristics, retention time information and theoretical database of the standard, the present invention applies the final complete database to biological sample analysis. This method can make up for the major deficiencies that the metabolite qualitative inaccuracy and inability to quantify caused by the lack of standard products during the current metabolite analysis.

[0028] The present invention realizes the separation of the existing 22 standards at the chromatographic level, providing valuable retention time data for the construction of the QSRR model. QSRR generates molecular descriptors from the chemical structure of the analyte and constructs a statistical model to obtain the relationship between the descriptor and the retention time of the analyte. This method can predict the retention time of new compounds that have not yet been synthesized artificially only through chemical structure descriptors. Relying on the powerful qualitative ability of the ultra-high resolution mass spectrometer QEplus and the ability of the quantitative structure-retention relationship (QSRR) model to predict chromatographic retention time through chemical structure, the present invention has established a high-coverage, stable and reliable qualitative analysis method for oxysterols, thereby providing technical support for the research of oxysterols. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 Schematic diagram of the structure of cholesterol;

[0030] Figure 2 is a schematic diagram of the structure of steroidal compounds;

[0031] Figure 3In the Full Scan mode, the extracted ion chromatogram shows that 1 is 4β-Hydroxycholesterol, 2 is 7α-Hydroxycholesterol, 3 is 7β-Hydroxycholesterol, 4 is 20S-Hydroxy cholesterol, 5 is 22R-Hydroxycholesterol, 6 is 22S-Hydroxycholesterol, 7 is 24S-Hydroxycholesterol, 8 is 24R-Hydroxycholesterol, 9 is 25-Hydroxy cholesterol, 10 is 27-Hydroxycholesterol, 11 is 7α,25-Dihydroxycholesterol, 12 is 7β,27-Dihydroxycholesterol, 13 is 7-Ketocholesterol, 14 is 5,6α-Epoxy cholesterol15 is 5,6β-Epoxycholesterol, 16 is 24,25-Epoxycholesterol, 17 is 7-Dihydrocholesterol, 18 is 8-Dihydrocholesterol, 19 is 24-Dihydrocholesterol, 20 is 7α-Hydroxycholestone, 21 is 19α-Hydroxycholesterol, 22 is 6α-hydroxycholestanol;

[0032] Figure 4 are extracted ion chromatograms of representative oxysterol standards, wherein A is the EIC graph of 7α-OH-Cholestenone and 7-Ketocholesterol in the sample, B is the EIC graph of 5,6β-Epoxycholesterol in the sample, C is the EIC graph of 8-Dihydrocholesterol in the sample, D is the EIC graph of 7α-OH-Cholestenone and 7-Ketocholesterol in methanol solution, E is the EIC graph of 5,6β-Epoxycholesterol in methanol solution, and F is the EIC graph of 8-Dihydrocholesterol in methanol solution;

[0033] Figure 5 It is the fitting diagram of predicted RT and actual RT;

[0034] Figure 6 is the fitting accuracy between the model and the actual RT;

[0035] Figure 7 This is a summary diagram of the mass spectrometry fragmentation rules of oxysterols;

[0036] Figure 8 The primary mass spectra of the oxidized sterol standard in the positive ion Full Scan mode and the secondary mass spectra in the PRM mode, wherein A and B are the primary mass spectra and secondary mass spectra of the representative hydroxyl oxidized sterol 24S-Hydroxycholesterol; C and D are the primary mass spectra and secondary mass spectra of the representative epoxy oxidized sterol 5,6β-Epoxycholesterol; E and F are the primary mass spectra and secondary mass spectra of the representative keto oxidized sterol 7-Ketocholesterol; G and H are the primary mass spectra and secondary mass spectra of the representative dehydrogenated sterol 7-Dihydrocholesterol;

[0037] Fig. 9 The application and validation results of QSRR in biological samples. DETAILED DESCRIPTION

[0038] The present invention provides a method for constructing an oxysterol chromatographic retention time prediction model, comprising the following steps:

[0039] (1) Construct a theoretical database of oxysterols;

[0040] (2) using E-Dragon and RDKit to calculate the descriptors of each oxysterol molecule in the theoretical database;

[0041] (3) determining the retention time of oxysterols according to the descriptors, using an internal standard to calibrate the retention time of oxysterols, and taking the ln value of the relative retention time to obtain a data set;

[0042] (4) For the obtained data set, a random partitioning method of 80:20 is used in combination with the regression algorithm to jointly construct the model;

[0043] (5) A grid search method and five-fold cross-validation were used to obtain the best hyperparameter combination, and the jointly constructed model was optimized to obtain a prediction model for the chromatographic retention time of oxysterols.

[0044] In the present invention, a theoretical database of oxysterols is constructed; by searching lipidmaps and pubchem databases, all oxysterols in the above databases are included in the theoretical database, and then through Pubmed and sci-finder literature databases, oxysterols that have not been reported in online databases but have been reported in literature are supplemented into the theoretical database; the classification, molecular formula, molecular weight, database source and number and other information of theoretical oxysterols are summarized, and duplicate items are merged.

[0045] In the present invention, E-Dragon and RDKit are used to calculate the descriptors of each oxidized sterol molecule in the theoretical database; the present invention searches for the compound name of the oxidized sterol in Pubmed to obtain the SDF file and SIMILES information of its three-dimensional structure, and for the oxidized sterol that cannot be retrieved in Pubmed, Chemdraw3.0 is used to draw its three-dimensional structure, and E-Dragon and RDKit are used to calculate the molecular descriptors of each oxidized sterol. The descriptors mainly cover the characteristics of lipophilicity, oxygen functional groups, hydrogen bonds and double bonds in the molecule, including fat solubility (ALOGP), number of hydrogen bonds (nHBonds), distance between oxygen atoms (G(O..O)), topological polar surface area (TPSA), number of double bonds (nDB), number of oxygen atoms (nO), intramolecular hydrogen bond strength (ITH) and molecular thermodynamic properties (HTm).

[0046] In the present invention, the retention time of oxidized sterols is determined according to the descriptor, the retention time of oxidized sterols is corrected using an internal standard, and the ln value of the relative retention time is taken to obtain a data set; since oxidized sterols generally differ only in the type and number of oxygen-containing functional groups, after analyzing the structure of existing oxidized sterols and their retention time on the chromatogram, it is found that there is a potential dependence relationship between the retention time of oxidized sterols and the position and number of their substituents, and the number and position of double bonds. For example, when the number of substituents is large, oxidized sterols usually elute earlier, such as dihydroxy oxidized sterols usually elute earlier than monohydroxy oxidized sterols; when the position of the substituent is on the side chain, the polarity of the oxidized sterol is relatively large, and the polarity is relatively small on the ring, such as in the existing standard products, the oxidized sterols with substituents on the side chain are all eluted earlier. Since the retention time distribution of oxidized sterols is not uniform, the effect of directly using retention time numerical modeling is not good. In addition, in order to avoid errors between different time periods and instruments, the present invention adopts the relative retention time (RRT) method, uses an internal standard to correct the retention time of each oxysterol, and takes the ln value of the relative retention time as the dependent variable of the model.

[0047] In the present invention, a random partitioning method of 80:20 is used to obtain the data set and combined with a regression algorithm to jointly construct a model; for a data set containing 22 standard products, the present invention uses a random partitioning method of 80:20 and an ElasticNet regression algorithm to model. In order to improve the expressive power of the model, the present invention performs polynomial feature expansion on the original features during the modeling process, and uses it in the modeling process, and determines through cross-validation that the order of the optimal polynomial is 2. Polynomial feature expansion not only increases the diversity of features, but also captures the nonlinear relationship between features. Afterwards, the expanded features are standardized, and the features are transformed into zero mean and unit variance using StandardScaler to eliminate the scale differences between features.

[0048] In the present invention, the grid search method and five-fold cross validation are used to obtain the best hyperparameter combination, and the jointly constructed model is optimized to obtain the prediction model of chromatographic retention time of oxysterols; in order to optimize the regularization parameter in ElasticNet regression, the present invention adopts the grid search method and obtains the best hyperparameter combination through five-fold cross validation, which further improves the prediction ability of the model. 2 It is used to measure the degree of fit of the model. 2 The closer to 1, the better the model fitting effect. MAE is a measure of the average deviation between the predicted value and the true value. The smaller the value, the more accurate the model prediction. The prediction model of the chromatographic retention time of oxysterols is y=-0.1574+0.0419*nO+(-0.1205)*G(O..O)+0.0036*nHBonds+0.0278*ALOGP*nO+(-0.2389)*ALOGP*G(O..O)+(-0.0291)*ALOGP*ITH+(-0.0614)*ALOGP*n HBonds+(-0.0210)*ALOGP*HTm+0.0894*nO^2+0.0027*nO*nDB+0.0025*nO*nHBonds+0.0132*nO*HTm+(-0.0860)*G (O..O)*nDB+(-0.1036)*G(O..O)*nHBonds+(-0.1001)*G(O..O)*HTm+0.0510*nDB^2+(-0.0328)*ITH^2+0.0010*IT H*nHBonds+0.0018*TPSA*nHBonds+0.0032*nHBonds*HTm.

[0049] The present invention provides a method for qualitative detection of oxysterols, comprising the following steps:

[0050] (5.1) Constructing MS database of oxysterol molecules;

[0051] (5.2) Use Ful MS mode and PRM mode to obtain the molecular ion peaks and secondary mass spectrometry fragmentation information of oxysterol standards, and summarize the additive form and secondary mass spectrometry fragmentation rules of oxysterols;

[0052] (5.3) using full scan mode to obtain the molecular ion peak of the oxysterol in the sample, and using PRM mode to obtain the secondary mass spectrometry fragmentation information of the oxysterol in the sample, combining the summation form obtained in step (5.2) with the secondary mass spectrometry mass spectrometry fragmentation law, to obtain the possible mass spectrometry fragmentation characteristic fragments of the oxysterol in the sample;

[0053] (5.4) Based on the chromatographic retention time prediction model of oxysterols obtained in claim 1, the mass spectrometry fragmentation characteristic fragments obtained in step (5.3) are compared with the MS database of oxysterol molecules constructed in step (5.1) to achieve qualitative detection of oxysterols in the sample.

[0054] In the present invention, an MS database of oxysterol molecules is constructed; constructing an MS database of oxysterol molecules is a prerequisite for accurate qualitative characterization of oxysterols. In order to increase the rigor and reliability of the database, the present invention will construct a theoretical oxysterol database containing MS information and MS / MS information. In the MS system, the chemical structure determines the primary and secondary ion performance of the molecule.

[0055] In the present invention, the molecular ion peaks and secondary mass spectrometry fragmentation information of oxysterol standards were obtained by using the Full MS mode and the PRM mode, and the additive form and secondary mass spectrometry fragmentation rules of oxysterols were summarized; in the positive ion mode, the Full MS mode was used to observe the optimal molecular ion peaks of different types of oxysterols, among which the optimal molecular ion peaks of monohydroxy oxysterols and epoxy groups were [M+H-H2O] + The optimal ion peak of oxysterols containing keto groups is [M+H] + The optimal ion peak of multi-substituted oxysterols is [M+H-2H2O] + The above summary of the addition form of oxysterols and the secondary mass spectrometry fragmentation rules are as follows Figure 7 As shown, the mass spectra of various oxysterol representatives are as follows Figure 8 shown.

[0056] In the present invention, the full scan mode is used to obtain the molecular ion peak of the oxysterol in the sample, and the PRM mode is used to obtain the secondary mass spectrometry fragmentation information of the oxysterol in the sample. The summation form obtained in step (5.2) is combined with the secondary mass spectrometry mass spectrometry fragmentation law to obtain the possible mass spectrometry fragmentation characteristic fragments of the oxysterol in the sample.

[0057] In the present invention, based on the obtained oxidized sterol chromatographic retention time prediction model, the mass spectrometry fragmentation characteristic fragments obtained in step (5.3) are used to compare with the oxidized sterol molecule MS database constructed in step (5.1), so as to achieve qualitative detection of oxidized sterols in the sample; the method based on the obtained oxidized sterol chromatographic retention time prediction model, the mass spectrometry fragmentation characteristic fragments obtained in step (5.3) are used to compare with the oxidized sterol molecule MS database constructed in step (5.1) is as follows: based on the characteristic fragment ions and precursor ions, a candidate structure is deduced in the database, and the retention time of the candidate structure is predicted according to the obtained oxidized sterol chromatographic retention time prediction model, and the qualitative detection is achieved by comparing the predicted retention time with the experimental retention time, and the sample is plasma.

[0058] The present invention also provides application of the method in qualitative detection of oxidized sterols lacking standard substances.

[0059] The technical solutions provided by the present invention are described in detail below in conjunction with the embodiments, but they should not be construed as limiting the protection scope of the present invention.

[0060] Example 1 Summary of theoretical database of oxysterols

[0061] Oxysterols are the oxidation products of cholesterol, and their skeleton usually has a double bond at C5 and C6 and a hydroxyl group at C3. Some oxysterols are derived from cholesterol precursors, such as dehydrocholesterol and cholesterol.

[0062] First, databases such as lipidmaps (https: / / www.lipidmaps.org / ) and pubchem (https: / / pubchem.ncbi.nlm.nih.gov) were searched, and all oxysterols in the above databases were included in the theoretical database.

[0063] Secondly, through the literature databases such as Pubmed (https: / / pubmed.ncbi.nlm.nih.gov / ) and sci-finder (https: / / scifinder-n.cas.org), the oxysterols that have not been reported in the online database but have been reported in the literature were added to the theoretical database. The classification, molecular formula, molecular weight, database source and number of theoretical oxysterols were summarized, and duplicates were merged.

[0064] Example 2 Detection and Analysis of Oxysterols

[0065] 1. Preparation of Oxysterol Standard Stock and Working Solutions

[0066] Preparation of stock solutions: Accurately weigh each oxysterol standard and prepare a 1.0 mg / mL standard stock solution in methanol.

[0067] Preparation of working solution: Accurately pipette the standard stock solution, dilute it with methanol to obtain a working solution concentration of 100 ng / mL, and simultaneously prepare a mixed standard solution with a concentration of 100 ng / mL. Store in a -20℃ refrigerator until testing. The standard information is shown in Table 1.

[0068] Table 1 Standard product information

[0069]

[0070]

[0071] 2. Chromatographic parameters

[0072] ACQUITY UPLC CORTECS C8 (1.6 μm, 2.1*100 mm, Waters, USA) was used. The column temperature was 40°C and the injection volume was 3 μL. Mobile phase: phase A: water (containing 0.1% formic acid), phase B: acetonitrile: methanol = 8:2, gradient elution was used. The HPLC mobile phase elution gradient is shown in Table 2.

[0073] Table 2 HPLC mobile phase elution gradient

[0074]

[0075]

[0076] 3. Mass spectrometry parameters

[0077] The full scan mode (Full Scan) was used to obtain the peak time and separation of each standard in the mixed standard solution. In the Full Scan mode, positive ion mode was used for acquisition, with a scan range of m / z = 50-500; a resolution of 70000; an electrospray ion source (ESI); a capillary temperature (Capillary Temp): 320°C; a spray voltage (Spray Voltage): 3.8 kV; a maximum spray current (Max Spray Current): 100 μA; a sheath gas flow (Sheath Gas Flow): 30 L / min; an auxiliary gas flow (Aux Gas Flow): 5 L / min; and a probe heating temperature (Probe Heater Temp): 300°C.

[0078] The PRM mode was used to obtain the secondary mass spectrometry fragmentation information of the standard. In the PRM mode, the positive ion mode was used for acquisition, the scanning range was: m / z 50-500; the resolution was 17500; the capillary temperature (Capillary Temp): 320℃; the spray voltage (Spray Voltage): 3.8kV; the maximum spray current (Max Spray Current): 100μA; the sheath gas flow (Sheath Gas Flow): 30L / min; the auxiliary gas flow (Aux Gas Flow): 5L / min; the probe heating temperature (Probe Heater Temp): 300℃; the collision energy (Collision Energy): nCE mode, and the nCE range can be selected as 25V, 35V, and 40V.

[0079] 4. Plasma sample pretreatment

[0080] Take 50 μL of plasma in a glass centrifuge tube and add 5 μL of 500 μg / mL BHT solution. Add 500 μL of Folch extractant (methylene chloride-methanol volume ratio 2:1), vortex for 30 seconds, add 100 μL of water to separate the phases, vortex for 30 seconds, and centrifuge at 12000 rpm for 15 minutes. Take 300 μL of the lower layer of clear liquid and add 6 μL of 500 ng / mL internal standard solution. Blow dry and re-dissolve with 30 μL of phase B mobile phase (acetonitrile-methanol volume ratio 8:2), centrifuge at 12000 rpm for 15 minutes, and take the upper layer of clear liquid into the injection vial for injection.

[0081] 5. Data Processing

[0082] ThermoFisher Xcalibur software was used for data acquisition, and ThermoFisher QualBrowser software was used for qualitative analysis.

[0083] 6. Experimental Results

[0084] 6.1 Liquid phase separation results of standard products

[0085] The mixed standard working solution in "1. Preparation of oxysterol standard stock solution and working solution" was loaded on the instrument under the conditions of "2. Chromatographic parameters" and "3. Mass spectrometry parameters", and the [M+H] of the Full Scan scan results were extracted in the ThermoFisher QualBrowser software analysis. + and [M+H-H2O] + Chromatographic peak. Mass accuracy deviation 5ppm. Extracted ion chromatogram as shown Figure 3 shown.

[0086] Depend on Figure 3 It can be seen that the chromatographic method of the present invention has a strong separation ability for oxysterol standards and is suitable for the detection and analysis of oxysterols.

[0087] 6.2 Retention time deviation between actual samples and standards

[0088] To ensure that the constructed retention prediction model is applicable to actual samples, the RT of the standard EIC of the same oxysterol in the FullScan mode of the same batch was compared with the RT of the EIC in the biological sample to observe whether the RT was shifted. The RT difference between the hydroxy oxysterol (7α-Hydroxycholesterol), epoxy oxysterol (5,6α-Epoxycholesterol), and keto oxysterol (7-Ketocholesterol) that existed in high concentrations in the actual sample was compared with the standard solution to determine whether the plasma matrix had an effect on the RT. The results are shown in Figure 4 .

[0089] Depend on Figure 4 It can be seen that the RT of the oxidized sterols in the sample is close to the RT of the standard, and ΔRT is less than or equal to 0.1 min. Therefore, the interference of the matrix on RT can be ignored. At the same time, it is shown that the RT value measured by the chromatographic method of the present invention is stable in the sample and standard solution. The chromatographic conditions based on the standard RT can be applied to oxidized sterols in the plasma matrix.

[0090] 6.3 Development and validation of a retention time prediction model for oxysterols

[0091] Since oxysterols generally differ only in the type and number of oxygen-containing functional groups, the most commonly used molecular descriptors such as logP are not applicable in describing oxysterol molecules. In this case, it is necessary to explore more molecular descriptors that may be able to distinguish these isomers, or even stereoisomers. After analyzing the structure of existing oxysterols and their retention on chromatography, it was found that there is a potential dependence between the retention time of oxysterols and their structural characteristics. These structural characteristics are summarized as the position and number of substituents, and the number and position of double bonds. For example, when the number of substituents is large, oxysterols usually elute earlier, such as dihydroxy oxysterols usually elute earlier than monohydroxy oxysterols; when the position of the substituent is on the side chain, the polarity of the oxysterol is relatively large, and the polarity is relatively small on the ring. For example, in the existing standard products, the elutions are all oxysterols with substituents on the side chain, which is consistent with previous studies.

[0092] Calculation of descriptors: First, search in Pubmed by the compound name of oxidized sterols to obtain the SDF file and SIMILES information of its three-dimensional structure. For the oxidized sterols that cannot be retrieved in Pubmed, Chemdraw3.0 is used here to draw its three-dimensional structure. E-Dragon and RDKit are used to calculate the molecular descriptors of each oxidized sterol. The selected descriptors mainly cover the characteristics of lipophilicity, oxygen functional groups, hydrogen bonds and double bonds in the molecule, including eight descriptors such as ALOGP (lipid solubility), nHBonds (number of hydrogen bonds), G (O..O) (distance between oxygen atoms), TPSA (topological polar surface area), nDB (number of double bonds), nO (number of oxygen atoms), ITH (intramolecular hydrogen bond strength) and HTm (molecular thermodynamic properties). In the selection of descriptors, the present invention combines the structural characteristics of oxidized sterols themselves to ensure that the input descriptors can fully reflect the key factors affecting the retention time of oxidized sterols.

[0093] Data preprocessing: Since the retention time distribution of oxysterols is not uniform, the effect of directly using the retention time numerical modeling is poor. In addition, in order to avoid errors between different time periods and instruments, this experiment adopted the relative retention time (RRT) method, using internal standards to correct the retention time of each oxysterol, and taking the ln value of the relative retention time as the dependent variable of the model.

[0094] Model construction and optimization: For a data set containing 22 standards, this experiment uses a random partitioning method of 80:20 to build a model to ensure that the model can be verified on certain independent samples. After comparing the performance of various algorithms such as random forest, support vector machine, Lasso regression, ElasticNet, etc., ElasticNet regression is finally used to build the model. This regression algorithm combines the advantages of ridge regression and Lasso regression, and can perform effective parameter selection in high-dimensional feature space. In order to improve the expressive power of the model, the modeling process of the present invention performs polynomial feature expansion on the original features, which is used in the modeling process, and the order of the optimal polynomial is determined to be 2 through cross-validation. Polynomial feature expansion not only increases the diversity of features, but also captures the nonlinear relationship between features. Afterwards, the expanded features are standardized, and the features are transformed into zero mean and unit variance using StandardScaler to eliminate the scale differences between features.

[0095] Model optimization: In order to optimize the regularization parameters in ElasticNet regression, the present invention adopts a grid search method and obtains the best hyperparameter combination through five-fold cross validation, which further improves the prediction ability of the model. The prediction model of chromatographic retention time of oxysterols is y=-0.1574+0.0419*nO+(-0.1205)*G(O..O)+0.0036*nHBonds+0.0278*ALOGP*nO+(-0.2389)*ALOGP*G(O..O)+(-0.0291)*ALOGP*ITH+(-0.0614)*ALOGP*n HBonds+(-0.0210)*ALOGP*HTm+0.0894*nO^2+0.0027*nO*nDB+0.0025*nO*nHBonds+0.0132*nO*HTm+(-0.0860)*G(O..O)*nDB+(-0.1036)*G(O..O)*nH Bonds+(-0.1001)*G(O..O)*HTm+0.0510*nDB^2+(-0.0328)*ITH^2+0.0010*IT H*nHBonds+0.0018*TPSA*nHBonds+0.0032*nHBonds*HTm. The R of the final model on the training set is 2 The R 2 is 0.9702, and MAE is 0.0537 (Table 3). 2 It is used to measure the degree of fit of the model. 2 The closer it is to 1, the better the model fit is. MAE measures the average deviation between the predicted value and the true value. The smaller the value, the more accurate the model prediction is.

[0096] Model validation: In order to avoid overfitting of the model, the present invention uses internal leave-one-out cross validation and four random four-fold cross validation to test the performance of the model. The final model is R 2 and MAE performed well, indicating that the model has accurate prediction performance (Table 3 and Table 4). Figure 5 and Figure 6), except for one deviation value, the error between the experimental value and the predicted value of the remaining data is within 10%, which can be used as the confidence interval of the model, and at the same time proves that the model has good retention prediction performance. In order to ensure that the model is suitable for RT prediction of other oxysterols with different structures, the present invention selects 4 new standards (not involved in the construction of the model) for external validation. The external validation results show that the error between the experimental value and the predicted value of the data is within 10%, which is within the confidence interval of the model.

[0097] In conclusion, the prediction accuracy of the standards was high, indicating that the model has good predictive ability and high accuracy and can be used for the qualitative characterization of oxysterols in human plasma.

[0098] Table 3 Model training results

[0099]

[0100] Table 4 Comparison of retention time between training set and test set

[0101]

[0102]

[0103] 6.4 Summary and verification of mass spectrometric characteristics of oxysterols

[0104] Constructing an MS database of oxysterol molecules is a prerequisite for accurate qualitative characterization of oxysterols. To increase the rigor and reliability of the database, the present invention will construct a theoretical oxysterol database containing MS information and MS / MS information. In the MS system, the chemical structure determines the primary and secondary ion performance of the molecule. Oxysterols have similar structures, including a 27-carbon skeleton and oxygen-containing functional groups on the ring or side chain, so they have similar cleavage patterns, see Figure 7 .

[0105] In positive ion mode, use Full MS mode to observe the best molecular ion peaks for different types of oxysterols. Usually the best molecular ion peak for monohydroxy oxysterols and epoxy groups is [M+H-H2O] + The optimal ion peak of oxysterol containing keto group is [M+H] + In addition, multi-substituted oxysterols may also appear [M+H-2H2O] + The best ion peak.

[0106] Taking the structure of 7α-Hydroxycholesterol as an example, under the action of collision voltage, the hydroxyl groups on the ring or side chain undergo dehydration reaction after the proton of the oxysterol dissociates to generate [M+H-H2O] + 、[M+H-2H2O]+ The fragment ions, such as Figure 7 m / z=385.34624, m / z=367.33640. Assuming that the first step of the cleavage of the oxidized sterol is the cleavage of the side chain and the ring, this part leads to the generation of m / z=255. In the ring skeleton, the fragmentation sites are usually between the carbon atoms with IUPAC numbers of 11-12 and 8-14, between the carbon atoms of 12-13 and 8-14, and between the carbon atoms of 1-10 and 5-6, and these fragmentation sites usually bring m / z=159 and m / z=95, m / z=173 and m / z=81, and m / z=109 characteristic ions. In addition, the fragmentation between the carbon atoms of 13-17 and 14-15 brings the characteristic ion of m / z=213. In addition to these main fragments, a series of mass bands with a mass difference of 14Da (CH2) appeared in the secondary mass spectrum. These mass bands may be caused by the breakage of the oxidized sterol carbon skeleton at each site, such as m / z=147, m / z=133, m / z=121, etc.

[0107] Example 3 Application of the LC-MS / MS-based retention time prediction and qualitative method for oxysterols in human plasma

[0108] 1. Sample Collection

[0109] Postmenopausal women who were treated in the Department of Orthopedics of Shanghai Fifth People's Hospital from February 2019 to April 2021 were selected as the research subjects, and the study was approved by the Medical Ethics Committee of Shanghai Fifth People's Hospital. Clinical information collection and sample collection were completed with the knowledge and consent of all participants. All participants were strictly matched for age and BMI and had no special diet, smoking and drinking habits. Patients with cancer, liver disease, kidney disease, or metabolic or hereditary bone diseases, and patients who used bisphosphonates, glucocorticoids, estrogens and other drugs that may affect bone metabolism within 6 months were excluded. 15 healthy human plasma and osteoporotic human plasma were evenly mixed and frozen in a -80℃ refrigerator for use.

[0110] 2. Data processing

[0111] ThermoFisher Xcalibur software was used for data acquisition, and ThermoFisher QualBrowser software was used for qualitative analysis.

[0112] 3. Results and Discussion

[0113] Through the methods of Example 1 and Example 2, a RT prediction method for 80 oxysterols summarized by literature retrieval and database was constructed. The predicted RT information was combined with the mass spectrometry fragmentation law of oxysterols, and finally a Full-MS full scan was used in human plasma. The theoretical parent ion mass-to-charge ratio and RT were imported into the inclusion list, and a PRM mode scan was performed to obtain secondary mass spectrometry information for the qualitative detection of oxysterols.

[0114] Taking P5 (Cholestenone) as an example, it is explained that based on the QSRR model, unknown oxysterols in biological samples are further identified. The experimental retention time of compound P5 (precursor ion m / z385.34607) in XIC is 30.62min. According to the characteristic fragment ions in its mass spectrum, it is inferred that the compound is an oxysterol. Based on the characteristic fragment ions and precursor ions, 7 candidate structures are derived in the database. These substances have the same mass-to-charge ratio in the mass spectrum and cannot be distinguished at the mass spectrum level. The retention time of these candidate structures is predicted using the QSRR model, such as Fig. 9 As shown, the predicted retention times (RT pre ) were 11.18min, 14.52min, 15.59min, 21.92min, 23.24min, 28.05min, and 28.50min, respectively. The error between the experimental value and the predicted value should be within the threshold of 10%. By comparing the predicted value with the experimental retention time, the compound was finally determined to be Cholestenone, and confirmed using synthetic standards, ensuring the reliability of the QSRR method for the identification of oxysterols.

[0115] Finally, 35 oxysterols were actually identified in human plasma (21 without commercial standards and 14 with commercial standards), as shown in Tables 5 and 6.

[0116] Table 5 Known oxysterols detected in mixed human plasma (including standards)

[0117]

[0118]

[0119] Table 6 Identification results combined with QSRR prediction

[0120]

[0121]

[0122]

[0123]

[0124] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for constructing a prediction model for chromatographic retention time of oxysterols, characterized in that: The following steps are involved: (1) Construct a theoretical database of oxysterols; (2) using E-Dragon and RDKit to calculate the descriptors of each oxysterol molecule in the theoretical database; (3) determining the retention time of oxysterols according to the descriptors, using an internal standard to calibrate the retention time of oxysterols, and taking the ln value of the relative retention time to obtain a data set; (4) For the obtained data set, a random partitioning method of 80:20 is used in combination with the regression algorithm to jointly construct the model; (5) A grid search method and five-fold cross-validation were used to obtain the best hyperparameter combination, and the jointly constructed model was optimized to obtain a prediction model for the chromatographic retention time of oxysterols.

2. The construction method according to claim 1, characterized in that: The theoretical database of oxysterols is obtained by collecting all oxysterols reported in the database; the database includes lipidmaps, pubchem, Pubmed, and sci-finder.

3. The construction method according to claim 1, characterized in that: The descriptors include lipid solubility, number of hydrogen bonds, distance between oxygen atoms, topological polar surface area, number of double bonds, number of oxygen atoms, strength of intramolecular hydrogen bonds, and molecular thermodynamic properties.

4. The construction method according to claim 1, characterized in that: The regression algorithm is the ElasticNet regression algorithm.

5. A method for qualitative detection of oxysterols, characterized in that: The following steps are involved: (5.1) Constructing MS database of oxysterol molecules; (5.2) Use Ful MS mode and PRM mode to obtain the molecular ion peaks and secondary mass spectrometry fragmentation information of oxysterol standards, and summarize the additive form and secondary mass spectrometry fragmentation rules of oxysterols; (5.3) using full scan mode to obtain the molecular ion peak of the oxysterol in the sample, and using PRM mode to obtain the secondary mass spectrometry fragmentation information of the oxysterol in the sample, and combining the additive form law and the secondary mass spectrometry mass spectrometry fragmentation law obtained in step (5.2) to obtain the possible mass spectrometry fragmentation characteristic fragments of the oxysterol in the sample; (5.4) Based on the chromatographic retention time prediction model of oxysterols obtained in claim 1, the mass spectrometry fragmentation characteristic fragments obtained in step (5.3) are compared with the MS database of oxysterol molecules constructed in step (5.1) to achieve qualitative detection of oxysterols in the sample.

6. The method for qualitative detection of oxysterols according to claim 5, characterized in that: The MS database includes a theoretical oxysterol database of MS information and MS / MS information.

7. The method for qualitative detection of oxysterols according to claim 5, characterized in that: The sample is plasma.

8. The method for qualitative detection of oxysterols according to claim 5, characterized in that: The method of using the oxidized sterol chromatographic retention time prediction model obtained in claim 1, supplemented by the mass spectrometry fragmentation characteristic fragments obtained in step (5.3) and comparing with the oxidized sterol molecule MS database constructed in step (5.1) is as follows: based on the characteristic fragment ions and precursor ions, candidate structures are derived from the database, According to the chromatographic retention time prediction model for oxysterols obtained in claim 1, the retention time of the candidate structure is predicted, and qualitative detection is achieved by comparing the predicted retention time with the experimental retention time.

9. Use of the method according to any one of claims 5 to 8 in the qualitative detection of oxysterols in the absence of standards.

Images