A plasma metabolite marker combination for assisting in identifying etomidate abuse and application thereof

Abstract

The application discloses a plasma metabolic marker combination for assisting in identifying etomidate abuse and application thereof, and can make up for a short board of a detection window period: traditional toxic substance detection is negative 24-48 hours after medication, and the application can still provide auxiliary metabolic evidence related to an abuse state after exceeding a direct toxic substance detection window period by detecting changes in endogenous metabolic homeostasis. The application does not depend on subjective self-reports of an evaluated person, can provide objective laboratory data support for evaluation of the abuse state in a judicial psychiatric appraisal or a forced drug rehabilitation scene, and can effectively identify etomidate abusers.

Description

Technical Field

[0001] This invention relates to a metabolic biomarker and its application, and more particularly to a combination of plasma metabolic biomarkers for assisting in the identification of etomidate abuse and its application. Background Technology

[0002] Etomidate, a short-acting non-barbiturate intravenous general anesthetic, has been used by criminals in recent years as a drug substitute, mixed into e-cigarette liquids or tobacco for inhalation. The abuse population is showing a trend towards younger ages, becoming a prominent public health and forensic toxicology issue. From October 1, 2023, etomidate was officially listed as a Class II psychotropic drug, possessing clear narcotic properties.

[0003] Currently, forensic identification and detection research on etomidate abuse mainly focuses on the qualitative and quantitative analysis of the original drug or its known metabolites, with detection matrices covering various types such as blood, urine, hair, and e-cigarette liquids. Mainstream detection methods can be divided into two main categories: immunological methods and chromatography-mass spectrometry (GC-MS). These methods directly detect the original drug or its known single metabolite (such as etomidate) in blood or urine. These methods are the gold standard for confirming recent drug exposure. This current situation of "primarily exogenous detection and lack of endogenous assessment" results in a lack of objective metabolic biomarkers that can be used to assist in assessing abuse status in forensic identification practice.

[0004] Metabolomics, as an important branch of systems biology, has demonstrated unique value in the field of forensic toxicology in recent years. Significant progress has been made in the analysis of metabolic characteristics and biomarker screening of addictive substances such as opioids, methamphetamine, and cocaine, with research samples covering various types including blood, urine, and brain tissue. Meanwhile, machine learning algorithms, with their ability to process high-dimensional, multicollinear data, have been gradually applied to the construction of forensic diagnostic models, including scenarios such as time of death estimation, injury identification, and toxicant exposure classification. The combined application of metabolomics and machine learning provides a new technical approach to assist in identifying substance abuse states from the perspective of endogenous metabolic alterations.

[0005] The existing technology has the following main drawbacks: 1. Short detection window: Etomidate is rapidly metabolized in the body, and its direct detection window in blood is relatively short (about 24-48 hours). Traditional methods are prone to missed detection for suspected abusers who are found after this window period.

[0006] 2. Inability to distinguish between abuse states: Traditional tests can only answer "whether it has been used recently", and cannot effectively distinguish between single, accidental use and chronic, repeated abuse, which is crucial in forensic practice.

[0007] 3. Susceptible to subjective influence: In judicial or compulsory drug rehabilitation settings, the assessment of the drug use history of the person being evaluated relies heavily on self-reporting, which is often unreliable due to denial or concealment, and lacks objective laboratory-assisted diagnostic evidence. Summary of the Invention

[0008] Objective of this invention: The objective of this invention is to provide a set of plasma metabolic biomarkers that can be used to assist in identifying whether an individual has been abusing etomidate long-term, thereby overcoming the technical shortcomings of traditional toxicology detection methods, such as short detection windows and inability to distinguish abuse states. Another objective of this invention is to provide a machine learning diagnostic model based on this biomarker set, to achieve objective and accurate identification of etomidate abusers.

[0009] This invention uses non-targeted and targeted metabolomics analysis of plasma samples from etomidate abusers (confirmed by judicial, clinical and administrative records) and healthy controls, combined with machine learning algorithms such as random forest, to screen and validate a set of plasma biomarkers that can reflect endogenous metabolic changes related to etomidate abuse: hexanoylcarnitine, adenosine monophosphate (AMP), and lysophosphatidylethanolamine (LysoPE(P-18:1(9Z) / 0 / 0)).

[0010] Technical solution: Application of reagents for detecting the concentration of metabolic markers in the preparation of diagnostic kits for etomidate abuse, wherein the metabolic markers include one or more of hexanocarnitine, adenosine monophosphate, and lysophosphatidylethanolamine.

[0011] In the aforementioned application, the reagent is used to determine the concentration of metabolic markers by liquid chromatography or liquid chromatography-tandem mass spectrometry (LC-MS / MS).

[0012] The aforementioned application describes the use of the diagnostic kit to detect isolated plasma samples.

[0013] The aforementioned application, etomidate abuse, includes long-term etomidate abuse where the metabolic window period has passed since the intake at the time of detection.

[0014] A method for identifying etomidate abuse includes the following steps: (1) Sample detection: The concentrations of three metabolites in isolated plasma samples were quantitatively detected by liquid chromatography or liquid chromatography-tandem mass spectrometry; the three metabolites were hexanocarnitine, adenosine monophosphate and lysophosphatidylethanolamine. (2) Model input: The concentration values ​​of the three metabolites described in step (1) are used as input features; (3) Model determination: The pre-built random forest model is used for calculation. The model parameters are: number of decision trees (ntree) selected from 100-300, number of features randomly selected for each node (mtry) selected from 1-3, and minimum sample size of the node (nodesize) selected from 4-6. (4) Output results: The model outputs the classification probability and determines whether it is etomidate abuse based on the preset threshold.

[0015] In the method for identifying etomidate abuse, step (4) uses the threshold value that maximizes the Youden index as the cutoff point. Preferably, the preset threshold is 0.4-0.5, and more preferably, the preset threshold is 0.44.

[0016] The method for identifying etomidate abuse uses a pre-constructed random forest model from step (3) based on plasma concentrations of three metabolites.

[0017] The identification model for etomidate abuse is a model constructed using the random forest algorithm based on the concentrations of three metabolites—hexanocarnitine, adenosine monophosphate, and lysophosphatidylethanolamine—in isolated plasma samples.

[0018] The application of the aforementioned identification model in identifying etomidate abuse.

[0019] The aforementioned application, etomidate abuse, includes long-term etomidate abuse where the metabolic window period has passed since the intake at the time of detection.

[0020] Specific detection and model building scheme: 1. Sample detection: The concentrations of hexanocarnitine, adenosine monophosphate and LysoPE (P-18:1(9Z) / 0 / 0) in plasma samples were quantitatively detected by liquid chromatography-tandem mass spectrometry (LC-MS / MS) in multiple reaction monitoring (MRM) mode.

[0021] 2. Model input: Use the concentration values ​​of the above three metabolites as input features.

[0022] 3. Model determination: The pre-built random forest model is used for calculation. The model parameters are: number of decision trees (ntree) = 200, number of features randomly selected for each node (mtry) = 2, and minimum sample size of each node (nodesize) = 5.

[0023] 4. Output results: The model outputs classification probabilities and determines whether an individual is an etomidate abuser based on a preset threshold.

[0024] Key points of the invention 1. Novel biomarker combination: The combination of three endogenous metabolites, hexanocarnitine, adenosine monophosphate and LysoPE (P-18:1(9Z) / 0 / 0), was discovered and confirmed for the first time to be significantly associated with etomidate abuse status.

[0025] 2. Reflects abuse status, rather than short-term exposure: The biomarkers of this invention capture systemic metabolic changes such as energy metabolism reprogramming (fatty acid oxidation, purine metabolism) and membrane phospholipid metabolism remodeling caused by repeated abuse, rather than the drug itself. Therefore, they can help distinguish between chronic abuse and single use.

[0026] 3. Machine learning model ensemble: By combining the three markers with the random forest algorithm, a high-accuracy classification model was constructed, achieving an AUC of 0.83 and an accuracy of 81.1% on the independent validation set.

[0027] 4. Objective auxiliary diagnostic value: It may provide a new auxiliary assessment tool for scenarios in judicial appraisal practice where objective evidence is lacking due to exceeding the window period for toxicology testing or unreliable self-reports.

[0028] Beneficial effects: Compared with the prior art, the present invention has the following advantages: (1) It makes up for the shortness of the detection window period: Traditional toxicology tests are negative 24-48 hours after drug administration, while the present invention, by detecting changes in endogenous metabolic homeostasis, may still provide auxiliary metabolic evidence related to abuse status after the direct toxicology detection window period has expired. (2) It provides objective metabolic evidence: It does not rely on the subjective self-report of the person being assessed, and provides objective laboratory data support for the assessment of abuse status in forensic psychiatric evaluation or compulsory drug rehabilitation scenarios. (3) It has good discrimination ability: Experimental data show that the combined biomarker model of the present invention has an AUC value of 0.831 (95% CI: 0.711-0.951) in the independent validation set, a prediction accuracy of 81.1%, a sensitivity of 90.0%, and a specificity of 69.6%, indicating that the model has good discrimination ability and above-average specificity, and can effectively identify etomidate abusers. Attached Figure Description

[0029] Figure 1 This is the research and design roadmap for the present invention;

[0030] Figure 2 This is a PCA analysis of etomidate abuse group and healthy control group under positive and negative ion modes in plasma non-targeted metabolomics analysis;

[0031] Figure 3 This is a score graph of OPLS-DA analysis of the etomidate abuse group in positive ion mode;

[0032] Figure 4This is a diagram of model parameters for OPLS-DA analysis of the etomidate abuse group in positive ion mode;

[0033] Figure 5 This is a displacement test plot of OPLS-DA analysis of the etomidate abuse group in positive ion mode;

[0034] Figure 6 This is a score graph of OPLS-DA analysis of the healthy control group under negative ion mode;

[0035] Figure 7 This is a graph of model parameters from the OPLS-DA analysis of the healthy control group under negative ion mode;

[0036] Figure 8 This is a displacement test plot of OPLS-DA analysis of the healthy control group under negative ion mode;

[0037] Figure 9 This is a volcano diagram of the etomidate abuse group (POS) in positive ion mode;

[0038] Figure 10 This is a volcano diagram of the healthy control group (NEG) under negative ion mode;

[0039] Figure 11 This is an analysis of the differential metabolite metabolism pathways between the etomidate abuse group and the healthy control group under positive and negative ion modes;

[0040] Figure 12 These are the ROC plot and concentration box plot of the candidate metabolite Hexanoylcarnitine after targeted validation;

[0041] Figure 13 The ROC plot and concentration box plot of candidate metabolite LysoPE (P-18:1(9Z) / 0 / 0) after targeted validation are shown.

[0042] Figure 14 These are the ROC plot and concentration box plot of the candidate metabolite Adenosine monophosphate after targeted validation;

[0043] Figure 15 These are the ROC plot and concentration box plot of the candidate metabolite Taurine after targeted validation;

[0044] Figure 16 These are the ROC plot and concentration box plot of the candidate metabolite Progesterone after targeted validation;

[0045] Figure 17The combined biomarker model performance graph includes: (A) ROC curves for the test set and independent validation set; (B) Confusion matrix for the test set and independent validation set; (C) Metabolite importance graph for the final model. Detailed Implementation

[0046] Example 1: Sample Collection and Metabolome Extraction The study included plasma samples from 82 healthy individuals who had not been exposed to drugs and 146 plasma samples from patients who abused etomidate. This study employed a three-stage data partitioning strategy (e.g., ...). Figure 1 (As shown). The discovery set (used for candidate biomarker screening) included plasma samples from 52 etomidate abusers and 29 healthy controls. The modeling set (used for model training, feature selection, and threshold determination) included plasma samples from 64 etomidate abusers and 30 healthy controls. The independent validation set (used for model performance evaluation) included plasma samples from 30 etomidate abusers and 23 healthy controls. All participants were male, and there was no statistically significant age difference between the two groups.

[0047] After collection, plasma samples were centrifuged at 4℃ and 3000 r / min for 10 min. The clear supernatant plasma was precisely aspirated and aliquoted, then flash-frozen in liquid nitrogen within 30 seconds to terminate the enzymatic metabolic reaction. Subsequently, the samples were transferred to an ultra-low temperature freezer at -80℃ for storage away from light. Long-term storage at -20℃ and repeated freeze-thaw cycles are strictly prohibited. Dry ice was used throughout the inter-regional transport process to maximize the stability of metabolites.

[0048] Metabolite extraction: Take 50 μL of plasma, add 50 μL of internal standard (L-2-chlorophenylalanine and ketoprofen, 3.2 μg / mL each), add 150 μL of extraction buffer (methanol:acetonitrile = 3:1), vortex for 5 minutes, let stand at room temperature for 10 minutes, centrifuge at 13,000 rpm for 10 minutes at 4°C, and take 150 μL of supernatant to dry under vacuum.

[0049] Example 2: Non-targeted metabolomics for screening candidate biomarkers The non-targeted metabolomics detection platform was a ThermoScientific Q Exactive HF-X ultra-high resolution mass spectrometer with a Vanquish ultra-high performance liquid chromatography system in series; the chromatographic column used was a Waters ACQUITY UPLC HSST3 column (2.1 mm × 100 mm, 1.8 μm); the mobile phase A was an aqueous solution containing 0.1% formic acid-10 mM ammonium formate, and the mobile phase B was acetonitrile; the sample was placed in an autosampler at 4 ℃, and the gradient elution program was as follows: 0–0.5 min, phase B maintained 2%; 0.5–2 min, phase B linearly increased from 2% to 50%; 2–5 min, phase B linearly increased from 50% to 98%; 5–8 min, phase B maintained 98% isocratic; 8–10 min, phase B linearly decreased from 98% to 2%; 10–12 min, phase B stabilized at 2%. The mass spectrometry parameters were set as follows: sheath gas flow rate 30 L / min, auxiliary gas flow rate 25 L / min; spray voltage for both positive and negative ion modes 3.6 kV; SLEns-RF value 40; capillary temperature 325 ℃, auxiliary gas heating temperature 300 ℃; secondary fragmentation collision energy (NCE) set to 20, 30, and 40; data-dependent acquisition mode (DDA) was used, prioritizing the fragmentation of the top 10 ions (Top-N=10) per cycle, with a mass-to-charge ratio scan range of 70–1050 m / z, and a total acquisition time of 12 min. One QC sample was interspersed with every 10 consecutive sample analyses to monitor the stability of the chromatographic mass spectrometry system and the reliability of the data throughout the experiment; metabolites with a peak area relative standard deviation (RSD) > 30% in the QC samples were removed. The QC samples showed tight PCA clustering, confirming the reliability and stability of the detection platform.

[0050] We performed untargeted metabolomics analysis on plasma samples in the discovery cohort, annotating a total of 9440 metabolic features in both positive and negative ion modes. Of these, 5685 were in the positive ion mode and 3755 in the negative ion mode. Our PCA analysis of the sample data revealed that all QC samples clustered together. Figure 2 This indicates that the QC sample data showed good repeatability during the analysis, the chromatographic system was stable, and the mass spectrometry signal showed no significant attenuation. A certain degree of differentiation was observed between the etomidate abuse group and the healthy group, indicating that some metabolites contributed to the differentiation between the samples. OPLS-DA analysis showed that samples from both groups were clearly separated. Figures 3-8 This indicates that some metabolites differed significantly between the two groups of samples within the constructed model. (Volcano plot) Figure 9 , 10The results also showed significant differences in metabolic characteristics between the healthy group and patients abusing etomidate. We screened differential metabolites using t-test (p<0.05 & FDR<0.05), VIP>1, and FC>2 or <0.5, and then further screened the differential metabolites to obtain endogenous differential metabolites. A total of 68 endogenous differential metabolites were found between the etomidate abuse group and the healthy group, including 32 positive ion patterns and 36 negative ion patterns. These mainly included lipids and lipid molecules, organic acids and their derivatives, nucleoside nucleotides, and other substances.

[0051] Pathway results of differential plasma metabolite enrichment between healthy individuals and etomidate abusers, as follows: Figure 11 As shown in the diagram, differential enrichment exists in the pathway enrichment map for purine metabolism, taurine and taurine metabolism, linoleic acid metabolism, and vitamin B6 metabolism. This suggests that the abuse of etomidate can cause energy metabolism reprogramming, leading to purine and glycerol ester metabolism disorders, causing nerve damage and compensatory protection manifested as abnormal taurine and vitamin B6 metabolism, while also causing inflammation and lipid homeostasis imbalance (alterations in linoleic acid, α-linolenic acid, and bile acid metabolism).

[0052] To assess the ability of the aforementioned differential metabolites to distinguish etomidate abusers, we performed ROC analysis on the peak areas of all differential metabolites. The area under the curve (AUC) of each differential metabolite was above 0.5, indicating a certain ability to distinguish etomidate abusers. Based on ROC performance and biological relevance, eight candidate metabolites were selected from 68 differential metabolites for targeted validation. Seven of these candidates had AUC > 0.9; taurine, due to its involvement in the taurine and hypotaurine metabolic pathways and its potential biological significance (AUC 0.875), was also included in subsequent validation. Information on the candidate biomarkers is shown in Table 2.

[0053]

[0054] Example 3: Screening and Target Validation of Candidate Metabolic Biomarkers Targeted LC-MS / MS detection: Absolute quantification of eight candidate metabolites was performed using a SCIEX Triple Quad 6500+ in MRM mode. PS (14:0 / 14:0), LysoPE (P-18:1(9Z) / 0:0), hexanocarnitine, adenosine monophosphate, and taurine were detected using a Waters ACQUITY UPLC Beh Amide column (2.1 mm × 100 mm, 1.8 μm). The mobile phase A was an aqueous solution of 0.1% formic acid and 10 mM ammonium acetate, and the mobile phase B was acetonitrile. Throughout the analysis, the samples were placed in an autosampler at 4°C. The samples were subjected to ultra-high performance liquid chromatography with the following gradient elution program: 0–3 min, 95% B; 3–10 min, 5% B; 10–12 min, 5% B; 12–12.1 min, 95% B; 12.1–18 min, 95% B. Prostaglandin B1, α-hydroxyhippuric acid, and progesterone were analyzed using a Waters ACQUITY UPLC Beh T3 column (2.1 mm × 100 mm, 1.8 μm). The mobile phase A was 0.1% formic acid aqueous solution, and phase B was acetonitrile. The gradient elution program was as follows: 0–3 min, 1% B; 3–10 min, 5% B; 10–12 min, 5% B; 12–12.1 min, 95% B; 12.1–18 min, 95% B. Method validation showed good linearity for each analyte (r > 0.99), intra- and inter-batch precision RSD < 15%, and accuracy between 85% and 115%. In targeted validation, PS(14:0 / 14:0) / prostaglandin B1 and α-hydroxyhippuric acid were below the limit of quantitation in most real-world samples, and these three were excluded. Ultimately, five metabolites (hexanocarnitine, adenosine monophosphate, LysoPE (P-18:1(9Z) / 0 / 0), progesterone, and taurine) were used for model construction. The differences in each biomarker between the abuse group and the healthy group were statistically significant. The AUC values ​​and concentration box plots for individual biomarkers are shown below. Figures 12-16 As shown.

[0055] Example 4: Construction and Validation of Machine Learning Diagnostic Models Model Construction: The modeling set data (n=94) for targeted validation was divided into training and test sets in a 7:3 ratio. Five algorithms were compared: logistic regression, random forest, SVM, LASSO, and Ridge. AUC was used as the main evaluation metric to compare model performance. The random forest algorithm showed good robustness to the high dimensionality and collinearity of metabolomics data, therefore, it was selected as the final modeling algorithm in subsequent analyses. Furthermore, the random forest algorithm achieved the highest AUC value on the internal validation set, demonstrating the best overall model performance (Table 3).

[0056]

[0057] Feature selection: Based on the feature importance scoring (Mean Decrease Accuracy and Mean Decrease Gini) of the random forest algorithm and the recursive feature elimination (RFE) strategy, the three metabolites with the largest contribution were finally determined as the final model features: hexanocarnitine, adenosine monophosphate, and LysoPE (P-18:1(9Z) / 0 / 0).

[0058] Model parameters: The random forest model parameters are as follows: ntree = 200, mtry = 2, nodesize = 5.

[0059] The model performance is shown in Table 4:

[0060] The combined biomarker model demonstrated good discriminative ability on the test set. ROC analysis showed that the model's AUC value for distinguishing between the etomidate abuse group and the healthy control group was 0.918 (95% CI: 0.801-1). Figure 17 A). Using the threshold of maximizing the Youden exponent (0.44) as the cutoff point, the model achieved an accuracy of 89.3%, a sensitivity of 94.7%, and a specificity of 77.8% on the test set (Table 4). To evaluate the model's generalization ability, the trained model was applied to the independent validation set. The results showed that the model's AUC on the independent validation set was 0.831 (95% CI: 0.711–0.951). Figure 17 (B) The prediction accuracy reached 81.1%, with sensitivity and specificity of 90% and 69.6%, respectively (Table 4). The final model's metabolite importance graph is shown below. Figure 17 As shown in C. Although the AUC on the independent validation set is slightly lower than that on the training set, it is still at a good level (AUC>0.8), indicating that the model has good generalization ability and no obvious overfitting phenomenon.

Claims

1. The application of a reagent for detecting the concentration of a metabolic marker in the preparation of a diagnostic kit for etomidate abuse, wherein the metabolic marker comprises one or more of hexanocarnitine, adenosine monophosphate, and lysophosphatidylethanolamine.

2. The application according to claim 1, characterized in that, The reagents are used to determine the concentration of metabolic markers by liquid chromatography or liquid chromatography-tandem mass spectrometry.

3. The application according to claim 1, characterized in that, The diagnostic kit is used to detect isolated plasma samples.

4. The application according to claim 1, characterized in that, Etomidate abuse includes long-term abuse of etomidate, but the metabolic window period has passed since the intake when it is detected.

5. A method for identifying etomidate abuse, characterized in that, Includes the following steps: (1) Sample detection: The concentrations of three metabolites in isolated plasma samples were quantitatively detected by liquid chromatography or liquid chromatography-tandem mass spectrometry; the three metabolites were hexanocarnitine, adenosine monophosphate and lysophosphatidylethanolamine. (2) Model input: The concentration values ​​of the three metabolites described in step (1) are used as input features; (3) Model determination: The pre-built random forest model is used for calculation. The model parameters are: the number of decision trees is selected from 100-300, the number of features randomly selected for each node is selected from 1-3, and the minimum sample size of the node is selected from 4-6. (4) Output results: The model outputs the classification probability and determines whether it is etomidate abuse based on the preset threshold.

6. The method for identifying etomidate abuse according to claim 5, characterized in that, The preset threshold in step (4) is cut off at the threshold where the Youden exponent is maximized.

7. The method for identifying etomidate abuse according to claim 5, characterized in that, The pre-built random forest model described in step (3) is constructed based on the plasma concentrations of three metabolites.

8. A model for identifying etomidate abuse, characterized in that, It is a model constructed using the random forest algorithm to measure the concentrations of three metabolites—hexanocarnitine, adenosine monophosphate, and lysophosphatidylethanolamine—in isolated plasma samples.

9. The application of the identification model according to claim 8 in identifying etomidate abuse.

10. The application according to claim 9, characterized in that, Etomidate abuse includes long-term abuse of etomidate, but the metabolic window period has passed since the intake when it is detected.

Images