A method and system for analyzing drug metabolites

Abstract

The present invention discloses an analysis method and system for drug metabolites, relating to the technical field of analysis of drug metabolites. This analysis method performs liquid chromatography-high resolution tandem mass spectrometry detection on the metabolites of the drug to be tested, filters the data by MDF, performs DP analysis, compares the MS<supgt;2< / supgt> spectrum similarity and fragment ion similarity between the parent drug and the metabolite, and retains the MS<supgt;2< / supgt> spectrum and MS<supgt;1< / supgt> spectrum that meet the requirements; predicts the molecular weight change of potential metabolites, constructs a metabolite prediction model; predicts the MS<supgt;1< / supgt> spectrum, and judges possible metabolites; speculates the cleavage pathway, constructs the MS molecular network of the parent drug, and judges characteristic ions; calculates NLF filtering; establishes the metabolic relationship between the parent drug and the metabolite fragments according to the MS molecular network of the parent drug and the MS<supgt;2< / supgt> spectrum, constructs a microscopic MS molecular network and confirms the characteristic ions. The method provided by the present invention can quickly discover and identify the metabolites of drugs (radioisotope tracer drugs), and has high feasibility, effectiveness and accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of drug metabolite analysis, and in particular to a drug metabolite analysis method and system. Background Art

[0002] The absorption, distribution, metabolism and excretion (ADME) properties of drugs affect their drugability, effectiveness and safety. Drug metabolism research is an important part and content in the drug research and development process. It runs through the entire process of drug discovery and development and is one of the key factors that determine whether a drug can be marketed. The tracking and structural identification of drug metabolites is one of the core contents of ADME tests for innovative drugs and is also a difficult point in drug metabolism research. The rapid and accurate identification of drug metabolites in biological matrices helps to understand the biotransformation pathways of drugs or candidate compounds, determine metabolic soft sites, and establish the relationship between "structure-metabolic pathways", which can guide drug design and structural optimization and screen out drugs with higher metabolic stability.

[0003] Most drug metabolism pathways are considered to be detoxification processes, that is, drugs undergo biotransformation under the mediation of drug metabolizing enzymes in the body to generate metabolites with greater polarity and higher water solubility. Compared with the parent drug, the pharmacological activity of these metabolites is usually weakened or completely eliminated. However, in a few cases, some drugs and candidate compounds may generate reactive metabolic intermediates under the mediation of drug metabolizing enzymes in the body. The reactive intermediates further covalently bind to biological macromolecules such as proteins and DNA, causing cell damage and thus toxicity. In recent years, the potential risk of screening new drug candidate compounds to generate reactive metabolites has attracted widespread attention. Therefore, the discovery of active metabolites and toxic metabolites is very important for pharmacodynamics and safety evaluation.

[0004] Due to the diversity of drug chemical structures and the complexity of in vivo biotransformation pathways, drug metabolite identification research is difficult. The interference of various endogenous substances in the biological matrix and the low content of the metabolites themselves further increase the difficulty of rapid and accurate structural identification. High performance liquid chromatography-mass spectrometry (HPLC-MS) has the high separation ability of liquid chromatography and the high sensitivity and excellent selectivity of mass spectrometry. It can be used to carry out high-throughput and high-efficiency metabolite identification research, and has gradually become the mainstream method in the field of drug metabolite identification research.

[0005] High performance liquid chromatography (HPLC) and ultra-high performance liquid chromatography (UPLC) have become effective means of analyzing drug metabolites due to their large sample loading capacity, good applicability, and high resolution. They generally use mass spectrometry as a detector. Compared with other detection technologies, mass spectrometry can provide higher selectivity, sensitivity, and richer structural information.

[0006] High performance liquid chromatography-high resolution mass spectrometry (HPLC-HR-MS) has good chromatographic separation, high sensitivity and high resolution, and can well separate, detect and characterize drug metabolites in biological matrices. Drug metabolites can be proposed based on accurate mass and ion fragments. However, the large amount of information provided by HR-MS makes it difficult and time-consuming to effectively extract drug metabolite characterization data.

[0007] Radioisotopes play a vital role in studying the ADME properties of innovative drugs and are gradually becoming an indispensable tool in ADME tests of innovative drugs. Since radioisotopes have low natural abundance values, the use of radioisotopes has the characteristics of high specificity, sensitivity, high sensitivity and easy detection. Most of the drug safety evaluation studies used in the application for new drug registration (NDA) of the US Food and Drug Administration (FDA) use pharmacokinetic data of radioisotope-labeled compounds in animals and humans after administration.

[0008] Isotope filtering is often used to screen the metabolites of isotope models of drugs containing halogen elements such as Cl and Br. At present, there is a lack of isotope filtering strategies for the identification of metabolites of radioactive isotope tracer drugs, that is, isotope filtering of hot substances (radioactive isotope tracer drugs) and cold substances (non-radioactive isotope tracer drugs). The commercial metabolite identification software developed by mass spectrometry equipment suppliers is expensive due to the lack of commonality between different mass spectrometry platforms; and, at present, due to certain technical barriers and regulatory requirements for radioactive isotope tracer drug metabolites, the number of institutions with radioactive isotope tracer drug research qualifications is very limited. Some mass spectrometry data analysis platforms only have relevant data processing strategies for compounds with natural isotope structures (such as halogens, sulfur atoms, etc.). However, there are obvious differences in isotope abundance and ratio between natural isotopes and radioactive isotope tracer drugs. Therefore, no mass spectrometry data analysis platform for the identification of radioactive isotope tracer drug metabolites has been found. Summary of the invention

[0009] The drug metabolite analysis method and system provided by the present invention is a mass spectrometry data analysis method and platform for identifying metabolites of conventional drugs, especially radioactive isotope tracer drugs, based on an open source platform (Python and mass spectrometry molecular network). The platform utilizes the experimental characteristics of radioactive isotope tracer drug metabolites. Cold substances and hot substances are administered in a fixed ratio, and the in vivo exposure also follows this ratio; the raw mass spectrometry data obtained is processed using special cold substance and hot substance isotope filtering to quickly discover and identify metabolic pathways and metabolites. It is specifically achieved through the following technologies.

[0010] A method for analyzing drug metabolites, comprising the following steps:

[0011] Taking the metabolites of the drug to be tested (in vitro or in vivo) for radioisotope flux detection; the drug to be tested is a non-radioisotope tracer drug or a radioisotope tracer drug; if the drug to be tested is a radioisotope tracer drug, the corresponding metabolite information is detected;

[0012] Liquid chromatography-high-resolution mass spectrometry detection was performed to obtain raw data in a common format, and the metabolites of the radiolabeled drug were located at the same retention time of the primary mass spectrometry peak;

[0013] The raw data in the general format are filtered by radioisotope and mass defect to obtain the MS of metabolites (cold substances and hot substances if it is a radioisotope tracer drug). 1 Spectra and MS 2 Atlas;

[0014] The MS of the parent drug and metabolites were compared using the point integration algorithm. 2 The similarity of the spectra and fragment ions is used to screen and retain the MS 2 Spectra and corresponding MS 1 Atlas;

[0015] Predict the molecular weight changes of potential metabolites of the drug to be tested, and use this as a training data set to build a metabolite prediction model; predict the MS data set, and determine the possible metabolites based on the output results of the metabolite prediction model;

[0016] Combined with MS 2 The spectral data (fragment ions of metabolites of the drug to be tested) are used to infer the fragmentation pathway of the drug to be tested, and the MS molecular network of the parent drug is constructed according to the fragmentation pathway to determine whether there are characteristic ions;

[0017] The bound metabolites were screened out by calculating neutral loss;

[0018] Based on the MS molecular network of the parent drug (i.e., prototype drug) and the metabolite fragment ion information (MS 2 Spectral data) to establish the metabolic relationship between the parent drug and metabolite fragments, and to construct a microscopic MS molecular network (i.e., MS 1 and MS 2 The characteristic ions are confirmed based on the microscopic MS molecular network.

[0019] The existing drug metabolite prediction and identification methods can only process compounds containing natural isotopes and cannot be used for the identification of metabolites of radioactive isotope tracer drugs. The present invention uses radioactive isotope filtration and radioactive isotope flow detection, and based on the same pharmacokinetic and drug metabolism characteristics of hot and cold substances in the body, the DP algorithm is used to determine the metabolites of hot substances.

[0020] Specifically:

[0021] (1) Metabolites produced by radioactive isotope-labeled drugs are first subjected to radioactive flux detection (liquid chromatography), so that all metabolites of radioactive labeled drugs are detected; non-radioactive labels cannot be detected;

[0022] (2) The metabolites of radiolabeled drugs are located at the same retention time of the primary mass spectrometric peak;

[0023] (3) then performing liquid chromatography-high-resolution mass spectrometry detection;

[0024] (4) Through radioisotope filtering, the radioactive labeling characteristics (such as isotope abundance, ratio, etc.) are found, and radioisotope filtering, mass loss filtering, nitrogen law filtering and other processing are performed to screen and determine the MS of hot and cold substances. 1 Atlas;

[0025] (5) Hot substances and cold substances have the same metabolic sites, MS 2 The spectra are highly similar, and the point integration algorithm (DP) is used to determine the metabolites of thermal substances; the metabolites of radioactive isotope tracer drugs are identified.

[0026] It should be noted that the metabolite prediction model is trained by using the predicted metabolite molecular weight as a training data set, traversing the MS data and outputting a mass spectrum containing the ion peaks in the data set.

[0027] The predicted metabolite molecular weights (data set) are mainly 385.3578, 389.3526, 399.3369, 401.3526, 405.3475, 417.3475, 417.3840, 419.3632, 433.3424, 433.3788, 435.3581, 451.3530, 483.3251, 499.3200, etc.

[0028] It should be noted that the metabolite molecular weight prediction result is a data set containing a large number of predicted molecular weights, and the above prediction result is only a part of the data set.

[0029] Furthermore, the mass defect filtering window was set to ±50 mDa for mass defect filtering;1 Medium [M+H] + Parity of the integer part of m / z.

[0030] Furthermore, whether to perform nitrogen law filtering is determined based on whether the chemical formula of the drug to be tested contains nitrogen atoms.

[0031] Furthermore, the point integration algorithm is used for analysis as follows: the top 50 peak intensity ion peaks are selected and the peak intensity normalized value W is calculated using the following formula (1): i ;

[0032] W i =[intensity of peak i ] m [m / z of peak i ] n (1)

[0033] Among them, W i represents the normalized value of peak intensity, m and n represent the mass weight and intensity proportional factor respectively; the DP value is calculated using the following formula (2):

[0034]

[0035] Among them, W u and W ai represent the normalized values ​​of the peak intensities of the measured spectrum and the reference spectrum, respectively.

[0036] Optionally, the peak intensity normalized value W i In the formula, m=0.5, n=2.

[0037] Furthermore, for the metabolites of non-radioactive isotope tracer drugs, MS with DP values ​​≥ 0.8 were screened. 2 Spectra and corresponding MS 1 Atlas.

[0038] Furthermore, the drug is a radioactive isotope tracer drug.

[0039] Furthermore, it also includes drawing a mirror image of the visualized parent drug MS molecular network and the microscopic MS molecular network through numpy and matplotlib libraries.

[0040] The present invention also provides an analysis system for drug metabolites, including a preprocessing module, a mass loss filtering module, a point integration algorithm module, a metabolite prediction model module, an MS molecular network module, and a neutral loss screening module;

[0041] A preprocessing module, used to convert the raw data output after liquid chromatography-high-resolution secondary mass spectrometry detection into raw data in a universal format;

[0042] The mass defect filtering module is used to filter the raw data in the general format to obtain MS 1 Spectra and MS 2 Atlas;

[0043] Point integration algorithm module for comparing the MS of parent drug and metabolites 2 The similarity of the spectra and fragment ions is used to screen and retain the MS 2 Spectra and corresponding MS 1 Atlas;

[0044] The metabolite prediction model module is used to predict the molecular weight changes of potential metabolites of the drug to be tested and use this as a training data set to build a metabolite prediction model; 1 Predict the spectral data, and determine the possible metabolites based on the output results of the metabolite prediction model; re-input or modify the training data set based on different drugs to be tested;

[0045] MS molecular networking module for combining MS 2 The cleavage pathway of the drug to be tested is inferred based on the spectral data and expert experience, and the MS molecular network of the parent drug is constructed based on the cleavage pathway to determine whether there are characteristic ions; based on the MS molecular network of the parent drug and MS 2 Establishing the metabolic relationship between the parent drug and the metabolite fragments based on the spectral data, constructing a microscopic MS molecular network; confirming the characteristic ions based on the microscopic MS molecular network;

[0046] Neutral loss screening module, used to calculate neutral loss and screen out bound metabolites.

[0047] Compared with the prior art, the advantages of the present invention are as follows: based on the open source Python and mass spectrometry molecular network platform technology, the present invention establishes a mass spectrometry data analysis method for drug metabolites, especially the identification of metabolites of radioactive isotope tracer drugs, and utilizes special cold and hot material isotope filtering to quickly discover and identify metabolites; the analysis and identification results have high feasibility, effectiveness and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 A workflow for a hybrid strategy for drug metabolite characterization based on Python platform-based intelligent MS data processing combined with MS molecular networking;

[0049] Figure 2 and Figure 3 These are the TIC diagrams of MDZ after incubation in human liver microsomes; Figure 2TIC incubation for MDZ phase I and TIC incubation for MDZ phase II; Figure 3 Negative control TIC for phase I incubation and negative control TIC for phase II incubation;

[0050] Figure 4 The chromatograms of phase I and phase II metabolites of midazolam processed by MassLynx software MDF are shown;

[0051] Figure 5 The EIC graphs of MDZ human liver microsomes metabolites in phase I and phase II incubations;

[0052] Figure 6 The chromatograms of phase I and phase II metabolites of midazolam processed by the MDF algorithm programmed in Python;

[0053] Figure 7 MS before MDF and NRF treatment 1 picture;

[0054] Figure 8 MS after MDF and NRF treatment 1 picture;

[0055] Fig. 9 It is the MDZ cleavage pathway;

[0056] Fig.10 This is the EIC diagram of MDZ metabolites in phase I incubation of human liver microsomes;

[0057] Fig.11 This is the EIC diagram of MDZ metabolites in phase II incubation of human liver microsomes;

[0058] Fig.12 for the MDZ metabolic pathway;

[0059] Fig.13 M 1 MS molecular network diagram, fragmentation diagram, and mirror image;

[0060] Fig.14 This is the MS molecular network diagram of midazolam metabolites;

[0061] Fig.15 This is the secondary mass spectrum of CVB-D standard;

[0062] Fig.16 is the CVB-D ion trap mass spectrum;

[0063] Fig.17 It is the CVB-D cleavage pathway;

[0064] Fig.18 Blank plasma chromatograms for rat injections;

[0065] Fig.19 Chromatogram of plasma metabolite TIC injected in rats;

[0066] Fig. 20 XIC plots of plasma metabolites injected into rats;

[0067] Fig.21 This is a chromatogram processed using MassLynx V4.1 software MDF;

[0068] Fig. 22 Chromatograms for MDF processing using Python programming;

[0069] Fig.23 The MS molecular network diagram, fragmentation diagram, and mirror image diagram of the metabolite M2-1;

[0070] Fig.24 MS molecular network of metabolites in plasma samples of rats injected with CVB-D. DETAILED DESCRIPTION

[0071] The technical solution of the present invention will be described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. 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.

[0072] In the following specific embodiments, midazolam (chemical formula C 18 H 13 CIF 3 ) and Cyclovirobuxinum D (CVB-D, chemical formula C 26 H 46 N 2 O) as an experimental object to analyze its metabolites. Among them, midazolam has been reported in the literature for its phase I and phase II metabolites. It is used as a model compound to compare whether the metabolites identified by the analytical method of the present invention are consistent with those reported in the known literature, so as to verify the reliability of the analytical method provided by the present invention. Then, the same analytical method is used to obtain the phase I and phase II metabolites and metabolic pathways of cyclovirobuxine D.

[0073] In the following specific implementation manner, the main instruments and equipment used are shown in Table 1 below, and the main reagents and drugs are shown in Table 2 below.

[0074] Table 1 Main instruments and equipment

[0075] Water bath constant temperature oscillator Tianjin Sateris Laboratory Analytical Instruments Manufacturer XH-C Vortex Mixer Changzhou Yuexin Instrument Manufacturing Co., Ltd. MiniSpin Centrifuge Eppendorf, Germany OASIS HLB Column Waters Corporation Nitrogen purge concentrator Tokyo Rikakki Co., Ltd. Ultrasonic Cleaner Kunshan Ultrasonic Instrument Co., Ltd. XEVO G2-S Q-TOF Tandem Mass Spectrometer Waters Corporation Orbitrap ion trap mass spectrometer Thermo Fisher Scientific, USA

[0076] Table 2 Main reagents and drugs

[0077] Cyclovirobuxine D (purity ≥ 98%) Sichuan Weiqi Biotechnology Co., Ltd. NADPH (purity ≥ 95%) Shanghai Jizhi Biochemical Technology Co., Ltd. UDPGA (purity ≥ 98%) Beijing Xinde Element Technology Co., Ltd. Alamethoxazole (purity> 90%) Nanjing Yifeixue Biotechnology Co., Ltd. Rat liver microsomes (20 mg / mL) Corning Corporation Human liver microsomes (20 mg / mL) Corning Corporation Sodium Carboxymethyl Cellulose Shanghai Haohong Biopharmaceutical Technology Co., Ltd. Trifluoroacetic acid Sinopharm Group Chemical Reagent Co., Ltd. EDTA-2K Shanghai BiDe Pharmaceutical Technology Co., Ltd. Magnesium chloride Beijing Bailingwei Technology Co., Ltd. Huang Yang Ning Tablets Kaifeng Mingren Pharmaceutical Co., Ltd. PFG Beijing Xuanze Weiye Technology Co., Ltd. Normal saline Shandong Qidu Pharmaceutical Co., Ltd. Ethanol Yantai Yuandong Fine Chemical Co., Ltd. Methanol (chromatographic grade) Sigma Acetonitrile (chromatographic grade) Sigma PBS buffer Beijing Coolaibo Technology Co., Ltd.

[0078] In the following specific implementation manner, the experimental reagents are prepared as follows:

[0079] 1. Cyclovirobuxinum D (CVB-D, chemical formula C 26 H 46 N 2 O) Standard solution

[0080] Take 1 mg of cyclovirobuxine D and 1 mL of methanol to prepare a 1 mg / mL cyclovirobuxine D standard solution;

[0081] Take 80.5 μL of cyclovirobuxine D standard solution and 119.5 μL of 0.1 M PBS (pH=7.4) solution to prepare a 1 mM cyclovirobuxine D working solution.

[0082] 2. Midazolam (chemical formula C 18 H 13 CIF 3 )Standard Solution

[0083] Take 1 mg of midazolam and 1 mL of methanol to prepare a 1 mg / mL midazolam standard solution;

[0084] Take 65.2 μL of midazolam standard solution and 134.8 μL of 0.1 M PBS (pH=7.4) solution to prepare a 1 mM midazolam working solution.

[0085] 3. 20 mM MgCl 2 Solution: Take 1.9 mg MgCl 2 The solution was prepared by mixing with 1 mL of 0.1 M PBS (pH = 7.4).

[0086] 4. NADPH solution (prepared before use): 1.9 mg NADPH and 51 μL 0.1 M PBS (pH = 7.4) solution were mixed and prepared.

[0087] 5. UDPGA solution (prepared before use): 1.68 mg UDPGA and 52 μL 0.1 M PBS (pH=7.4) solution were mixed and prepared.

[0088] Example 1: In vitro incubation of liver microsomes with midazolam and CVB-D

[0089] 1. Incubation of Phase I Metabolites of Midazolam and CVB-D

[0090] (1) Accurately pipette 432.5 μL of 0.1 M PBS buffer (pH = 7.4) and 50 μL of 50 mM NADPH solution into a 1.5 mL EP tube and mix well; add 12.5 μL of rat liver microsomes (20 mg / mL) to two of the four replicates and 12.5 μL of human liver microsomes (20 mg / mL) to the remaining two replicates.

[0091] After vortex mixing, the cells were placed in a 37°C water bath thermostat and pre-incubated for 5 min.

[0092] (2) Add 5 μL of 1 mM CVB-D solution to the rat liver microsome incubation medium and the human liver microsome incubation medium to initiate the reaction, and continue incubation at 37°C for 90 min.

[0093] Similarly, 5 μL of 1 mM midazolam solution was added to the above two incubation solutions to start the reaction, and the incubation was continued at 37°C for 5 min.

[0094] (3) After the incubation, 500 μL of ice-cold 10% trifluoroacetic acid was added to terminate the reaction and vortexed for 5 min. The mixture was centrifuged at 10,000 rpm for 10 min, and the supernatant was collected to obtain liver microsome phase I incubation samples containing midazolam and CVB-D phase I metabolites.

[0095] Rat and human liver microsomes were boiled in a 100°C water bath for 15 min, and a similar incubation system as above (the only difference being that it did not contain NADPH) was added. After vortex centrifugation under the same conditions, the supernatant was taken as a negative control.

[0096] 2. Incubation of Phase II Metabolites of Midazolam and CVB-D

[0097] (1) Accurately pipette 327.5 μL of 0.1 M PBS buffer (pH = 7.4) into a 1.5 mL EP tube and add 50 μL of 20 mM MgCl 2 solution, 5 μL 2.5 mg / mL alamethicin solution, 50 μL 50 mM NADPH solution, and 50 μL 50 mM UDPGA solution, mixed evenly; four parallel batches were prepared, with 12.5 μL rat liver microsomes (20 mg / mL) added to two of them and 12.5 μL human liver microsomes (20 mg / mL) added to the remaining two.

[0098] After vortex mixing, the cells were placed in a 37°C water bath thermostat and pre-incubated for 5 min.

[0099] (2) Add 5 μL of 1 mM CVB-D solution to the rat liver microsomes and human liver microsomes incubation medium to initiate the reaction, and continue incubation at 37°C for 90 min.

[0100] Similarly, 5 μL of 1 mM midazolam solution was added to the above two incubation solutions to start the reaction, and the incubation was continued at 37°C for 20 min.

[0101] (3) After the incubation, 500 μL of ice-cold 10% trifluoroacetic acid was added to terminate the reaction, vortexed for 5 min, and centrifuged at 10,000 rpm for 10 min. The supernatant was collected to obtain the liver microsome phase II incubation samples containing midazolam and CVB-D phase I metabolites.

[0102] Another sample of liver microsomes was boiled in a 100°C water bath for 15 min, and a similar incubation system as described above (the only difference being that it did not contain NADPH) was added. The supernatant was taken as a negative control after vortex centrifugation under the same conditions.

[0103] 3. Processing of liver microsome incubation samples

[0104] Liver microsome incubation samples were processed using HLB columns.

[0105] First, the HLB column was activated with 1 mL of methanol and equilibrated with 1 mL of water, and then the sample was loaded into the HLB column;

[0106] After washing the water-soluble impurities with 1 mL of water, the sample was eluted with 1 mL of methanol, and the eluate was collected.

[0107] The eluate was dried using a constant flow of nitrogen, reconstituted by adding 200 μL of methanol-water solution (1:1, V / V), vortexed for 5 min, and sonicated for 5 min until the sample was completely dissolved. The sample was centrifuged at 10,000 rpm for 10 min, and the supernatant was taken for sampling and analysis.

[0108] Example 2: Metabolism of CVB-D in rats and UHPLC-Q-TOF MS / MS process

[0109] 1. Animal experiments

[0110] Twelve Wistar rats (200-250 g), half male and half female, were purchased from Jinan Pengyue Experimental Animal Breeding Co., Ltd., license number SCXK (Lu) 20190003. The experimental animal research was reviewed by the Ethics Committee, with the approval number 2019SDL092.

[0111] According to the administration method, the rats were divided into an injection group and an oral administration group (6 rats in each group, half male and half female), and were adaptively raised in the experimental environment for one week. Blank blood, urine and feces samples were collected before the experiment. Before administration, all rats were fasted for 12 hours, with unlimited drinking water.

[0112] Injection group: rats were injected with CVB-D injection (PEG: 0.9% saline: ethanol, 5:4:1, v / v / v, concentration: 0.8 mg / mL) via tail vein, with a dose of 2.5 mg / kg. After administration, the rats were placed in metabolic cages, and urine and feces were collected at 0-4, 4-8, 8-12, and 12-24 h. Blood was collected from the orbit at 0.5, 1, 2, 4, 8, and 12 h, and 300 μL was placed in a K-filtered jar. 2 The collected whole blood was centrifuged at 3500 rpm for 5 min to obtain plasma samples. Urine, feces and plasma samples were temporarily stored in a -20℃ refrigerator for further processing.

[0113] Intragastric administration group: Rats were intragastricly administered with Huang Yang Ning tablets (ground into powder and dissolved in 0.5% CMC) at a dose of 10 mg / kg (equivalent to the labeled amount of CVB-D in Huang Yang Ning tablets). After administration, the rats were placed in metabolic cages, and urine and feces were collected at 0-4, 4-8, 8-12, and 12-24 h. Blood was collected from the orbits at 0.5, 1.5, 3, 6, and 12 h, and 300 μL was placed in a K-filtered jar. 2 The collected whole blood was centrifuged at 3500 rpm for 5 min to obtain plasma samples. Urine, feces and plasma samples were temporarily stored in a -20℃ refrigerator for further processing.

[0114] 2. Biological sample processing

[0115] (1) Plasma: Equal amounts of plasma samples at each time point were mixed, 200 μL of mixed plasma was taken, protein was precipitated with 5 times ice-cold acetonitrile, vortexed for 5 min, and then centrifuged at 10,000 rpm for 10 min at 4°C. The supernatant was blown dry under a nitrogen stream and reconstituted with 200 μL acetonitrile-water (1:1, v / v). Vortexed for 5 min, sonicated for 5 min, and then centrifuged at 10,000 rpm for 10 min. The supernatant was taken for sampling and analysis.

[0116] (2) Urine: Centrifuge the urine sample at 12000 rpm for 10 min, and take the supernatant to remove food residues. After mixing equal amounts of urine samples from different time periods, use an HLB column for treatment. Pretreat the HLB column with 1 mL of methanol and 1 mL of water, respectively. Take 1 mL of urine sample and load it into the HLB column. After washing with 1 mL of water, elute the sample with 1 mL of methanol. Then blow dry the eluate with nitrogen at room temperature, and add 200 μL of acetonitrile-water (1:1, v / v) to reconstitute it. Vortex for 5 min, sonicate for 5 min, and then centrifuge at 10000 rpm for 10 min. Take the supernatant for sampling and analysis.

[0117] (3) Feces: After removing food residues from the feces, place them in an oven at 60°C and dry them. Then grind them into powder and mix equal amounts of feces powder from different time periods. Take 0.3 g of feces powder and add 1 mL of methanol for ultrasonic extraction for 10 min. Centrifuge at 10,000 rpm for 10 min. Take the supernatant, blow dry with nitrogen, and add 200 μL of acetonitrile-water (1:1, v / v) for re-dissolution. Vortex for 5 min, ultrasonicate for 5 min, and centrifuge at 10,000 rpm for 10 min. Take the supernatant for sampling and analysis.

[0118] 3. UHPLC-Q-TOF MS / MS conditions

[0119] (1) Chromatographic conditions

[0120] An ACQUITY UHPLC CSH C18 chromatographic column (1.7 μm, 2.1 mm×100 mm) was used; the column temperature was 45°C; the mobile phase A was 0.1% formic acid in water; the mobile phase B was acetonitrile; the flow rate was 0.3 mL / min; the injection volume was 10 μL; the gradient elution program was: 0-2.5 min 15% B, 2.5-25 min 15-95% B, 25-28 min 95% B, 28-28.01 min 95-15% B, 28.01-32.00 min 15% B.

[0121] (2)Q-TOF conditions

[0122] ESI-MS positive ion mode full scan, ion source temperature, 120℃; capillary voltage, 3.0kV; cone voltage, 40V; desolvation gas temperature, 500℃; desolvation gas flow rate, 50.0L / Hr; scanning frequency, 20sec; collision energy, 6V; scanning range, 50-1200Da.

[0123] (3) Ion trap conditions

[0124] ESI-MS scan in positive ion mode, mass spectrometry conditions are the same as 2.3.2; parent ion: 403.3; daughter ions: 385.3, 354.3, 323.2, 372.3, 302.2.

[0125] Example 3: Intelligent mass spectrometry data (MS data) processing based on Python

[0126] The UHPLC-Q-TOF-MS / MS raw data were first converted into the mzML universal format by the MSConvert function of Proteo Wizard 3.0 software so that Python could read, process, and visualize the data.

[0127] PyCharm 2021.2 software is configured with Python 3.8 environment to write intelligent MS data processing programs and read MS data based on the pyopenms open source package. MS data is processed algorithmically based on databases such as numpy and pandas. The workflow diagram of drug metabolite characterization strategy based on Python intelligent MS data processing program combined with MS molecular network is shown in the figure below. Figure 1 shown.

[0128] The Python intelligent MS data processing program mainly includes: mass defect filtering (MDF), nitrogen rule filtering (NRF), point integration (DP), characteristic ion filtering (CIF), metabolite prediction and neutral loss filtering (NLF), as well as data visualization process.

[0129] 1. Mass Deficit Filtration (MDF) and Nitrogen Regulation Filtration (NRF)

[0130] (1) Mass Defect Filter (MDF)

[0131] The Accurate Mass Filter function in MassLynx V4.1 software was used to set the parameters: Function = 1; mass range: 326 ± 50 Da, 502 ± 50 Da; mass defect window: 0.0855 ± 50 mDa, 0.1181 ± 50 mDa, and mass defect filtering (MDF) was performed on the midazolam metabolism data.

[0132] The following parameters were set: Function = 1; mass range: 403 ± 50 Da, 579 ± 50 Da; mass defect window: 0.3683 ± 50 mDa, 0.4004 ± 50 mDa, and the MS data from CVB-D in vitro metabolism were subjected to MDF processing.

[0133] The following parameters were set: Function = 1; mass range: 403-50Da to 403+100Da, 579±50Da; mass defect window: 0.3683±50mDa, 0.4004±50mDa, and MDF processing was performed on the MS data from the in vivo metabolism of CVB-D.

[0134] The same MDF operation can be achieved by editing the above parameter window in the programming program and entering the same parameter values ​​as above.

[0135] The MDF chromatograms of each metabolic data were obtained through the above operations to verify the feasibility of the MDF algorithm written in Python.

[0136] (2) Nitrogen Ratio Filtration (NRF)

[0137] Midazolam and DCVB-D have 3 and 2 nitrogen atoms, respectively. In the positive ion mode of ESI-MS, according to the nitrogen rule [M+H] + m / z are odd and even numbers respectively. 1 Medium [M+H] + The parity of the integer part of m / z is used to implement nitrogen rule filtering (NRF).

[0138] 2. Analysis of Point Integration (DP) Algorithm

[0139] The DP algorithm is a similarity matching algorithm used to evaluate the MS of parent drugs and metabolites. 2 Spectral similarity and fragment ion similarity. In this MS data processing program, the DP algorithm is implemented by writing the following formulas (1) and (2):

[0140] W i =[intensity of peak i ] m [m / z of peak i ] n (1)

[0141] Select the top 50 peak intensity ion peaks to calculate formula (1), where W i represents the peak intensity normalization value, m=0.5, n=2. Peak intensity normalization can increase the proportion of high-quality peaks to reduce the interference of low-intensity noise peaks.

[0142]

[0143] In formula (2), W u and W ai represent the normalized values ​​of the peak intensities of the measured spectrum and the reference spectrum, respectively, and the ∑ symbol represents the sum.

[0144] The DP value is between 0 and 1. The closer the DP value is to 1, the higher the spectral similarity. In this screening, MS with DP ≥ 0.8 is retained. 2 and the corresponding MS 1 .

[0145] 3. Characteristic ion filtering (CIF), metabolite (including fragmentation pathway) prediction and neutral loss filtering (NLF) summarize the metabolites of midazolam that have been characterized and their molecular weight information, as shown in Table 1.

[0146] Table 1 Midazolam metabolites and their molecular weights

[0147]

[0148] For information on the metabolites of midazolam that have been characterized and their molecular weights, sources include but are not limited to:

[0149] "A highly sensitive LC-MS-MS assay for analysis of midazolam and its major metabolite in human plasma: applications to drug metabolism", (VAJabor, EBCoelho, NADos Santos, PSBonato, VLLanchote; J. Chromatogr. BAnalyt. Technol. Biomed. Life Sci., 822 (2005) 27-32).

[0150] "In vitro and in vivo glucuronidation of midazolamin humans" (RuthHyland, Toby Osborne, Anthony Payne, Sarah Kempshall, Y.Raj Logan, KhaledEzzeddine3&Barry Jones, British Journal of Clinical.Pharmacology.67 (2009) 445-454);

[0151] "Mechanistic Modeling to Predict Midazolam Metabolite Exposure fromIn Vitro Data" (Drug Metab. Dispos, 44 (2016) 781-791)

[0152] According to the biotransformation rules known in the art and Biotransformer 3.0, the molecular weight changes of potential metabolites of CVB-D were predicted and summarized in Table 2 below.

[0153] Table 2 Predicted CVB-D metabolites and their molecular weights

[0154]

[0155] Table 1 and Table 2 were used as training data sets to build metabolite prediction models, and the metabolite prediction models were used to predict the MS data set.

[0156] The method for constructing the metabolite prediction model is as follows: the precise molecular weight of the original drug ± the precise molecular weight change produced by biological conversion, all the precise molecular weights obtained, all the precise molecular weight outputs are used as the training data set, the metabolic MS data are traversed, and the molecular weight that can be traversed in the data set is used as the metabolite prediction model.

[0157] Based on the output of the model, the possible metabolites in the MS data are determined. The training data set can be changed to predict the metabolites of other compounds.

[0158] The fragmentation pathway of CVB-D was inferred based on the fragment ions of CVB-D and related references. The MS molecular network of the parent drug was constructed based on the fragmentation pathway inferred by mass spectrometry fragmentation analysis. The MS molecular network of the parent drug showed that CVB-D had characteristic ions of m / z 121.10 and 135.11. When there were characteristic ions, characteristic ion filtering (CIF) was triggered. By identifying the characteristic ions, the MS that may be derived from CVB-D metabolites was screened. 2 picture.

[0159] Midazolam has no characteristic ions and will not trigger this step of data processing.

[0160] Neutral loss filtering (NLF) screened out the combined metabolites such as glucuronidation and sulfation by calculating the neutral loss (176.0321 Da for glucuronidation and 79.9568 Da for sulfation).

[0161] 4. Data Visualization

[0162] The mirror image of the MS molecular network is drawn using the numpy and matplotlib libraries.

[0163] The mirror image more intuitively reflects the similarities and differences between the main fragment ions of metabolites and parent drugs.

[0164] 5. Mass spectrometry data processing results

[0165] Total ion current chromatograms (TIC) of midazolam human liver microsome phase I and phase II incubation samples and negative control samples are shown in Figure 2. Figure 2 and Figure 3 As shown in Figure 2, the midazolam metabolite peak is masked by a large number of endogenous matrix interference peaks and is not obvious in TIC.

[0166] The results of MDF processing of midazolam human liver microsomal phase I and phase II metabolism data using MassLynx V4.1 software are as follows: Figure 4 As shown in the figure above and below, it can be seen that after MDF processing, two metabolite peaks in the phase I metabolism data are displayed from the complex biological matrix peaks. Figure 5As shown in the figure above, the retention time of the metabolite peaks in the extracted ion current chromatogram (EIC) is consistent with that of the metabolite peaks, and one metabolite peak is still buried in the biological matrix peak. Figure 5 As shown in the figure below, the three metabolite peaks in the phase II metabolism data are all separated from the biological matrix peaks, and the retention time is consistent with the metabolite peaks in the EIC. The results show that MDF can remove most of the endogenous interference from the biological matrix and effectively distinguish the metabolite ion peaks from the biological matrix ion peaks.

[0167] Figure 6 (The upper and lower figures) are the results of the MDF programming algorithm processing the human liver microsome phase I and phase II metabolism data of midazolam. Figure 4 The following figure and Figure 6 Above, and Figure 5 The above picture and Figure 6 As can be seen in the figure below, the retention time and peak shape of each chromatographic peak are the same, which verifies the feasibility and accuracy of the MDF algorithm written in Python. 1 Figure Figure 7 As shown, the MS 1 Contains numerous and chaotic ion peaks. Figure 8 MS after MDF and nitrogen filtration at the same retention time 1 Figure 1 screened and retained a very small number of molecular ion peaks that may come from metabolites, eliminating the interference of a large number of non-metabolite molecular ion peaks. This result also proves the effectiveness of the MDF and nitrogen law filtering algorithms.

[0168] Midazolam standard MS 2 The DP algorithm retained the sample spectra with a DP value ≥ 0.8 compared with the reference spectrum. Phase I metabolite data of midazolam incubated in human liver microsomes obtained by Q-TOF MS / MS analysis (MS after data processing) 1 There are 2833 mass spectra in the original mass spectrometry data. After DP algorithm processing, only 590 MS spectra with high similarity to the reference spectra are retained. 2 and its corresponding MS 1 Phase II metabolite data of midazolam (MS after data processing) 1 Among the 2833 mass spectra, 140 mass spectra were retained after NLF processing, and only 48 mass spectra with high similarity to the reference spectra remained after further processing by the DP algorithm. It can be seen that the DP algorithm greatly narrowed the metabolite search range and further improved the accuracy of metabolite characterization.

[0169] The error range was allowed to be within 10ppm, and two types of metabolites were predicted in the phase I and phase II metabolism data of midazolam, respectively. According to the mass spectrometry data and fragment ions processed by the above algorithm, the specific metabolites were further characterized in combination with the MS molecular network.

[0170] Example 4: MS molecular network construction, analysis, and qualitative analysis of MDZ metabolites

[0171] The fragment ion information of the parent drug and the fragment ion information of the metabolite (i.e., MS 2 Spectral data were visualized using Cytoscape software, which used nodes to describe the fragmentation relationship between the parent drug and metabolites, and searched for characteristic ions of the compounds based on the network relationship.

[0172] Specifically, the mass spectrometry data and fragment ion information obtained by the Python intelligent MS data processing program were entered into an Excel spreadsheet and then imported into the Cytoscape software. After sorting, a visual MS molecular network diagram was obtained for further confirmation of compound metabolites. At the same time, the fragment ion relationship network diagram between the parent drug and metabolites was improved, each metabolite ion fragment was assigned, and the metabolic relationship between the metabolite and the parent drug was clarified.

[0173] 1. Fragmentation pathway, MS molecular network and qualitative analysis of midazolam

[0174] (1) Cleavage pathway

[0175] like Fig. 9 As shown in the figure, the cleavage pathway of midazolam has been clearly characterized. It can be seen that the chlorine atom first breaks to produce a molecular ion at m / z 291, and then the five-membered ring and the seven-membered ring break to produce two ions at m / z 209 and 250 respectively. In addition, the direct cleavage of the five-membered ring of midazolam eliminates a molecule of CH 3 CN produces a fragment ion of m / z 285, which is then further eliminated by one molecule of HCN to produce a fragment ion of m / z 258, and in the process, the loss of a chlorine atom also produces fragment ions of m / z 250 and 223. Direct cleavage of the seven-membered ring of midazolam produces a fragment ion of m / z 209.

[0176] (2) MS molecular network and qualitative analysis

[0177] Based on Python intelligent MS data processing results, MS molecular network, midazolam fragmentation diagram, and fragment ion information, a total of 6 midazolam metabolites were identified (see Table 3, Fig.10 , Fig.11 ); including 3 phase I metabolites and 3 phase II metabolites, the specific metabolic pathways are shown in Fig.12 .

[0178] Table 3 MDZ hepatic microsomal metabolites

[0179]

[0180] The intelligent MS data processing integrated with the MS molecular network strategy constructed in this specific embodiment accurately identified 6 midazolam metabolites, which are consistent with the midazolam metabolites that have been reported and clearly characterized in the literature. This shows that the MS data processing method provided by the present invention can be used for the mining of potential drug metabolites and can effectively realize the qualitative analysis of metabolites.

[0181] (3) Analysis results of metabolite M1

[0182] Taking the metabolite M1 of midazolam as an example, the metabolite characterization process and MS molecular network analysis are described in detail below:

[0183] In the MS molecular network, each ion is represented by a node, where the orange and blue nodes represent midazolam and M1 parent ions, respectively, and the pink nodes are fragment ions.

[0184] The retention time of M1 is 6.399min, and the quasi-molecular ion peak [M+H] + The m / z is 358.0777, which is 32Da higher than that of protonated midazolam, that is, two oxygen atoms are added. According to the MS molecular network diagram ( Fig.13 From the upper left figure, we can see that: (1) M 1 The fragment ions m / z 260.0275 and 225.0570 are derived from the midazolam fragment ions m / z 244.0324 and 209.063, respectively, with an increase of 16 Da, which are the fragment ions generated after the seven-membered ring is broken; (2) M 1 The fragment ions m / z 239.0762, 274.0451, 301.0459, and 266.0856 are derived from the midazolam fragment ions m / z 223.0792, 258.0480, 285.0589, and 250.0901, respectively, with an increase of 16Da. They are all fragment ions generated after the five-membered ring is broken. It is inferred that one hydroxylation occurs in the seven-membered ring or benzene ring, and the other hydroxylation occurs at the C' position. 1 The difference between the fragment ion 323.1165 and the fragment ion m / z 291.1166 of midazolam is 32Da, which further indicates that M 1 It is a product of midazolam undergoing two hydroxylations. Fig.13 The following figure shows M visualized by Python 1 The mirror image of the secondary mass spectrum and the secondary mass spectrum of midazolam shows that the metabolite ions in the mirror image have mass deviations of 16Da and 32Da relative to the parent drug ions.

[0185] Fig.14 This is the MS molecular network diagram of midazolam metabolites. There are 6 metabolite clusters in the figure, representing 6 midazolam metabolites. Each metabolite cluster is composed of its fragment ions, and the fragmentation relationship between the fragment ions constitutes the network relationship between the nodes. For example, in the M1 metabolite cluster, the cluster nodes represent 7 fragment ions from M1, and there are 7 lines representing the fragmentation pathways and molecular weight changes.

[0186] 2. CVB-D cleavage pathway and metabolite qualitative analysis

[0187] (1) CVB-D cleavage pathway

[0188] Before characterizing CVB-D metabolites in various biological matrices, the fragmentation pathways of CVB-D were first summarized. CVB-D was analyzed by mass spectrometry using QExactive Orbitrap. Fig.15 As shown, the [M+H] of CVB-D was detected in positive ion mode. + The quasi-molecular ion peak is m / z 403.3678, and the main fragment ions include m / z 385.3583, 372.3257, 354.3155, 341.2844, 323.2733, 302.2840, 297.2564, 203.1787, 201.1638, 189.1638, 187.1481, 135.1169, and 121.1013. Combined with the CVB-D ion trap mass spectrum; and speculated that the CVB-D fragmentation pathway, such as Fig.16 shown.

[0189] The protonated ion of CVB-D loses a water molecule neutrally to generate a fragment ion m / z 385.3577, and the A ring further breaks to generate a fragment ion m / z 297.2583. CVB-D C 3 bit or C 20 The cleavage of the methylamino group produces m / z 372.3261, followed by the loss of a water molecule to produce m / z 354.3155, or the cleavage of the A ring to produce m / z 302.2842. 3 , C 20 Simultaneous cleavage of two methylamino groups produces m / z 341.2839. The fragment ion m / z 323.2739 is obtained by the loss of two aminomethyl groups and a water molecule from CVB-D. 12 , C 13 and C 8 , C 14 The single bond between the two rings breaks to produce fragment ions m / z 203.1800 and m / z 121.1018. 11 , C 12 and C8 , C 14 The single bond between the two rings breaks to produce fragment ions m / z 189.1644 and m / z 135.1174. 6 , C 7 , C 9 , C 10 and C 9 , C 19 The single bond cleavage produces fragment ions m / z 135.1174 and m / z 187.1487. 5 , C 6 , C 9 , C 10 and C 9 , C 19 The single bond cleavage between the two molecules produced fragment ions m / z 121.1018 and m / z 201.1643.

[0190] (2) Qualitative analysis of CVB-D metabolites in vivo and in vitro

[0191] According to the Python intelligent MS data processing results, MS molecular network, and CVB-D cleavage pathway, a total of 39 CVB-D metabolites were inferred in rat plasma, urine, feces samples, and human and mouse in vitro liver microsome incubation samples after administration, of which 21 metabolites were identified in in vitro metabolism and 36 metabolites were identified in in vivo metabolism. The molecular mass deviation of all metabolite structural characterizations was controlled within 5ppm. The results are shown in Table 4, and the metabolic pathways are shown in Fig.17 .

[0192] Table 4 Metabolites of CVB-D in vivo and in vitro

[0193]

[0194]

[0195]

[0196] Take the plasma sample obtained after rats were injected with CVB-D standard as an example, the chromatogram is as follows Figure 18-22 shown. Fig.18 Blank plasma chromatograms for rat injections; Fig.19 Chromatogram of plasma metabolite TIC injected in rats;

[0197] Fig. 22 This is a chromatogram processed by Python programming MDF. It can be seen that MDF filters out a large number of chromatographic peaks from biological matrices, and most metabolite peaks are displayed. Finally, a total of 17 metabolites were identified. Each metabolite was identified in MDF ( Fig.21 The retention time of the chromatographic peaks and their extracted ion peaks ( Fig. 20 By comparing Fig.18 , 19 , and found that CVB-D metabolites were detected in the plasma of rats after administration; by comparison Fig.18 , 19 and 21, which proves that the MDF algorithm can filter out a large number of chromatographic peaks from biological matrices. Fig.21 , 22 , proving the consistency between the MDF algorithm results written in Python and the MDF results processed by the software.

[0198] Fig.23 The MS molecular network diagram, cleavage diagram, and mirror diagram of the metabolite M2-1 fragments are shown. In the MS molecular network diagram, the fragment ions at the same cleavage site are connected by dotted lines. From the MS molecular network diagram, it can be observed that the parent ion of M2-1 differs from the parent ion of CVB-D by 14Da, that is, one carbon atom and two hydrogen atoms are lost. The fragment ions of M2-1 m / z 358.3069, 340.2988, 327.2608, and 309.2673 are derived from CVB-D, respectively. The fragment ions of m / z 372.3261, 354.3155, 341.2839, and 323.2739 are reduced by 14Da, corresponding to the fragments of the parent ring after the cleavage of each group on the parent ring; the fragment ions of M2-1 m / z 189.1649 and 175.1433 are derived from CVB-D, respectively. The fragment ions of m / z 203.1800 and 189.1644 are reduced by 14Da, both corresponding to the AB ring fragments produced by the rupture of the C ring; the fragment ions of M2-1 m / z 107.0891 and 283.2593 come from the CVB-D fragment ions 121.1018 and 297.2583, which are reduced by 14 Da, respectively, and both correspond to A-ring fragments. Therefore, it is speculated that CVB-D undergoes 4α-demethylation to produce M2-1.

[0199] according to Fig.23 The molecular network diagram in Figure 3 and the fragmentation pathway of CVB-D show that the fragmentation of CVB-D and M2-1 both have m / z 135.11 and 121.10 fragment ions, which are inferred to be the diagnostic ions of CVB-D and its metabolites. Fig.23 The mirror image also reveals that the fragment ions that undergo metabolic reactions differ by 14 Da, while the fragment ions at the sites where no metabolic reactions occur remain unchanged.

[0200] Metabolite MS molecular network diagram Fig.24 As shown in Figure 2, taking the MS molecular network diagram of plasma metabolites in rats after injection of CVB-D as an example, the network diagram contains 15 clusters representing 15 metabolites, and the extracted ion current chromatogram is shown in Figure 2. Fig. 22As shown in the figure, m / z 121.10 and 135.11 fragment ions can be seen in all 15 metabolite clusters, which can be confirmed as CVB-D characteristic ions. The fragmentation relationship of the fragment ions is further assisted in the confirmation of each metabolite by connecting lines in the cluster.

[0201] Specifically, the qualitative analysis of each metabolite of CVB-D is as follows:

[0202] M1[M+H] + The m / z is 385.3569, which is 18 Da less than that of protonated CVB-D. 2 The fragment ions were the same as those of the parent drug. Therefore, M1 was determined to be the dehydration product of CVB-D.

[0203] M2-1 and M2-2 have the same [M+H] + m / z 389.3526, retention times 28.443min and 28.299min, respectively, which are 14Da less than protonated CVB-D. It is speculated that both are demethylation products of CVB-D. MS 2 In the analysis, they all have fragment ions at m / z 358.32, 340.29, 327.27, and 309.24, which are derived from the sequential cleavage of the hydroxyl group and the aminomethyl group on the rigid ring of CVB-D. The fragment ions of M2-1 at m / z 175.1433 and 107.0892 are 14Da different from the AB ring obtained after the C and B rings of CVB-D, respectively. It is speculated that the demethylation of M2-1 occurs at the C4 position of CVB-D. Similarly, the fragment ions of M2-2 at m / z 173.1763 and 187.1422 are 14Da different from the CD ring obtained after the B ring of CVB-D, respectively. It is speculated that the demethylation of M2-2 occurs at the C13 or C14 position of CVB-D.

[0204] M3-1 and M3-2 are isomers of each other, [M+H] + m / z 399.3370 and CVB-D [M+H] + Compared with the decrease of 4Da, it is speculated that it is the product of two consecutive dehydrogenation of the parent drug. Its fragment ions m / z 368.2902, 350.2993, 337.2660, and 319.2407 are all obtained by the loss of four hydrogen atoms of the CVB-D fragment ion.

[0205] M4-1, M4-2, M4-3, and M5 are isomers of each other, [M+H] +The m / z is 401.3526, which is 2Da less than that of protonated CVB-D, i.e., it loses two hydrogen atoms, and is speculated to be a metabolite of CVB-D dehydrogenation or secondary alcohol to ketone. The M4-1 A ring fragment ion m / z 133.1001 is 2Da different from the CVB-D A ring fragment ion m / z135.1169; the M4-2 CD ring fragment ion m / z199.1515 is 2Da different from the CVB-D CD ring fragment ion m / z 201.1638; the m / z 201.1606 and 187.1466 AB ring fragment ions in M4-3 are 2Da different from the CVB-D AB ring, so it is speculated that the metabolism of M4-1, M4-2, and M4-3 occurs in the A ring, CD ring, and AB ring, respectively.

[0206] The fragment ions m / z 151.1112 and 217.1471 produced by M5 increased by 16Da compared with the fragment ions at the CD ring of CVB-D, adding an oxygen atom. It is speculated that M5 is the hydroxyl group on the D ring of CVB-D oxidized to ketone.

[0207] M6-1, M6-2[M+H] + Compared with protonated CVB-D, it increased by 2Da, suggesting that demethylation and hydroxylation metabolism occurred. The fragment ions of M6-1 at m / z 191.1432 and 205.1436 differed by 2Da from the AB ring fragments of CVB-D, suggesting that metabolism occurred at the C4 position of CVB-D. Similarly, based on the fragment ions of M-2 at m / z 189.1262 and 203.1452, it was suggested that metabolism occurred at the C13 or C14 position of CVB-D.

[0208] M7-1, M7-2 and M8, M9 [M+H] + The m / z 417.3476 and 417.3839 fragments of M7-1 and M7-2 were 14Da higher than those of protonated CVB-D, respectively. The fragment ions m / z 217.1503 and m / z 215.1400 of M7-1 and M7-2 were 14Da higher than those of CVB-D AB ring and CD ring, respectively, suggesting that the metabolism of M7-1 and M7-2 occurred in the AB ring and CD ring, respectively.

[0209] M8 and M9 are metabolites of CVB-D after adding a methylene molecule. The fragment ions of M8, m / z 386.3515 and 355.2812, differ by 14 Da from the fragment ions of CVB-D, m / z 372.3257 and 341.2844. The remaining fragment ions are the same as the fragment ions of CVB-D, so it is speculated that the methylation occurs at the hydroxyl group and breaks with the hydroxyl group. The fragment ion of M9, m / z 316.2776, is a fragment generated after the M9 D ring breaks, and it is speculated that the methylation occurs at the C3 adjacent primary amino group of the D ring.

[0210] M10-1, M10-2, and M11 are isomers of each other, [M+H] + The m / z of both is 419.3632, which is 14Da higher than that of protonated CVB-D, i.e., an oxygen atom is added. The fragment ions of M10-1, m / z 137.0790 and 151.1109, are from the D-ring fragments, and the fragment ions of M10-2, m / z 219.1841 and 205.1589, are from the AB-ring fragments, and they differ by 14Da from the D-ring fragment ions and AB-ring fragment ions of CVB-D, respectively. It is speculated that the hydroxylation of M10-1 and M10-2 occurs in the D-ring and AB-ring, respectively. The fragment ions of M11, m / z 318.2420, 401.3375, and 370.3244, differ by 16 Da from the fragment ions of CVB-D, m / z 302.2840, 385.3583, and 354.3155. The remaining fragment ions are the same as those of CVB-D, so it is speculated that the C3 ortho-secondary amine is metabolized to hydroxylamine.

[0211] M12-M14[M+H] + Compared with the protonated CVB-D, which increases by 30Da, M12 and 13 are isomers of CVB-D with two oxygen atoms added and two hydrogen atoms reduced. The fragment ions m / z 233.1512, 219.1245, and 165.0925 of M12-1 differ by 30Da from the fragment ions from the AB ring of CVB-DA, the fragment ions m / z 233.1610 and 219.1258 of M12-2 differ by 30Da from the fragment ions from the ABC ring of CVB-D, and the fragment ions m / z 217.1593 and 231.1393 of M12-3 differ by 30Da from the fragment ions from the BCD ring of CVB-D. It is speculated that M12-1, M12-2, and M12-3 are the metabolites of hydroxylation and ketone formation of the AB ring, ABC ring, and BCD ring of CVB-D, respectively. Similarly, based on the fragment ions m / z 219.1237, 165.0876, 151.0787 from M13-1 and m / z 217.1358, 231.1497 from M13-2, it is speculated that M13-1 and M13-2 are metabolites of the oxidation of the C4 methyl group of CVB-D to carboxylic acid and the oxidation of the C13 or C14 methyl group of CVB-D to carboxylic acid, respectively. M14-1, 14-2 [M+H] + Both are m / z 433.3789, which adds one molecule of oxygen and one molecule of methylene compared with protonated CVB-D. The fragment ions from M14-1 and M14-2 are m / z 219.1737 and m / z 217.1482, which differ by 30Da from the CVB-D fragment ions m / z 189.1638 and 187.1481, respectively, indicating that the metabolism of M14-1 and M14-2 occurs in the AB ring and CD ring, respectively.

[0212] M15-1, M15-2[M+H] + Both were m / z 435.3581, which was 32Da higher than that of protonated CVB-D, i.e., two oxygen atoms were added. It was speculated that both were products of two consecutive hydroxylations. The fragment ions m / z 235.1636 and 221.1570 of M15-1 differed from the AB ring fragment ions m / z 203.1787 and 189.1638 of CVB-D, and the fragment ions m / z 219.1479 and 233.1655 of M15-2 differed from the CD ring fragment ions m / z 203.1787 and 189.1638 of CVB-D by 32Da, indicating that the metabolic reactions of M15-1 and M15-2 occurred in the AB ring and CD ring, respectively.

[0213] M16-1 and M16-2 are isomers of each other, [M+H] + Both are m / z 451.3530, which is 48Da higher than protonated CVB-D, that is, three oxygen atoms are added. It is speculated that both are three consecutive hydroxylation metabolites. The difference of 46Da between the fragment ions m / z 251.1611 and 237.1457 of M16-1 and the fragment ions m / z 203.1787 and 189.1638 of CVB-D AB ring is 46Da, indicating that the metabolic reaction of M16-1 occurs in the AB ring and the metabolic reaction of M16-2 occurs in the CD ring.

[0214] M17-M26 are phase II metabolites of CVB-D, and glucuronidation and sulfation are the main phase II metabolic reactions of CVB-D. M17 and M20-M26 are all glucuronidation metabolites, all of which lose 176Da neutrally, that is, lose one glucose neutrally, and the m / z after neutral loss are 403.3569, 435.3544, 419.3608, 401.3683, 417.3427, and 399.3397, respectively. M18 and M19 are sulfation metabolites, all of which lose 80Da neutrally, and the m / z after neutral loss are 403.3668 and 419.3627, respectively.

[0215] LC-HR-MS coupling instrument has become the preferred analytical tool for detecting and identifying metabolites due to its high sensitivity and strong selectivity, but the analysis of MS data is still a complex and cumbersome task. The drawback of traditional MS analysis is that it is time-consuming on the one hand, and on the other hand, most MS analysis software is provided by instrument suppliers, which is not only expensive, but also such software can only process the MS data provided by the corresponding analytical instrument, and has data limitations. Therefore, the present invention provides a drug metabolite characterization method based on Python artificial intelligence MS data processing combined with MS molecular network, which can be used for the processing of MS data in the mzML general format, and saves a lot of time for MS analysis through various algorithm processing, and further performs qualitative analysis on drug metabolites, and finally visualizes the qualitative results through MS molecular network diagram and mirror diagram.

[0216] This specific implementation method uses midazolam as a model compound, and verifies the effectiveness and accuracy of the integration strategy by processing the metabolic data of human liver microsomes and mouse liver microsomes of midazolam. In the experiment, the results of each algorithm in the integration strategy were analyzed step by step: the accuracy of the MDF algorithm was verified by comparing the processing results of MassLynx V4.1 software, and the effectiveness of the nitrogen law filtering, DP, and NLF algorithms was verified by comparing the results before and after processing. According to the data processing results combined with the clearly described fragmentation route of midazolam and the MS molecular network, 6 metabolites of midazolam were finally identified, including 3 phase I metabolites and 3 phase II metabolites. The metabolites of midazolam obtained by analysis are consistent with the literature reports mentioned above, confirming that the integration strategy can be applied to the characterization of potential drug metabolites.

[0217] In summary, UHPLC-Q-TOF-MS technology was used to detect and analyze biological matrix samples (plasma, urine, feces) of rats before and after administration, as well as human liver microsomes and rat liver microsome incubation samples, and the data processing and analysis method of the present invention was used to process and analyze the original MS data. Before characterizing CVB-D metabolites, the CVB-D cleavage pathway was first inferred based on the secondary mass spectrometry, ion trap mass spectrometry and known literature of CVB-D standards, such as Fig.19As shown. Subsequently, combined with the MS molecular network, it was determined that CVB-D had characteristic ions of m / z 121.10 and 135.11, which were used to distinguish CVB-D metabolites from other compounds. Finally, a total of 39 CVB-D metabolites were identified through this research strategy, of which 21 metabolites were identified in in vitro metabolism and 36 metabolites were identified in in vivo metabolism. The identification results are shown in Table 4 above. In vitro metabolism experiments can quickly screen out the main metabolites of drugs, while in vivo metabolism experiments can simulate the complex metabolic environment in the body (such as the presence of multiple enzymes, cofactors and transporters), and take into account the metabolites produced by the first-pass effect of drugs, so more types of metabolites are detected in in vivo metabolism experiments.

[0218] The artificial intelligence MS data processing method constructed in this specific embodiment can be successfully applied to the characterization of midazolam and CVB-D drug metabolites. It not only clearly explains the metabolic pathway of CVB-D, but also provides new ideas and methods for the characterization and analysis of drug metabolites.

[0219] The above specific embodiments describe the implementation of the present invention in detail, but the present invention is not limited to the specific details in the above embodiments. Within the scope of the claims and technical concept of the present invention, the technical solution of the present invention can be modified and changed in many simple ways, and these simple modifications all belong to the protection scope of the present invention.

Claims

1. A method for analyzing drug metabolites, characterized in that: The following steps are involved: Taking the in vitro / in vivo metabolites of the drug to be tested for radioactive isotope flux detection; the drug to be tested is a non-radioactive isotope tracer drug or a radioactive isotope tracer drug; if the drug to be tested is a radioactive isotope tracer drug, the corresponding metabolite information is detected; Liquid chromatography-high-resolution mass spectrometry detection is performed to obtain raw data in a universal format, and the metabolites of radiolabeled drugs are located at the same retention time of the primary mass spectrometry peak; The raw data in the general format were filtered for radioisotopes and mass defects to obtain the MS of metabolites. 1 Spectra and MS 2 Atlas; The MS of the parent drug and metabolites were compared using a point integration algorithm. 2 The similarity of the spectra and the fragment ions are used to screen and retain the MS 2 Spectra and corresponding MS 1 Atlas; Predict the molecular weight changes of potential metabolites of the drug to be tested, and use this as a training data set to build a metabolite prediction model; 1 The spectral data is used for prediction, and the possible existence of metabolites is determined according to the output results of the metabolite prediction model; Combined with MS 2 The spectral data is used to infer the fragmentation pathway of the drug to be tested, and the MS molecular network of the parent drug is constructed according to the fragmentation pathway to determine whether there are characteristic ions; By calculating MS 1 Neutral loss of spectral data screened out bound metabolites; According to the MS molecular network of the parent drug and MS 2 Establishing the metabolic relationship between the parent drug and the metabolite fragments based on the spectral data, constructing a microscopic MS molecular network, and confirming the characteristic ions based on the microscopic MS molecular network; The point integration algorithm is used for analysis as follows: select the top 50 peak intensity ion peaks and use the following formula (1) to calculate the peak intensity normalization value W: i ; （1） Among them, W i represents the normalized value of peak intensity, m and n represent the mass weight and intensity scaling factor, respectively; The DP value is calculated using the following formula (2): （2） Among them, W u and W ai represent the normalized values ​​of the peak intensities of the measured spectrum and the reference spectrum, respectively.

2. The method for analyzing drug metabolites according to claim 1, characterized in that: The mass defect filtering window was set to ±50 mDa for mass defect filtering; 1 In the spectral data [M+H] + Parity of the integer part of m / z.

3. The method for analyzing drug metabolites according to claim 1, characterized in that: Whether nitrogen filtering is performed is determined based on whether the chemical formula of the drug to be tested contains nitrogen atoms.

4. The method for analyzing drug metabolites according to claim 3, characterized in that: For metabolites of non-radioactive isotope tracer drugs, MS with DP values ​​≥ 0.8 were screened and retained. 2 Spectra and corresponding MS 1 Atlas.

5. The method for analyzing drug metabolites according to claim 1, characterized in that: The drug is a radioactive isotope tracer drug.

6. The method for analyzing drug metabolites according to claim 1, characterized in that: It also includes the use of numpy and matplotlib libraries to draw and visualize the mirror image of the parent drug MS molecular network and the microscopic MS molecular network.

Images