Enrichment detection method for trace drug metabolites in clinical biological sample

Abstract

The invention relates to the technical field of biological medicine detection, and particularly discloses an enrichment detection method for trace drug metabolites in a clinical biological sample. In order to solve the problems that samples are complex, trace substances are difficult to detect and the like, through a nano material-aptamer collaborative enrichment module, a micro-fluidic chip-ultrasonic extraction module, a multi-dimensional chromatography-high-resolution mass spectrometry module and the like, a quantitative algorithm fused by a background noise suppression algorithm based on deep learning and a dynamic programming-genetic algorithm is combined; the efficient enrichment, accurate detection and accurate analysis of the metabolite are realized. Meanwhile, the intelligent quality control module ensures that the result is reliable. According to the method, the detection sensitivity and accuracy are remarkably improved, and powerful support is provided for clinical drug monitoring and disease diagnosis.

Description

Technical Field

[0001] This invention belongs to the field of biomedical detection technology, specifically relating to a method for the enrichment and detection of trace drug metabolites in clinical biological samples. Background Technology

[0002] There are many problems that urgently need to be solved in the detection of trace drug metabolites in clinical biological samples:

[0003] High sample complexity: Clinical biological samples (such as blood, urine, tissues, etc.) have complex compositions, containing a large number of proteins, lipids, endogenous small molecules, etc. These substances can interfere with the detection of drug metabolites, leading to inaccurate test results.

[0004] Trace substance detection is challenging: drug metabolites are present in extremely low concentrations in biological samples, and conventional detection methods are insufficient to meet the requirements for sensitivity and accuracy, easily leading to missed or false detections.

[0005] Low enrichment efficiency: Existing sample pretreatment enrichment methods, such as solid phase extraction and liquid-liquid extraction, have problems such as low enrichment efficiency, cumbersome operation, and long time consumption, and cannot rapidly and effectively enrich trace drug metabolites.

[0006] Data analysis is challenging: The large amount of data obtained lacks effective analysis and processing methods. Traditional data analysis methods struggle to extract valuable information from complex data and cannot accurately identify and quantify trace drug metabolites. Summary of the Invention

[0007] Based on this, the present invention provides a method for the enrichment and detection of trace drug metabolites in clinical biological samples. Through innovative sample processing techniques, detection methods, and algorithm models, it achieves efficient enrichment and accurate detection of trace drug metabolites, providing important evidence for clinical drug monitoring, efficacy evaluation, and disease diagnosis. This includes:

[0008] The nanomaterial-aptamer complex was used to specifically enrich trace amounts of drug metabolites in the sample; rapid extraction was achieved by combining microfluidic chip with ultrasound.

[0009] Separation and detection were performed using multidimensional chromatography-high-resolution mass spectrometry; data processing was performed using a deep learning algorithm based on the fusion of convolutional neural networks and long short-term memory networks.

[0010] Metabolite peak matching and quantification were accomplished using a dynamic programming-genetic algorithm fusion strategy.

[0011] The structure of metabolites was analyzed using a multimodal data fusion algorithm;

[0012] The reliability of testing is ensured by utilizing intelligent quality control and result evaluation modules.

[0013] Furthermore, in the deep learning-based data processing algorithm, the mean squared error loss function is used for...

[0014] Model training, the formula is: Where N is the number of training samples. It is the true signal value of the i-th sample. It is the predicted signal value of the model for the i-th sample. The model parameters are updated iteratively through the Adam optimizer to achieve background noise suppression and enhancement of drug metabolite features.

[0015] Furthermore, in the dynamic programming-genetic algorithm fusion strategy:

[0016] The cost function C constructed by the dynamic programming algorithm is expressed as: ,in, Let be the weighting coefficient, satisfying , For the difference in peak retention time, Due to the difference in mass-to-charge ratio, To determine the peak intensity difference, the genetic algorithm evaluates the quality of individuals using a fitness function and performs iterative optimization according to the following steps:

[0017]

[0018] in, Denotes the population of generation t. For individuals fitness value, For crossover probability, The mutation probability, This indicates the combination of operations to ultimately obtain the optimal peak matching and quantitative parameters.

[0019] Furthermore, in the nanomaterial-aptamer composite, the nanomaterial is a magnetic nanoparticle or a mesoporous silica nanoparticle, and the aptamer is modified on the surface of the nanomaterial through chemical coupling.

[0020] Furthermore, the microfluidic chip is equipped with a serpentine microchannel and a vortex reaction chamber, and the ultrasonic frequency is 20-40kHz.

[0021] Furthermore, in the multidimensional chromatography-high-resolution mass spectrometry technique, two-dimensional reversed-phase hydrophilic liquid chromatography is coupled with time-of-flight mass spectrometry, and the chromatographic gradient elution conditions are dynamically adjusted according to the sample characteristics.

[0022] Furthermore, the multimodal data fusion algorithm employs a sparse representation model, projecting different modal data onto a shared feature space for fusion.

[0023] Furthermore, in the intelligent quality control and result evaluation module, the fuzzy logic evaluation rules are set based on the signal-to-noise ratio of the detection signal, peak area repeatability, and standard substance recovery rate.

[0024] Furthermore, the clinical biological samples undergo centrifugation and filtration pretreatment before testing.

[0025] Furthermore, after the nanomaterial-aptamer complex is enriched, solid-liquid separation is performed using magnetic separation or centrifugal separation.

[0026] Beneficial effects:

[0027] Highly efficient enrichment: Based on the specific enrichment module of nanomaterial-aptamer synergy and the rapid extraction module of microfluidic chip-ultrasound synergistic enhancement, the enrichment efficiency and extraction speed of trace drug metabolites are significantly improved, the interference of sample matrix is ​​reduced, and high-quality samples are provided for subsequent detection.

[0028] Precise detection: The separation and detection module of multidimensional chromatography-high-resolution mass spectrometry, combined with deep learning-based background noise suppression and feature enhancement algorithms, achieves high sensitivity and high resolution detection of trace drug metabolites, improves the quality and accuracy of detection signals, and reduces the probability of missed detections and false detections.

[0029] Accurate Analysis: The dynamic programming-genetic algorithm-integrated metabolite peak matching and quantification algorithm module and the multimodal data fusion metabolite structure analysis algorithm module can accurately perform peak matching, quantitative analysis and structural analysis of drug metabolites, providing reliable data support for clinical drug research and disease diagnosis.

[0030] Intelligent quality control: The intelligent quality control and result evaluation module enables real-time monitoring of the testing process and intelligent evaluation of results, ensuring the reliability and accuracy of test results and improving the quality and efficiency of testing work.

[0031] Wide range of applications: The enrichment detection method of the present invention is applicable to the detection of various clinical biological samples and drug metabolites. It has good versatility and scalability and can be widely used in clinical drug monitoring, new drug development, disease diagnosis and other fields. Attached Figure Description

[0032] Figure 1 Flowchart of the metabolite peak matching and quantification algorithm module integrating dynamic programming and genetic algorithm;

[0033] Figure 2 Flowchart of the algorithm module for metabolite structure analysis based on multimodal data fusion;

[0034] Figure 3Flowchart of a specific enrichment module based on the synergistic effect of nanomaterials and aptamers. Detailed Implementation

[0035] Example 1

[0036] This embodiment verifies the effectiveness of the enrichment and detection method for human plasma samples (200 μL volume, expected metabolite concentration 5-100 pg / mL) containing metabolites of the antidepressant sertraline. Through specific enrichment using nanomaterial-aptamer complexes, microfluidic ultrasonic extraction, multidimensional chromatography-high-resolution mass spectrometry, and deep learning data processing, trace amounts of metabolites (such as norsertraline) are detected, with a target detection limit of 0.5 pg / mL and a quantitative error ≤5%.

[0037] II. Materials and Equipment Preparation

[0038] Preparation of nanomaterial-aptamer complexes

[0039] Nanomaterials: Magnetic nanoparticles (Fe3O4@SiO2, particle size 50-100nm, specific surface area 200m2 / g, purchased from Sigma-Aldrich);

[0040] Aptamer: Single-stranded DNA aptamer for norsertraline (Sequence:

[0041] 5I-GCTACGTACGTCGATCGATCG-3I, purified by HPLC, synthesized by Shanghai Sangon Biotech.

[0042] Coupling reagents: 3-aminopropyltriethoxysilane (APTES) and glutaraldehyde (25% aqueous solution), purchased from AIfaAcesar.

[0043] microfluidic chip

[0044] Material: PDMS (polydimethylsiloxane), with internally etched serpentine microchannels (500μm width, 200μm depth, 10cm total length) and eddy current reaction chambers (1mm diameter, 300μm depth, 10 chambers), ultrasonic probe (adjustable frequency 20-40kHz, 50W power, Shenzhen Komeda). Chromatography-Mass Spectrometry Equipment

[0045] Multidimensional chromatographic system: Two-dimensional reversed-phase hydrophilic liquid chromatography (first dimension: C18 column,

[0046] 150mm × 2.1mm, 3.5μm; Second dimension: HILIC column, 100mm × 2.1mm, 2.7μm, Agilent).

[0047] High-resolution mass spectrometry: Time-of-flight mass spectrometry (TripleTOF6600+, Sciex), electrospray ionization source (ESI+).

[0048] Data processing platform

[0049] Hardware: NVIDIA RTX 3090 GPU, Intel i9-12900K CPU;

[0050] Software: Python 3.8 (TensorFlow 2.8 framework), self-written dynamic programming-genetic algorithm fusion program (MATLAB R2021b).

[0051] III. Specific Experimental Steps

[0052] 1. Preparation and enrichment of nanomaterial-aptamer complexes

[0053] Magnetic nanoparticle modification:

[0054] Fe3O4@SiO2 particles (100 mg) were added to 10 mL of LAPTES ethanol solution (5% v / v), stirred at 60 °C for 24 h, centrifuged (8000 rpm, 10 min) and washed 3 times to obtain amination magnetic particles.

[0055] The amination particles were reacted with glutaraldehyde (final concentration 2.5%) at room temperature for 2 h, centrifuged, and then aptamer solution (10 μM, 1 mL Tris-HCl buffer, pH 7.4) was added. The mixture was incubated overnight at 4 °C with shaking. After magnetic separation, the mixture was used with 0.1 M...

[0056] The aptamer-modified magnetic nanoparticles (MNPs-Apt) were obtained by washing three times with NaCl and PBS buffer.

[0057] Plasma sample enrichment:

[0058] A 200 μL plasma sample (containing norsertraline metabolites) was centrifuged at 12,000 rpm for 10 min, and the supernatant was filtered through a 0.22 μm filter membrane.

[0059] Add MNPs-Apt (10 mg / mL, 50 μL), incubate at 37°C with shaking for 30 min (shaking speed 200 rpm), separate with an external magnetic field for 1 min, discard the supernatant, and wash the precipitate twice with 100 μL PBS (pH 7.4).

[0060] 2. Microfluidic chip ultrasonic extraction

[0061] Chip pretreatment: The PDMS chip was ultrasonically cleaned with ethanol-water (1:1) for 10 min, dried with nitrogen, and activated with 0.1% Triton X-100 solution for 5 min;

[0062] Extraction process:

[0063] Add 50 μL of methanol-water (8:2, v / v) extraction solution to the MNPs-Apt precipitate and transfer it to the chip injection port;

[0064] The ultrasonic frequency was set to 30kHz, the power to 50%, the peristaltic pump was driven at a flow rate of 100μL / min, the extract was kept in the serpentine channel for 3min, and the vortex chamber was used to enhance mass transfer.

[0065] The extract was collected into a centrifuge tube, dried under nitrogen at 40°C, reconstituted with 20 μL of mobile phase (0.1% formic acid-acetonitrile, 9:1), and filtered through a 0.22 μm filter membrane before analysis.

[0066] 3. Multidimensional chromatography-high-resolution mass spectrometry separation and detection

[0067] Two-dimensional chromatographic conditions:

[0068] First dimension (reversed-phase chromatography): C18 column, column temperature 35℃, mobile phase A: 0.1% formic acid water, mobile phase B: acetonitrile, gradient elution: 0-5 min, 5-20% B; 5-10 min, 20-40% B; 10-15 min, 40-80% B; flow rate 0.3 mL / min, injection volume 5 μL;

[0069] Second dimension (hydrophilic chromatography): HILIC column, column temperature 30℃, mobile phase C: 10mM ammonium acetate (pH 4.5), mobile phase D: acetonitrile, gradient elution: 0-3 min, 90-70% D; 3-6 min, 70-50% D; flow rate 0.2 mL / min, two-dimensional separation achieved using valve switching technology. Mass spectrometry parameters:

[0070] Ion source: ESI+, spray voltage 5.5kV, atomizing gas pressure 50psi, auxiliary gas temperature 500℃;

[0071] Scan range: m / z 100-1000, acquisition rate 2 spectra / s, resolution ≥40,000 FWHM (m / z 200).

[0072] 4. Deep Learning Data Processing and Metabolite Analysis

[0073] Data preprocessing based on CNN-LSTM:

[0074] The raw mass spectrometry data were converted into a three-dimensional matrix of m / z intensity-retention time and divided into a training set (80%), a validation set (10%), and a test set (10%).

[0075] Construct a CNN-LSTM fusion network: CNN layers (3 convolutional layers, kernel size 3×3, stride 1) extract local features, and LSTM layers (2 layers, 128 units) process time series correlations;

[0076] Training parameters: mean squared error loss function, Adam optimizer (learning rate 0.001, β1=0.9, β2=0.999), 50 iterations, batch size 32, achieving background noise suppression (signal-to-noise ratio improved by 3.2 times).

[0077] Dynamic Programming-Genetic Algorithm Peak Matching:

[0078] Construction of the dynamic programming cost function: In the dynamic programming-genetic algorithm fusion strategy, the cost function C constructed by the dynamic programming algorithm is expressed as: , where, , Let be the weighting coefficient, satisfying , For the difference in peak retention time, Due to the difference in mass-to-charge ratio, The difference in peak intensity is considered; the genetic algorithm evaluates the quality of individuals through a fitness function and performs iterative optimization according to the following steps:

[0079]

[0080] in, Denotes the population of generation t. For individuals fitness value, For crossover probability, The mutation probability, This indicates the combination of operations to obtain the optimal peak matching and quantitative parameters. The differences are: retention time (≤0.5 min), mass-to-charge ratio (≤5 ppm), and peak intensity (≤20%).

[0081] Genetic algorithm parameters: population size 100, crossover probability =0.8, mutation probability =0.05, iterated for 100 generations, and finally obtained the optimal matching peak of the norsertraline metabolite (retention time 12.3 min, m / z 276.1543).

[0082] Multimodal data fusion and structure analysis:

[0083] Data such as chromatographic retention time, precise mass number from mass spectrometry, and secondary fragment ions are integrated and projected onto a shared feature space using a sparse representation model (projection matrix dimension 128).

[0084] By matching with databases (HMDB, Metlin), the structure of the metabolite was determined to be norsertraline (theoretical m / z 276.1547, error 2.1 ppm).

[0085] To address the complexity of detecting trace drug metabolites in clinical biological samples (such as blood and urine), the overall workflow is optimized across the entire chain through modular design:

[0086] Sample pretreatment: Centrifugation (12000 rpm, 10 min) and filtration through a 0.22 μm filter membrane were used to remove macromolecular impurities and particulate matter, laying the foundation for subsequent enrichment.

[0087] Specific enrichment: Using aptamers modified on the surface of nanomaterials (magnetic nanoparticles or mesoporous silica), trace metabolites are targeted and captured through the principle of antigen-antibody specific binding. Combined with magnetic separation or centrifugation (8000 rpm, 10 min), rapid solid-liquid separation is achieved, and the enrichment efficiency is improved by more than 40% compared with traditional methods.

[0088] High-efficiency extraction: The microfluidic chip has a built-in serpentine microchannel (500μm wide and 200μm deep) and a vortex reaction chamber. Combined with 20-40kHz ultrasonic-assisted extraction, the extraction time is shortened to 3-5 minutes by enhancing mass transfer through turbulence, and the extraction efficiency is increased by 2.8 times.

[0089] Separation and detection: Two-dimensional reversed-phase hydrophilic liquid chromatography coupled with quadrupole time-of-flight mass spectrometry is used. Quadrupole pre-screening of target ions reduces interference, and time-of-flight mass spectrometry provides high-resolution mass numbers (accuracy ≤ 5 ppm), enabling accurate separation and signal acquisition of metabolites in complex matrices.

[0090] Intelligent analysis: Deep learning algorithms (CNN-LSTM fusion model) suppress background noise (improving signal-to-noise ratio by 3.2 times), dynamic programming-genetic algorithm fusion strategy achieves peak matching (success rate of over 95%) and accurate quantification (error ≤ 5%), multimodal data fusion analyzes metabolite structure, and finally, an intelligent quality control module (based on signal-to-noise ratio ≥ 10, RSD ≤ 5%, and recovery rate of 80-120%) ensures reliable results.

[0091] The key role of column equilibration: After each chromatographic gradient elution, a 5-10 minute column equilibration step is forcibly introduced to restore the mobile phase to its initial ratio (e.g., 5% acetonitrile for one-dimensional reversed-phase chromatography, and 90% acetonitrile for two-dimensional hydrophilic chromatography). This ensures that the stationary phase of the chromatographic column is fully balanced, guaranteeing that the retention time deviation is ≤0.5 minutes during continuous injection, and meeting the repeatability requirements of high-throughput detection (e.g., 100+ samples per day).

[0092] The synergistic advantages of quadrupole-time-of-flight mass spectrometry: The quadrupole acts as an ion filter, selectively transmitting target mass-to-charge ratio (m / z) ions and reducing matrix interference ions entering the time-of-flight mass spectrometer; the time-of-flight mass spectrometer accurately determines the mass number through the ion time-of-flight difference. Combining the advantages of both, the detection limit is as low as 0.5 pg / mL, making it suitable for the analysis of pg-level trace metabolites.

[0093] 5. Intelligent quality control and result evaluation

[0094] Quality control parameters:

[0095] Signal-to-noise ratio (S / N) ≥ 10 (defined as peak intensity / baseline noise);

[0096] Peak area repeatability (RSD) ≤ 5% (n = 6 repeated detections);

[0097] The recovery rate of the standard substance was 80-120% (with the addition of 5 pg / mL norsertraline standard).

[0098] Result evaluation:

[0099] Fuzzy logic rule: If S / N≥10, RSD≤5%, and recovery rate 80-120%, the result is reliable; otherwise, trigger duplicate detection or method optimization.

[0100] In this embodiment, the detection signal S / N=18.5, RSD=3.2%, and recovery rate of 92.3% meet the quality control standards.

[0101] IV. Experimental Results and Discussion

[0102] Enrichment efficiency verification

[0103] The enrichment recovery of 5 pg / mL norsertraline using MNPs-Apt reached 91.7% (n=3), which is 40.6% higher than the traditional solid phase extraction (SPE) method (65.2% recovery), demonstrating the advantage of aptamer-specific binding.

[0104] Magnetic separation takes only 1 minute, which is 10 times more efficient than centrifugal separation (10 minutes).

[0105] Microfluidic ultrasonic extraction effect

[0106] At an ultrasonic frequency of 30kHz, the extraction efficiency is increased by 2.8 times compared to conditions without ultrasound. The vortex chamber increases the mass transfer coefficient by 1.5 times, and the extraction time is shortened from 30 minutes for traditional liquid-liquid extraction to 3 minutes.

[0107] Chromatography-mass spectrometry separation and detection performance

[0108] Two-dimensional chromatography effectively separates interfering substances from target metabolites in complex plasma matrices, with a peak capacity of 4000, which is 5 times higher than that of one-dimensional chromatography (peak capacity 800).

[0109] The high-resolution mass spectrometry has a mass accuracy of ≤5ppm, enabling accurate identification of metabolites, which meets the technical features of claim 7.

[0110] Data processing and quantitative accuracy

[0111] The CNN-LSTM algorithm reduces background noise by 67% and achieves a feature peak recognition accuracy of 98.3%, which is a significant improvement over the traditional baseline correction method (accuracy of 82.5%).

[0112] The dynamic programming-genetic algorithm fusion strategy achieved a peak matching success rate of 95.6% and a quantitative error of 2.8%, meeting the accuracy requirements.

[0113] V. Conclusion

[0114] This embodiment achieves efficient detection of trace drug metabolites in clinical biological samples through specific enrichment of nanomaterial-aptamer complexes, microfluidic ultrasonic extraction, multidimensional chromatography-high-resolution mass spectrometry, and deep learning data processing. Specifically, it verifies that:

[0115] Preparation process and enrichment efficiency of magnetic nanoparticle-aptamer complex.

[0116] The structural design of the serpentine channel and vortex chamber of the microfluidic chip, and the ultrasonic extraction parameters of 20-40kHz;

[0117] A technique for coupling two-dimensional reversed-phase hydrophilic chromatography with time-of-flight mass spectrometry;

[0118] Training process of CNN-LSTM deep learning algorithm and peak matching strategy of dynamic programming-genetic algorithm;

[0119] Evaluation rules for intelligent quality control modules.

[0120] Experimental results show that this method has a low detection limit, fast analysis speed, and accurate quantification, making it suitable for high-throughput detection of trace drug metabolites in clinical biological samples and fully supporting the technical solution in the claims.

Claims

1. A method for the enrichment and detection of trace drug metabolites in clinical biological samples, characterized in that, include: s1. Utilizing nanomaterial-aptamer complexes to specifically enrich trace amounts of drug metabolites in samples; s2. Rapid extraction is achieved through a combination of microfluidic chip and ultrasound; s3. Separation and detection were performed using multidimensional chromatography-high-resolution mass spectrometry. s4. Data processing is performed using a deep learning algorithm based on the fusion of convolutional neural networks and long short-term memory networks; s5. Metabolite peak matching and quantification were achieved based on a dynamic programming-genetic algorithm fusion strategy; s6. A multimodal data fusion algorithm was used to analyze the structure of metabolites; s7. Utilize intelligent quality control and result evaluation modules to ensure testing reliability.

2. The enrichment detection method according to claim 1, characterized in that, In the deep learning-based data processing algorithm, the mean squared error loss function is used for model training, and the formula is: Where N is the number of training samples. It is the true signal value of the i-th sample. It is the predicted signal value of the model for the i-th sample. The model parameters are updated iteratively through the Adam optimizer to achieve background noise suppression and enhancement of drug metabolite features.

3. The enrichment detection method according to claim 1, characterized in that, In the dynamic programming-genetic algorithm fusion strategy: the cost function C constructed by the dynamic programming algorithm is expressed as: ,in, For the weighting coefficients, satisfying , For the difference in peak retention time, Due to the difference in mass-to-charge ratio, The peak intensity difference is considered; the genetic algorithm evaluates the quality of individuals through a fitness function and performs iterative optimization according to the following steps: in, Denotes the population of generation t. For individuals fitness value, For crossover probability, The mutation probability, This indicates the combination of operations to ultimately obtain the optimal peak matching and quantitative parameters.

4. The enrichment detection method according to claim 1, characterized in that, In the nanomaterial-aptamer composite, the nanomaterial is a magnetic nanoparticle or a mesoporous silica nanoparticle, and the aptamer is modified on the surface of the nanomaterial by chemical coupling.

5. The enrichment detection method according to claim 1, characterized in that, After the nanomaterial-aptamer complex is enriched, solid-liquid separation is performed using magnetic separation or centrifugal separation.

6. The enrichment detection method according to claim 1, characterized in that, The microfluidic chip is equipped with a serpentine microchannel and a vortex reaction chamber, and the ultrasonic frequency is 20-40kHz.

7. The enrichment detection method according to claim 1, characterized in that, In the aforementioned multidimensional chromatography-high-resolution mass spectrometry (LC-MS), two-dimensional reversed-phase hydrophilic liquid chromatography is coupled with time-of-flight mass spectrometry (TOFMS), and the chromatographic gradient elution conditions are dynamically adjusted according to the sample characteristics.

8. The enrichment detection method according to claim 1, characterized in that, The multimodal data fusion algorithm employs a sparse representation model, projecting different modal data onto a shared feature space for fusion.

9. The enrichment detection method according to claim 1, characterized in that, In the intelligent quality control and result evaluation module, the fuzzy logic evaluation rules are set based on the signal-to-noise ratio of the detection signal, peak area repeatability, and standard substance recovery rate.

10. The enrichment detection method according to any one of claims 1-8, characterized in that, The clinical biological samples were pretreated by centrifugation and filtration before testing.

Images