Drug sensitivity prediction method based on blood NMR editing spectrum

By using blood NMR editing spectrum detection technology combined with machine learning models, the uncertainty and sampling damage problems of existing drug sensitivity prediction methods are solved, and more accurate individualized drug sensitivity prediction is achieved.

CN120613144AActive Publication Date: 2025-09-09INNOVATION ACAD FOR PRECISION MEASUREMENT SCI & TECH CAS
View PDF 6 Cites 0 Cited by

Patent Information

Application Number
CN202510689603.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-05-27
Publication Date
2025-09-09
Estimated Expiration
2045-05-27

AI Technical Summary

Technical Problem

Existing drug sensitivity prediction methods have uncertainty and low accuracy in the treatment of chronic non-communicable diseases such as cancer. In particular, prediction methods based on upstream biological indicators cannot meet the needs of personalized precision treatment, and the sampling process causes great damage to the body.

Method used

A method based on blood NMR editing spectrum is used. Through one-dimensional proton NMR editing spectrum detection technology, molecular interactions are retained to obtain NMR fingerprint data of small molecule metabolites, glycoproteins, and lipoproteins in plasma or serum before and after drug treatment. Combined with random forest and support vector machine classification models, individual differences are identified and drug sensitivity predictions are made.

Benefits of technology

It improves the accuracy of drug sensitivity prediction, reduces the damage to the body caused by sampling, realizes non-destructive testing and standardized analysis, and can more accurately predict an individual's response to drugs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120613144A_ABST
    Figure CN120613144A_ABST
Patent Text Reader

Abstract

The invention belongs to the technical field of medical examination, and relates to a drug sensitivity prediction method based on a blood NMR (nuclear magnetic resonance) editing spectrum, for example, a cancer immunotherapy drug-immune checkpoint inhibitor (ICIs) with ubiquitous drug resistance. The method comprises the following steps: firstly, acquiring pathological indexes before and after drug treatment and NMR (nuclear magnetic resonance) editing spectrum fingerprint data of plasma or serum, then constructing a drug sensitivity prediction model and screening predictive biomarkers by adopting a machine learning algorithm, and finally performing drug sensitivity prediction on blood NMR data of a patient needing to be treated by using the biomarkers. The NMR editing spectrum adopted by the invention is non-destructive to a sample, interaction information of a drug and an organism is retained while large and small molecules in blood are comprehensively observed, uncertainty caused by prediction based on upstream data sets such as genomics and the like can be avoided by blood metabolism phenotype analysis, prediction accuracy is improved, and the method has a good application prospect. The method is expected to be applied to clinical medication guidance of drugs with drug resistance.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of medical testing and relates to a method for predicting drug sensitivity based on blood NMR editing spectrum. Background Art

[0002] Chronic non-communicable diseases, such as cancer and cardiovascular and cerebrovascular diseases, pose a serious threat to human health. Due to their complex pathogenesis and diverse influencing factors, many drugs used to treat these diseases suffer from varying degrees of resistance or intolerance, making treatment challenging. Developing accurate and effective methods to predict drug sensitivity to specific drugs, and thus predicting a patient's sensitivity to a particular drug in advance, will help promote personalized, precision medicine treatment for these diseases.

[0003] Currently, most drug sensitivity prediction methods are developed to serve drug screening and are usually based on high-throughput drug sensitivity data and cell line genomics databases of related diseases, such as the Cancer Cell Line Encyclopedia (CCLE) and Genomics of Drug Sensitivity to Cancer (GDSC). At the cellular level, machine learning methods are used to non-targetedly predict drugs that may respond to the disease.

[0004] Predicting sensitivity to a specific drug at the body fluid or tissue level is often done by selecting upstream biological indicators based on the drug's mechanism of action. Taking cancer as an example, despite the recent development of immunotherapy approaches based on ICIs, such as PD-1 / PD-L1 inhibitors, which have significantly improved overall survival and reduced side effects in patients with various types of cancer, significant intergroup differences in the efficacy of ICIs exist, with only a small proportion of patients benefiting. Therefore, it is necessary to detect predictive biomarkers of efficacy before treatment and guide medication use. Biomarker testing is currently the primary approach for assessing patients who may benefit from ICIs and guiding their rational use. In recent years, researchers have discovered numerous biomarkers for this therapy at the genetic and immune microenvironmental levels, such as PD-L1 expression, tumor mutational burden (TMB) levels associated with cancer, cancer microenvironment-related factors, and intestinal microbial flora. Medication guidelines based on PD-L1 expression have been applied in clinical treatment. However, even with biomarker-based predictions of drug efficacy, the clinical objective response rate to ICIs alone remains generally below 30% for most cancer types. Some cancer patients even experience accelerated disease progression, developing hyper-progressive disease (HPD), even after receiving targeted medication. This demonstrates that predictive methods relying on upstream biomarkers still cannot meet the needs of precision medicine.

[0005] Although upstream biological indicators such as gene expression and gene mutation can indicate possible future events, they are inevitably subject to numerous escape mechanisms. With the widespread application of metabolic phenotyping in the medical field, it has been discovered that most chronic non-communicable diseases are closely related to metabolic disorders. For example, cardiovascular and cerebrovascular diseases are often accompanied by abnormal lipid metabolism, and cancer is closely related to energy metabolism disorders. Unlike upstream biological indicators, metabolic phenotyping information is certain to have already occurred. Therefore, developing drug sensitivity prediction methods for treating chronic diseases based on metabolic phenotyping can avoid the uncertainty brought about by upstream biological indicator predictions and improve the accuracy of predictions.

[0006] In recent years, metabolic phenotyping has been widely used in medical-related fields such as drug metabolism research, disease diagnosis and typing, and biomarker discovery related to chronic non-communicable diseases. In addition to changes in common small molecule metabolites, the metabolism of lipids and some biological macromolecules (such as lipoproteins and glycoproteins) is also closely related to chronic non-communicable diseases. Commonly used techniques for metabolic phenotyping include mass spectrometry and nuclear magnetic resonance (NMR) technology. Among them, nuclear magnetic resonance is very effective in detecting disease characteristics in biological fluids, and it has been proven to be an effective diagnostic option for COVID-19 infection. For biological fluid samples, nuclear magnetic resonance can achieve non-destructive testing, which means that it can achieve simultaneous detection of small molecule metabolites and some biological macromolecules without separating the samples. This also means that the interactions between molecules (including drugs and small molecules, and biological macromolecules) can be preserved, so that additional evaluation information can be obtained. Metabolic phenotyping analysis based on nuclear magnetic resonance technology reveals the relationship between metabolite diversity in human plasma and plasma biomacromolecule-drug interaction (Du Y, Lan W, Ji Z, Zhang X, Jiang B, Zhou X, et al. NMR spectroscopic approach reveals metabolic diversity of human blood plasma associated with protein-drug interaction. Analytical chemistry 85 (2013) 8601-8.). Since the diversity of metabolites in the blood is caused by individual metabolic differences, blood metabolic phenotyping analysis based on nuclear magnetic resonance can be used to evaluate the relationship between individual metabolic differences and drug efficacy. By using machine learning methods to model the metabolic data of blood before and after drug treatment, the shortcomings in predictive accuracy of using only a single biological indicator can be compensated. Summary of the Invention

[0007] Based on the above background, this application intends to separate from the upstream biological mechanism, focus on the differences in blood metabolism of drugs in patients with chronic non-communicable diseases, and propose a prediction method that can avoid the uncertainty brought by unknown escape mechanisms and improve prediction accuracy. The sampling process causes little damage to the body, and the detection and analysis process is non-destructive and standardized.

[0008] In order to solve the problems existing in the prior art, the present invention first establishes two one-dimensional proton NMR editing spectrum detection methods (T2W, DIRE) for plasma or serum samples, retaining the intermolecular interactions while obtaining as much metabolite information as possible, and obtaining the NMR fingerprint data of small molecule metabolites and glycoproteins and lipoproteins in the plasma or serum before and after drug treatment; then, through physiological and pathological indicators, identify individual differences in drug treatment effects and divide the cohort into high-sensitivity and low-sensitivity groups, and train unsupervised classifiers based on group labels and fingerprint data of blood samples before treatment, including random forest and support vector machine classification models, evaluate and screen the optimal classification model, and screen out biomarkers with high predictive ability for drug sensitivity prediction through ROC analysis of features. The technical route of this method is shown in the attached manual. Figure 1 shown.

[0009] Specifically, the present invention achieves the above-mentioned object by the following method:

[0010] A method for predicting drug sensitivity based on blood NMR editing spectrum, the specific steps are as follows:

[0011] 1. Queue Design

[0012] Strictly follow the ethical requirements of the trial to recruit a volunteer cohort of patients with a certain chronic non-communicable disease. The research cohort needs to be divided into two groups:

[0013] (1-1) Treatment group: patients who received a certain drug treatment;

[0014] (1-2) Blank control group: patients who did not receive treatment. Hierarchical cluster analysis was performed on the blood biochemical and prognostic indicators of the treatment group after treatment and the blood biochemical and prognostic indicators of the blank control group at the later stage of progression. The patients in the treatment group were divided into a high-sensitivity group and a low-sensitivity group;

[0015] (1-3) Blood sample retention

[0016] (1-3-1) Keep the plasma or serum samples from all patients at baseline as pre-dose samples;

[0017] (1-3-2) Retaining plasma or serum samples from patients in the treatment group at the late stage of treatment (the late stage of treatment in the present invention refers to the relatively stable stage of the disease) and patients in the blank control group at the late stage of natural progression (the late stage of progression in the present invention is determined with reference to the late stage of treatment in the treatment group);

[0018] 2. Collection and Processing of NMR Fingerprint Data

[0019] (2-1) Obtain one-dimensional plasma or serum samples from each group in step 1 1 H NMR spectral data, the one-dimensional 1 H NMR spectra were relaxation-weighted T2W-NMR and diffusion-edited relaxation DIRE;

[0020] (2-2) obtaining a two-dimensional nuclear magnetic resonance spectrum of any one of the plasma or serum samples in step 1 to assist in the identification of T2W-NMR spectral signals;

[0021] (2-3) Data processing:

[0022] For the one-dimensional 1 The H NMR spectrum was phase and baseline corrected, and the chemical shift of the T2W-NMR spectrum was calibrated using the proton signal of the glucose α-isomer at δ5.23 ppm. The DIRE spectrum was calibrated using the peak with the best resolution (preferably the -N(CH3)3 proton signal at δ3.20 ppm for SPC). The line width factor was set to 0.3 Hz, and the frequency domain spectrum was then generated using Fourier transformation.

[0023] For the above 1 The H NMR spectrum was corrected for chemical shift drift and then segmented integrated with an integration width ranging from 6 to 24 Hz. The segmented integration was finally normalized based on the sum of the total integrations of each sample to obtain the data sets of the T2W-NMR spectrum and the DIRE spectrum, respectively.

[0024] (2-4) T2W-NMR spectral signal identification

[0025] Signal identification of the T2W-NMR spectrum of blood samples was performed based on the two-dimensional nuclear magnetic resonance spectrum (2-2) and the human metabolome database HMDB;

[0026] (2-5) DIRE spectrum signal identification

[0027] Signal identification of blood sample DIRE spectra was performed with reference to the following literature: Liu M, Nicholson JK, London JC. High-resolution diffusion and relaxation edited one- and two-dimensional 1HNMR spectroscopy of biological fluids.Analytical chemistry 68(1996)3370-6;

[0028] The two fingerprint spectrum signals are identified through the above (2-4) and (2-5), wherein the T2W-NMR spectrum is the fingerprint spectrum of small molecule metabolites in the blood sample, and the DIRE spectrum is the fingerprint spectrum of glycoproteins, lipoproteins and lipid macromolecular metabolites containing small molecule branches in the blood sample;

[0029] 3. Model Construction and Evaluation

[0030] Random forest and support vector machine algorithms were used to model T2W-NMR and DIRE fingerprint datasets from pre-treatment high-sensitivity and low-sensitivity plasma or serum samples. Four prediction models were constructed. The random forest model consisted of 500 decision trees, and the generalization ability of the model was evaluated using out-of-bag error (OOB). The generalization ability of the support vector machine was evaluated using leave-one-out cross-validation (LOOCV). Based on the evaluation results, the model with the lowest classification error was selected for drug susceptibility prediction.

[0031] Evaluating the performance of random forests by out-of-bag error has also been shown to be effective (Mitchell MW. Bias of the Random Forest Out-of-Bag (OOB) Error for Certain Input Parameters [J]. Open Journal of Statistics, 2011, 01 (3): 205-211.).

[0032] For T2W-NMR spectra, the characteristic data set used for modeling is a matrix consisting of the chemical shifts and their segmented integral values ​​of the characteristic peaks of the small molecule metabolites identified in step (2-4); the characteristic data set of DIRE spectra is a matrix consisting of the chemical shifts and their segmented integral values ​​of the glycoproteins, lipoproteins and lipid macromolecular metabolites containing small molecule branches identified in step (2-5);

[0033] 4. Biomarker Screening

[0034] The BiomarkerAnalysis module of the MetaboAnalyst analysis platform was used to screen important features based on the model selected in (4-1) and perform ROC curve analysis on all features. The ROC curve in this module was generated by Monte Carlo cross-validation (MCCV) with balanced subsampling. Through this process, important features with biomarker functions were identified. The classification prediction ability of the feature variables was then evaluated based on the area under the ROC curve (AUC) of different feature variable combinations, and one or more biomarkers were selected for drug sensitivity prediction.

[0035] 5. Drug sensitivity prediction

[0036] ① Prepare the data to be predicted: According to step 2, collect and process the blood NMR data of the patients to be treated, and convert the NMR data into the input format expected by the model;

[0037] ② The obtained NMR data were input into the model through the Biomarker Analysis module of the MetaboAnalyst analysis platform, and the biomarkers or their combinations screened in step 4 were used to predict the classification of the high-sensitivity group and the low-sensitivity group.

[0038] Furthermore, in step (2), a transverse relaxation time-weighted T2W-NMR spectrum is obtained using a water-presaturated 1D carr-purcell-meiboomm-gill pulse sequence with a total spin echo time of 70 ms and a cycle delay of 2.0 s; and / or

[0039] When acquiring the diffusion and relaxation editing spectra DIRE, three smooth square gradient fields were used to achieve diffusion editing, with intensities of -17.13%, 13.13%, and 80%, respectively. The gradient length and diffusion delay were 1.53 ms and 120 ms, respectively. Relaxation editing was achieved using a total spin echo time of 33 ms.

[0040] Furthermore, the method selects drugs and disease types with large individual differences in efficacy in clinical applications, the drugs including but not limited to PD-1 monoclonal antibodies: Pembrolizumab, Nivolumab, Camreliz; or PD-L1 monoclonal antibodies: Atezolizumab, Avelumab, Adebelimumab; the diseases including but not limited to: non-small cell lung cancer, gastrointestinal cancer, breast cancer.

[0041] The prediction model and method proposed in this paper are essentially designed to solve the binary classification problem of the data being predicted (high-sensitivity group and low-sensitivity group). Machine learning is a powerful tool for solving this classification problem. To achieve this classification, a classifier is trained using a machine learning algorithm. This classifier then predicts drug sensitivity based on the NMR data characteristics of the patient's blood sample.

[0042] This paper uses two classifiers: the supervised Random Forest (RF) and Support Vector Machine (SVM) algorithms. The Random Forest algorithm is an ensemble learning-based algorithm that uses bagging and random feature selection to build multiple decision trees and combine their prediction results to improve model performance. The Support Vector Machine (SVM) is widely used in classification and regression tasks. Its core goal is to find a hyperplane that maximizes the gap between two classes, thereby achieving good classification results.

[0043] Compared with the prior art, this application has the following advantages and beneficial effects:

[0044] 1) The present invention establishes two one-dimensional hydrogen spectrum detection methods: relaxation-weighted spectrum T2W-NMR and diffusion relaxation editing spectrum DIRE. Using T2W-NMR spectrum and setting the appropriate echo time, large molecules with relatively small T2 can be quickly attenuated, while small molecules are retained due to their larger T2 and slower signal decay. DIRE spectrum can enhance the signals of glycoproteins and lipoproteins that diffuse slowly but have small molecule chains with longer relaxation times.

[0045] 2) Non-targeted prediction and non-destructive testing of blood samples preserve the interaction between drugs and biomolecules, improving prediction accuracy;

[0046] 3) The method proposed in the present invention is based on the detection of micro-blood samples. Detection and analysis can be carried out when the sample volume is greater than 30 μL. Therefore, the sampling process causes very little damage to the body, the detection process is non-destructive to the sample, and the sample can be used for other tests after the detection is completed. In addition, the analysis process is fast and can be standardized. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 .Flowchart of the method for predicting drug sensitivity of ICIs inhibitors proposed in this application.

[0048] Figure 2.Animal experimental timeline and physiological and pathological results in the examples. A is the experimental process of RMP1-14 treatment. B and C are the changes in body weight and tumor volume over time, respectively. D is the final tumor photograph of 8 random mice in the treatment group and 6 mice in the control group. E is the comparison of the final tumor weight of the treatment group and the control group. F and G are the final tumor section histological examinations of two representative mice after treatment with PBS and RMP-14, respectively. AVT: average tumor volume. ATV: average tumor volume. Control : Average tumor volume of the control group.

[0049] Figure 3 . The pulse sequence of the DIRE spectrum in the embodiment, wherein T e is the eddy current delay time of 5ms, t s is the T2 relaxation time (2t s Diffusion editing was achieved by using a gradient length δ of 1.5 ms and a diffusion delay time Δ of 120 ms.

[0050] Figure 4 . Signal identification of T2W-NMR (A) and DIRE (B) spectra of mouse plasma before administration in the treatment group in the example. Figure 4 The numbers in A represent: 1. cholesterol; 2. lipid; 3. isoleucine; 4. leucine; 5. valine; 6. 3-hydroxybutyric acid; 7. lactic acid; 8. alanine; 9. acetic acid; 10. N-acetyl glycoprotein (NAG); 11. O-acetyl glycoprotein (OAG); 12. methionine; 13. acetone; 14. pyruvate; 15. succinate; 16. glutamine; 17. citric acid; 18. lysine; 19. creatine; 20. phosphocholine; 21. glycerophosphocholine; 22. inositol; 23. glycerol; 24. threonine; 25. β-glucose; 26. α-glucose; 27. urea; 28. tyrosine; 29. ​​histidine; 30. phenylalanine; 31. tryptophan.

[0051] Figure 5 Hierarchical clustering heat map based on post-drug administration plasma biochemical indicators (see figure for details) and tumor growth rate in the examples. In the figure, "C" indicates the control group; in the legend, "T" indicates the treatment group.

[0052] Figure 6 .Out-of-bag error rate (OOB) of the random forest model trained on the T2W (A) and DIRE (B) spectrum datasets of pre-dose plasma in the Example.

[0053] Figure 7 . Error rate of leave-one-out cross validation (LOOCV) of the support vector machine model trained based on the T2W (A) and DIRE (B) spectrum datasets of pre-dose plasma in the examples.

[0054] Figure 8 . In the example, the random forest classification feature screening based on the DIRE spectrum of the pre-dose plasma sample is as follows: A-1: ​​the top 15 features with the highest selection frequency; A-2: ROC analysis of features 3-53 of the DIRE dataset; A-3: predictive ability of the top 5 high-frequency selected features. DETAILED DESCRIPTION

[0055] The technical solution of the present invention is further described below with reference to the accompanying drawings and embodiments.

[0056] Example

[0057] The prediction method will be used in a preclinical setting using a tumor-bearing mouse model and the mouse PD-1 inhibitor RMP1-14 as a therapeutic drug.

[0058] In the present invention, for the storage of plasma samples, it is necessary to select anticoagulant tubes that have passed NMR detection and store them at -80°C after quick freezing in liquid nitrogen within 30 minutes.

[0059] Preparation of plasma samples:

[0060] For blood samples with a plasma volume less than 100 μL, mix 30 μL of plasma with an equal volume of phosphate-buffered saline (45 mM, pH = 7.4 ± 0.1). Centrifuge at 11,000 g, 4°C for 10 minutes, and transfer the supernatant (50 μL) to a 1.7 mm NMR tube for analysis.

[0061] For blood samples with a plasma volume greater than 200 μL, 200 μL of plasma was mixed with 400 μL of phosphate-buffered saline (45 mM, pH = 7.4 ± 0.1), and after centrifugation at 11,000 g and 4°C for 10 min, the supernatant (500 μL) was transferred to a 5 mm NMR tube for testing.

[0062] 1. Animal Experiments and Blood Sample Collection

[0063] 1.1 Materials and Methods

[0064] Colorectal cancer MC38 cells (purchased from The Global Bioresource Center, cell line number ATCC#CRL-2638) were cultured according to established standard procedures. Female C57BL / 6J mice, 6-8 weeks old, were obtained from Cyagen Biosciences Inc., Suzhou, China, and housed under a 12-h light / dark cycle with free access to food and water. 1 x 10 6 After MC38 cells were grown, the tumor size was measured every 2 to 3 days using a digital caliper. The tumor volume was calculated as V = (W 2*L) / 2, where W is the width of the tumor and L is the length of the tumor. The formula for calculating tumor growth is: tumor growth (%) = tumor volume (day n) / tumor volume (day 0) × 100%.

[0065] The average volume reaches 90±20mm 3 On day 0, 22 mice were randomly divided into a treatment group (n=16) and a control group (n=6). The treatment group received 4.0 mg / kg (mpk) of the anti-mouse PD-1 antibody RMP1-14 (BioXcell, BE0146), while the control group received a phosphate-buffered saline (45 mM, pH=7.4±0.1) vehicle injection. Both groups received intraperitoneal injections on days 0, 3, 7, and 10.

[0066] Figure 2 As shown in A, the animal experiment in this example lasted for 31 days, with a total of 4 doses. Plasma samples were collected from mice before and after treatment. Plasma samples (30-40 μL) were collected at two time points: Plasma sample ① - one day before the first treatment (day 2, average tumor volume was approximately 55 mm 3 ) and plasma samples②-before the mice were killed after the end of the experiment (on the 21st day, the average tumor volume was about 1600mm 3 All plasma samples were collected in 1.5 mL EP tubes using K2EDTA solution (25 μL, 15 mg / mL) as an anticoagulant. Serum was also collected from euthanized mice for biochemical analysis.

[0067] Tumors from mice in the control and experimental groups were randomly selected for formalin fixation, paraffin embedding, sectioning, and hematoxylin and eosin (H&E) staining, followed by microscopic analysis.

[0068] 1.2 Results Analysis

[0069] The results showed that RMP1-14 treatment significantly inhibited tumor growth in mice, but did not cause significant differences in body weight ( Figure 2 B). Compared with the control group, the tumor volume and weight of the treatment group were significantly reduced (P<0.01) ( Figure 2 D, E), the relative standard deviation (RSD) of tumor weight in the treatment group was significantly higher than that in the control group. Histological examination of tumor sections from two representative mice showed that the tumor lesions almost completely disappeared after RMP1-14 treatment ( Figure 2 F, G). These results indicate that RMP1-14 treatment was generally effective in this experiment. Importantly, the changes in tumor weights in the treatment groups were more pronounced (higher RSD), suggesting that drug efficacy varied significantly between individuals.

[0070] 2. Plasma Samples 1 H NMR spectrum acquisition

[0071] Prior to NMR analysis, plasma (30 μL) was mixed with an equal volume of phosphate-buffered saline (45 mM, pH = 7.4 ± 0.1). After centrifugation at 11,000 g for 10 minutes at 4°C, the supernatant (50 μL) was transferred to a 1.7 mm NMR tube. NMR data were collected using a Bruker AVIII 600 MHz spectrometer equipped with a cryogenic probe.

[0072] All NMR data were acquired at 25°C using an automated sample exchange system (BACS, Bruker Biospin, Germany). 1 H NMR fingerprint:

[0073] (1) Transverse relaxation time-weighted (T2W-NMR) spectra were obtained using a water-presaturated 1D carr-purcell-meiboomm-gill (CPMG) pulse sequence with a total spin echo time of 70 ms and a recycle delay of 2.0 s.

[0074] (2) When collecting diffusion and relaxation edited spectra DIRE (reference (Liu M, Nicholson JK, London JC. High-resolution diffusion and relaxation edited one- and two-dimensional 1 HNMR spectroscopy of biological fluids. Analytical chemistry 68 (1996) 3370-6), diffusion editing was achieved using three smooth square gradient fields with intensities of -17.13%, 13.13%, and 80%, and gradient lengths and diffusion delays of 1.53 ms and 120 ms, respectively; relaxation editing was achieved using a total spin echo time of 33 ms.

[0075] Two plasma samples were randomly selected to collect their two-dimensional nuclear magnetic resonance spectra ( 1 H- 1 H COSY and TOCSY, 1 H- 13 CHSQC) was used for T2W-NMR spectrum signal identification.

[0076] 3. Data Processing and Identification of Two NMR Fingerprint Signals

[0077] 1.1 Data Processing

[0078] The obtained NMR spectra were processed using TOPSPIN software from Bruker Biospin: all spectra were phase and baseline corrected, the chemical shift of the T2W-NMR spectrum was calibrated with the proton signal of the glucose α-isomer at δ5.23 ppm (commonly used in existing literature), and the DIRE spectrum was calibrated with the -N(CH3)3 proton of the SPC signal at δ3.20 ppm (the best separated peak in the spectrum was selected as the calibration signal), the line width factor was set to 0.3 Hz, and the frequency domain spectrum was then generated by Fourier transformation.

[0079] Then, the signal coverage range of the plasma in the NMR hydrogen spectrum reflected in the two fingerprint spectra (in this embodiment, the chemical shift range of δ0.50ppm to δ9.00ppm) was corrected and segmented integrated using NMRSpec software, with an integration width of 6Hz. In order to eliminate the baseline distortion caused by the suppression of the water peak, the chemical shift regions of δ4.50ppm to δ5.11ppm of the T2W spectrum and δ4.50ppm to δ5.01ppm of the DIRE spectrum were excluded during the segmented integration process to ensure the accuracy of the spectral analysis results. The segmented integration was finally normalized based on the sum of the total integrals of each sample to obtain the above two 1 Dataset of H NMR fingerprint spectra.

[0080] 1.2 Signal Identification

[0081] Based on two-dimensional nuclear magnetic resonance spectroscopy and related literature (Cloarec O, Dumas ME, Craig A, Barton RH, Trygg J, Hudson J, et al. Statistical Total Correlation Spectroscopy: An Exploratory Approach for Latent Biomarker Identification from Metabolic 1HNMR Data Sets. Anal Chem 77 (2005) 1282-89; Nicholson JK, Foxall PJ, Spraul M, Farrant RD, Lindon JC. 750MHz 1H and 1H-13C NMR spectroscopy of human blood plasma. Anal Chem 67 (1995) 793-811.) and the Human Metabolome Database (HMDB) (Wishart DS, Feunang YD, Marcu A, Guo AC, Liang K, Vázquez-Fresno R, et al. HMDB 4.0: the humanmetabolome database for 2018. Nucleic acids research 46 (2018) D608-d17, https: / / doi.org / 10.1093 / nar / gkx1089), Thirty-one metabolites were identified in the T2W-NMR spectra of plasma of mice in the treatment group before administration ( Figure 4 A and Table 1). These metabolites include lipids, amino acids, carbohydrates (glucose, inositol), tricarboxylic acid (TCA) cycle intermediates (citrate, succinate, fumarate), glycoproteins (NAG, OAG), organic acids (formate, acetate), ketones (3-hydroxybutyrate, acetone), choline-containing compounds (choline, phosphocholine), as well as lactate, glycerol, creatine, and urea.

[0082] The main reference for identification of DIRE spectral signals of plasma of mice in the treatment group before administration (Liu M, Nicholson JK, London JC. High-resolution diffusion and relaxation edited one- and two-dimensional 1 H NMR spectroscopy of biological fluids. Analytical chemistry 68 (1996) 3370-6). DIRE (B) spectrum signal identification see Figure 4 B, Three major branched macromolecules were identified: (1) acetylation signals GlycA and GlycB of various N-acetylated glycoproteins, mainly α-1-acid glycoprotein; (2) -N(CH3)3 signals from supramolecular phospholipid complexes; and (3) phospholipid signals of different subfractions of high-density lipoprotein and low-density lipoprotein (SPC-A and SPC-B, respectively).

[0083] 4. Drug sensitivity grouping

[0084] The mice in the treatment and control groups were divided into a high-sensitivity group (HS group) and a low-sensitivity group (LS group) based on the biochemical indicators of the serum after administration and the growth rate of the tumor. A Chemray 800 fully automatic chemical analyzer was used to detect 15 serum biochemical indicators, including liver function (8 items), kidney function (3 items), and blood lipids (4 items). Specifically:

[0085] The above index data were imported into MetaboAnalyst 5.0 online data analysis platform (https: / / www.metaboanalyst.ca / ) for hierarchical clustering heat map analysis. The mice in the treatment group could be divided into two groups: some mice were clustered with the control group, while the remaining mice were clustered in one group. Based on this clustering result, the mice clustered with the control group were classified as the LS group, and the remaining mice were classified as the HS group ( Figure 5 ), and the grouping results are used as classification labels for constructing classification / prediction models based on pre-dose NMR data in the later stage.

[0086] 5. Construction and Evaluation of Classification Models

[0087] The Statistical Analysis [one factor] module of the MetaboAnalyst 5.0 online data analysis platform (https: / / www.metaboanalyst.ca / ) was used to construct a classification model using supervised random forest and support vector machine algorithms based on the T2W and DIRE spectral data of the plasma of the treatment group before administration and the group labels obtained from the above grouping results.

[0088] For T2W spectra, the feature data set is a matrix consisting of the chemical shifts of the characteristic peaks of 31 metabolites and their segmented integral values. For DIRE spectra, the feature data set is a matrix consisting of the chemical shifts of multiple characteristic peaks (53) and their segmented integral values ​​for each identified metabolite.

[0089] Each random forest model consists of 500 decision trees. In order to evaluate the predictive performance of the classifier, the out-of-bag error rate (OOB) was used to evaluate the classification error. The results showed that the OOB error of the random forest classification model constructed based on T2W spectrum data was 0.125 ( Figure 6 A), while the random forest classification model based on DIRE spectrum data showed better classification ability with an OOB error of 0.0625 ( Figure 6 B).

[0090] The results of the support vector machine classifier showed that when the variable levels were less than 6, the error rate of the support vector machine classification model built based on T2W spectrum data was 0.312 ( Figure 7A), while the error rate of the model based on DIRE spectrum data is 0.125 ( Figure 7 B).

[0091] The above results show that the random forest classification model based on pre-dose DIRE spectrum data has the smallest classification error, and this classifier will be selected to screen biomarkers for prediction in the future.

[0092] 6. Biomarker screening and prediction capabilities

[0093] In the BiomarkerAnalysis module of the MetaboAnalyst analysis platform, the pre-dose DIRE spectrum data feature set was input, and the high-frequency selected variable features (8A-1) were screened by the random forest classification algorithm of the module. ROC analysis was performed on the top 5 high-frequency selected variable features (including methylene signals Lipid_8 and Lipid_24 from lipids, N-acetyl signals from GlycA, and sugar signals from GlycA and GlycB), and the AUC was 0.94, with a 95% confidence interval (CI) between 0.667 and 1 ( Figure 8 A-2). Using these five variables, we predicted the 8 HS samples and 8 LS samples in the treatment group. The prediction results showed that except for one sample in the LS group that was misclassified as the HS group, the rest of the samples were correctly grouped ( Figure 8 A-3).

[0094] The above results show that the signal intensity of lipid methylene signal, GlycA and GlyB on the DIRE spectrum of tumor-bearing mouse plasma before treatment can accurately predict their sensitivity to mPD-1 inhibitors (RMP1-14). Figure 8 A-1, lipid signal intensity was significantly positively correlated with low drug sensitivity, while the signal intensity of GlycA and GlyB was significantly correlated with high drug sensitivity.

[0095] Table 1

[0096]

[0097]

[0098]

[0099]

Claims

1. A method for constructing a drug sensitivity prediction model based on blood NMR editing spectrum, the specific steps are as follows: 1) Queue Design According to the ethical requirements of the experiment, a volunteer cohort of patients with a certain chronic non-communicable disease was recruited and divided into two groups: (1-1) Treatment group: blood samples of patients receiving a certain drug treatment, including plasma or serum samples before treatment and plasma or serum samples after treatment; (1-2) Blank control group: plasma or serum samples from patients in the late stage of disease progression who did not receive treatment; 2) Hierarchical cluster analysis was performed on the blood biochemical and prognostic indicators of the treatment group after treatment and the blank control group, and the patients in the treatment group were divided into a high-sensitivity group and a low-sensitivity group; 3) Obtaining modeling data sets for the high-sensitivity group and the low-sensitivity group respectively, wherein the modeling data sets include T2W-NMR spectrum data and DIRE fingerprint data, wherein: The modeling dataset is obtained as follows: (3-1) Obtain one-dimensional plasma or serum samples from patients in the treatment group before treatment in step 1 1 H NMR spectral data, the one-dimensional 1 H NMR spectra were relaxation-weighted T2W-NMR and diffusion-edited relaxation DIRE; (3-2) obtaining a two-dimensional nuclear magnetic resonance spectrum of any plasma or serum sample of the patient in the treatment group before treatment in step 1 to assist in the identification of T2W-NMR spectrum signals; (3-3) The one-dimensional 1 The H NMR spectrum data were processed to obtain the T2W-NMR spectrum and DIRE spectrum data sets: (3-4) T2W-NMR spectral signal identification Signal identification of T2W-NMR spectra of plasma or serum samples based on the two-dimensional nuclear magnetic resonance spectrum (3-2) and the human metabolome database HMDB; (3-5) DIRE spectrum signal identification Signal identification of plasma or serum sample DIRE spectra; 4). Model construction and biomarker screening (4-1) Model construction and evaluation Supervised random forest and support vector machine algorithms were used to model the modeling data sets of plasma or serum samples in the high-sensitivity group and low-sensitivity group before treatment, and four models were constructed. The random forest model consisted of 500 decision trees, and the out-of-bag error (OOB) was used to evaluate the generalization ability of the model; the leave-one-out cross-validation (LOOCV) was used to evaluate the generalization ability of the support vector machine model, and the model with the lowest classification error was selected for drug sensitivity prediction based on the evaluation results.

2. The method according to claim 1, characterized in that In the step (3-1), a transverse relaxation time-weighted T2W-NMR spectrum is obtained using a water-presaturated 1D carr-purcell-meiboomm-gill pulse sequence with a total spin echo time of 70 ms and a recycle delay of 2.0 s; and / or When acquiring the diffusion and relaxation editing spectra DIRE, three smooth square gradient fields were used to achieve diffusion editing, with intensities of -17.13%, 13.13%, and 80%, respectively. The gradient length and diffusion delay were 1.53 ms and 120 ms, respectively. Relaxation editing was achieved using a total spin echo time of 33 ms.

3. The method according to claim 1, characterized in that The specific operations of the (3-3) data processing are: For the one-dimensional 1 H NMR spectra were phase and baseline corrected, and the chemical shift of the T2W-NMR spectra was calibrated using the proton signal of the glucose α-isomer at δ5.23 ppm. The DIRE spectra were calibrated using the best-resolved peak in the spectrum. The linewidth factor was set to 0.3 Hz, and the frequency domain spectra were then generated using Fourier transform. For the one-dimensional 1 The H NMR spectrum was corrected for chemical shift drift and then integrated piecewise with an integration width ranging from 6 to 24 Hz. The piecewise integration was finally normalized based on the sum of the total integrations of each sample to obtain the data sets of the T2W-NMR spectrum and DIRE spectrum, respectively.

4. The method according to claim 1, wherein Refer to the following reference for signal identification of plasma or serum sample DIRE spectra: Liu M, Nicholson JK, London JC. High-resolution diffusion and relaxation edited one- and two-dimensional 1 H NMR spectroscopy of biological fluids.Analytical chemistry 68(1996)3370-6; The identification of two fingerprint spectrum signals is achieved through (3-4) and (3-5), among which the T2W-NMR spectrum is the fingerprint spectrum of small molecule metabolites in the sample, and the DIRE spectrum is the fingerprint spectrum of glycoproteins, lipoproteins and lipid macromolecular metabolites containing small molecule branches in the sample.

5. The method according to claim 4, characterized in that For the T2W-NMR spectrum, the characteristic data set used to establish the model is a matrix consisting of the chemical shifts and segmented integral values ​​of the characteristic peaks of the small molecule metabolites identified in step (3-4); the characteristic data set of the DIRE spectrum is a matrix consisting of the chemical shifts and segmented integral values ​​of the glycoproteins, lipoproteins and lipid macromolecular metabolites containing small molecule branches identified in step (3-5).

6. The method according to claim 1, characterized in that In the method, drugs and disease types with large individual differences in efficacy in clinical applications are selected, where the drugs are cancer immunotherapy drugs PD-1 inhibitors, PD-L1 inhibitors or CTLA-4 inhibitors; and the diseases are non-small cell lung cancer, gastrointestinal cancer or breast cancer.

7. A method for predicting drug sensitivity based on blood NMR edited spectra, using the prediction model constructed by the method of any one of claims 1 to 6, wherein the specific method is as follows: (4-2) Screening of biomarkers: Based on the model constructed by the method of claims 1-6, important features are identified and ROC curve analysis is performed on all features. The ROC curve is generated by Monte Carlo cross validation (MCCV) with balanced subsampling to screen out important features with biomarker functions. The classification prediction ability of the feature variables is then evaluated based on the area under the ROC curve (AUC) of different feature variable combinations to screen out one or more biomarkers for drug susceptibility prediction; 5) Drug sensitivity prediction (5-1) Prepare the data to be predicted: Collect and process the blood NMR data of the patient to be treated according to step 3) and convert the NMR data into the input format expected by the model; (5-2) The obtained NMR data is input into the model, and the biomarkers or their combination screened in step (4-2) are used to predict the classification of the data into the high-sensitivity group and the low-sensitivity group.

Citation Information

Patent Citations

  • Nuclear magnetic resonance model for detecting metabolism micromolecules related to Alzheimer's disease and construction

    CN106198599A

  • Tumor drug sensitivity prediction method, system and equipment and storage medium

    CN115966316A

  • Metabolic phenotyping

    US20050074745A1

  • Lipoprotein insulin resistance indexes and related methods, systems and computer programs for generating same

    US20100100334A1

  • Spectographic metabolite-signature for identifying a subject's susceptibility to drugs

    US20200116703A1