A nanopore-based non-targeted detection device, system and detection method thereof

Abstract

The application provides a non-target detection device and system based on a nanopore and a detection method thereof, the non-target detection device and system based on a nanopore are used for non-target detection of a complex matrix through a nanopore, feature extraction is performed on a nanopore detection graph, and sample classification is realized through model construction, the detection method does not make a prior definition on a to-be-detected target, and various molecular information existing in serum is acquired in a net catching manner, non-target capture and overall state analysis of various molecules in serum are realized, and a new way of rapid and portable disease screening is provided.The application breaks through the limitation of traditional single marker detection by combining the nanopore technology and machine learning, and is expected to promote the development of precision medicine in a more efficient and comprehensive direction.

Description

Technical Field

[0001] This invention relates to the field of biomedical detection technology, specifically to a non-targeted detection device, system, and detection method based on nanopores. Background Technology

[0002] The onset and development of most diseases are accompanied by the appearance or concentration changes of certain specific molecular markers. Detecting disease-related molecular markers in complex body fluid matrices such as serum is an effective method for early disease diagnosis. However, the concentrations of relevant molecular markers are generally at extremely low levels in the early stages of disease, and some currently known markers lack high specificity (i.e., changes in many molecular markers are not limited to a single specific disease but may occur in multiple diseases and even under normal physiological conditions). Although comprehensive diagnostic methods using multiple molecular markers can greatly improve detection accuracy, many diseases still lack clearly defined molecular markers with high specificity and sensitivity, directly affecting early diagnosis, accurate classification, and effective treatment.

[0003] Serum contains a wealth of disease-related molecular information, including proteins, nucleic acids, and small metabolic molecules. Comprehensive analysis of serum biomarkers is crucial for early screening, diagnosis, and classification of diseases. Currently, clinical and laboratory detection of serum biomarkers primarily relies on the specificity of known disease-related molecular biomarkers. Specific detection of one or more disease-related molecular biomarkers is employed; for example, traditional PCR and ELISA techniques are suitable for detecting nucleic acid and protein biomarkers, respectively. Non-targeted omics-based techniques utilize a "full-spectrum analysis" strategy, where changes in the composition and concentration of various molecular biomarkers in different samples are reflected by differences in omics spectra or spectral data, which are then combined with machine learning algorithms for rapid classification. However, these methods have significant limitations: PCR and ELISA for specific detection of molecular biomarkers heavily rely on predefined targets and specific probes, allowing only "fishing"-style single-indicator detection. They cannot unbiasedly capture a population of molecules with low specificity or undefined molecular biomarkers in serum, easily missing important, unexpected biomarker information. Even some emerging nanopore sequencing technologies mostly focus on nucleic acid sequence analysis, which is difficult to apply directly to serum environments with complex components and interfering substances such as high salt and high protein, and cannot characterize the molecular state spectrum of serum as a whole.

[0004] It is evident that methods for quantitative detection of nucleic acid and protein molecular markers based on PCR and ELISA, respectively, are limited to the diagnosis of diseases with known markers, and their detection sensitivity and throughput are insufficient. These methods often fail in disease detection scenarios where the markers are unclear. Methods based on nanopores for simultaneous detection of multiple types of molecular markers improve detection sensitivity and throughput, but are still limited by the predictability of marker information.

[0005] The method of classification based on non-targeted omics combined with machine learning is an innovation of the traditional concept of detection based on specific biomarkers. By incorporating molecular information other than known biomarkers in the original sample into the analysis model, it provides a more comprehensive description of the overall state of the sample. However, omics technology often involves relatively complex sample labeling and sample pretreatment, resulting in a long analysis process, relatively high cost, and insufficient detection sensitivity.

[0006] There is an urgent need to find a more portable, faster, and more accurate method for non-targeted detection of complex samples, which can truly achieve rapid and comprehensive molecular information capture of complex biological samples without the need for predefined targets, and can effectively perform qualitative analysis of the overall state. Summary of the Invention

[0007] To address the problems existing in the prior art, this invention provides a non-targeted detection device, system, and method based on nanopores. This system utilizes solid-state quartz nanopore devices with a specific pore size to detect electrical signals in complex human serum matrices. The detection target is not predefined; instead, it collects multiple molecular information from the serum matrix through the combination of positive and negative potentials and the enrichment effect of microfluidic channel sample cells. This technical solution can obtain nanopore electrical signal spectra corresponding to different serum samples. Multidimensional feature extraction is then performed on the signal spectra, and further combined with machine learning algorithms to classify serum samples with different states (e.g., disease / health). Compared to traditional methods that only detect one or more biomarkers (PCR, ELISA, etc.), this invention has a wider range of applications and obtains richer information. Compared to broad-spectrum biomarker detection based on non-targeted omics, this invention achieves single-molecule detection sensitivity and eliminates the need for complex sample pretreatment, which is beneficial for large-scale, low-cost sample detection.

[0008] On one hand, the present invention provides a non-targeted detection device based on nanopores, the device comprising nanopores and a microfluidic sample cell; the microfluidic sample cell is used to collect samples, which are then detected through the nanopores.

[0009] Unlike the traditional "fishing" model of qualitative and quantitative detection and analysis (which only detects specific molecular markers related to diseases), the single-molecule nanopore detection technology provided by this invention combines artificial intelligence analysis methods and adopts a "full-spectrum analysis" strategy to establish a "net-trapping" biosensor model (comprehensive detection of multiple types of molecules). It reflects the changes in the composition and content of various molecular markers in different samples through differences in electrical signal spectra. Furthermore, it combines machine learning algorithms to achieve rapid classification of different types of samples. This not only improves the sensitivity of disease detection to the single-molecule level, but also breaks through the bottleneck of early disease detection by simultaneously detecting multiple types of biomolecules to fully obtain sample information.

[0010] The nanopores described in this invention are channels with a nanoscale size (generally a few nanometers to tens of nanometers). Under the action of an applied electric field, biomolecules migrate through the pores in sequence, and sensing is achieved by detecting the changes in current generated when molecules pass through the pores.

[0011] In some embodiments, the nanopore is a solid-state quartz nanopore device. Solid-state quartz nanopores can withstand complex biochemical environments and, unlike biological nanopores, will not deform or degrade when exposed to proteases, lipids, or extreme environments. Solid-state quartz nanopores can stably and reliably capture a wider range of diverse molecular information from complex serum matrices, providing a robust, versatile, and high-performance physical platform.

[0012] Microfluidic channel sample cells, with their microliter-level volume, utilize precise microchannel design to guide the sample into directional laminar flow, actively and continuously transporting spatially dispersed analyte molecules to the nanopore sensing region. This fundamentally alters the traditional "passive waiting" mode of molecule-dependent random diffusion in homogeneous solutions, significantly increasing the collision frequency between molecules and nanopores, thereby enhancing capture efficiency. Simultaneously, the combined effect of reduced detection volume, active transport, and local enrichment effectively reduces background noise interference in the bulk solution, resulting in a higher signal-to-noise ratio. Therefore, microfluidic channel sample cells can capture more effective molecular information per unit time, achieving a simultaneous improvement in detection throughput and sensitivity.

[0013] Furthermore, the pore size of the nanopore is 20-50 nm.

[0014] Research has shown that by using solid quartz nanoporous devices with a certain pore size to detect electrical signals in complex human serum matrices, the electrical signal spectra of nanopores corresponding to different serum samples can be obtained. Then, multidimensional feature extraction is performed on the signal spectra, and accurate classification of different samples is achieved through machine learning models.

[0015] When a smaller pore size is selected, fewer nanopore signals are detected; when a larger pore size is selected, nanopore blockage is more likely to occur during the detection process. Therefore, the preferred nanopore size is 20-50 nm.

[0016] In some approaches, it is preferable to use nanopores with a diameter of 30-40 nm to ensure relatively more stable signal acquisition when detecting samples.

[0017] Furthermore, the microfluidic sample cell includes a glass tube with an inner diameter of 1.5 mm to 2.5 mm and its support frame; the nanopores are placed in the glass tube for sample detection.

[0018] The inner diameter of the glass tube is 1.5 mm to 2.5 mm, a result that balances hydrodynamic performance, ease of use, and manufacturing cost. The glass tube diameter is set to be slightly larger than the overall outer diameter of the nanopore probe. The inner diameter of the glass tube cannot be too large, otherwise an effective capillary effect cannot be generated, and the sample consumption also increases with the increase of the glass tube diameter. The inner diameter of the glass tube should not be too small either, otherwise the risk of damage to the nanopores during the insertion of the nanopore probe may increase, and the preparation difficulty will also increase.

[0019] Furthermore, the support frame allows the glass tube to be placed at an angle of 10 to 30 degrees; it also includes an electrical control module, which includes a working electrode, a reference electrode, a patch-clamp amplifier, and a digital-to-analog converter.

[0020] Microchannel sample cells can be fabricated by tilting a glass tube and placing it on a support frame.

[0021] Placing the glass tube (microchannel) at an angle facilitates the placement and replacement of nanopore probes during testing.

[0022] In some configurations, the tilt angle is 15 degrees.

[0023] The inner diameter and tilt angle of the glass tube (microfluidic channel) are set by theoretical calculation to meet the range that can generate effective capillary action, that is, to balance the hydrostatic pressure through capillary pressure, thereby confining the solution inside the tube.

[0024] In some methods, the working electrode is located inside the capillary containing the nanopore, while the reference electrode is located in the sample within the glass tube. By setting a bias voltage between the working and reference electrodes (bias voltage = voltage of the working electrode - voltage of the reference electrode), non-targeted detection of the sample solution can be achieved at a specific bias voltage.

[0025] A patch-clamp amplifier and a digital-to-analog converter (DAC) are used to control the bias voltage between the working electrode and the reference electrode and to acquire nanopore detection spectra. The patch-clamp amplifier amplifies the weak ion current signal (pA level) from the nanopore and converts it into a voltage signal. The DAC further converts this into a digital signal for subsequent computer storage and analysis. During detection, the user first sets the target bias voltage value (digital signal) in the computer software. This command is converted into an analog voltage command by the DAC and sent to the patch-clamp amplifier. Upon receiving this command, the patch-clamp amplifier applies a stable and precise bias voltage to the electrolyte on both sides of the nanopore through its internal high-precision voltage circuit and negative feedback system, actively maintaining this voltage constant, unaffected by environmental interference. Simultaneously, the ion current within the nanopore is detected, amplified, and converted into a voltage signal in real time by the patch-clamp amplifier. This analog signal, reflecting current changes, is digitized at high speed and with high precision by the DAC and transmitted back to the computer for analysis and recording.

[0026] In another aspect, the present invention provides a non-targeted detection system based on nanopores, the system comprising a detection module and an analysis module; the detection module includes detecting samples using the non-targeted detection device described above; the analysis module is used to analyze the nanopore detection spectrum obtained by the detection module, extract features and construct a model to classify the samples.

[0027] The non-targeted detection system provided by this invention consists of two parts: a detection device and an intelligent analysis algorithm. The detection device includes a microfluidic channel sample cell and support, a solid-state quartz nanopore device, and an electrical control module. The microfluidic channel sample cell has a microliter volume, which greatly reduces sample consumption, enabling micro-volume detection, and introduces a capillary effect to improve the nanopore's molecular capture efficiency through local enrichment. The electrical control module can control the bias voltage between the working electrode and the reference electrode and acquire the nanopore detection spectrum.

[0028] Furthermore, the detection module uses an alternating positive and negative potential method to detect the sample.

[0029] This invention employs an alternating method of positive and negative potentials for non-targeted detection, which greatly enhances the long-term stability and reliability of the system when detecting complex samples, while also solving the technical pain point of easy contamination of solid pores.

[0030] The combined use of positive and negative potentials can force molecules with different net charges in the same batch of samples to pass sequentially through nanopores. Positively charged molecules are efficiently captured under negative pressure, while negatively charged molecules are captured under positive pressure. This ensures unbiased and comprehensive capture of molecular groups with diverse charges in the sample, avoiding the detection of certain types of molecules due to unidirectional electrophoretic forces. When the same molecule passes through the nanopore under positive and negative voltages, it generates distinctly different blocking current signals due to differences in its charge, conformation, and interaction with the pore channel. By capturing and correlating this bidirectional signal, a unique "electrical fingerprint" is generated for each molecule. This provides high-dimensional feature data combining charge properties and structural information for subsequent machine learning algorithms, forming the core information foundation for accurate and rapid qualitative analysis of serum status.

[0031] Meanwhile, the dynamic reversal of the potential can generate strong electroosmotic and electrophoretic forces, which can effectively remove biomolecules adsorbed or temporarily blocked in the pores, allowing the nanopores to maintain their optimal working state.

[0032] In some methods, the alternation of positive and negative potentials is achieved by setting a bias voltage between the working electrode and the reference electrode, where a bias voltage greater than 0 represents a positive potential, and a bias voltage less than 0 represents a negative potential. The use of positive and negative potentials means that after a period of time, a positive potential is applied, followed by a period of time of negative potential. For example, a bias voltage of 100mV is applied to the working electrode for 5 minutes, followed by a bias voltage of -100mV for 5 minutes.

[0033] The duration of the specific positive and negative potentials can be selected as needed. Experiments have shown that when the duration of the positive and negative potentials is no less than 3 minutes, it can be used for accurate and highly sensitive non-targeted detection of complex samples. In some methods, a duration of 5 minutes for the positive and negative potentials is preferred.

[0034] Furthermore, it also includes a sample diluent, which includes any one or more of KCl, NaCl, LiCl, TE buffer, Tris-HCl buffer, HEPES buffer, and phosphate buffer.

[0035] The present invention requires only 2 μL of serum sample for a single test, and no complicated pretreatment is required. The serum sample can be directly detected by nanopore after being diluted with buffer.

[0036] Using different buffer solutions can affect the signals collected by nanopores. For example, when the type of buffer, the type of salt ions, the KCl concentration, and the pH value of the buffer are changed, the nanopore signal spectrum of the same sample will change. This change can sometimes even further affect the classification accuracy of machine learning models.

[0037] In some methods, using 1M KCl and 1×TE buffer for non-targeted detection further improves detection efficiency. The high concentration of KCl (1M) provides a strong monovalent ion environment, ensuring a high signal-to-noise ratio and a large baseline current in the nanopore, making the blocking signal generated during serum molecule translocation more significant and easier to detect. The 1×TE buffer maintains the pH near the physiological range to preserve the stability of biomolecules in serum. The EDTA in the buffer chelates any divalent metal ions (such as Mg²⁺ and Ca²⁺) that may be present in the serum, preventing them from affecting the nanopore current or causing non-specific interactions.

[0038] Furthermore, the features include feature parameters extracted from four dimensions: time domain, frequency domain, time-frequency domain, and nonlinear domain.

[0039] Furthermore, the sample includes any one or more of serum, plasma, urine, tears, bile, cerebrospinal fluid, saliva, and body fluids.

[0040] The non-targeted detection system provided by this invention can be extended to other complex matrices similar to human serum, such as plasma, urine, tears, bile, cerebrospinal fluid, saliva, and other body fluids. Combined with machine learning algorithms, it enables rapid differentiation of various sample types with high classification accuracy.

[0041] Furthermore, the classification of the samples includes any one or more of the following: differentiation between disease and health, differentiation between different types of disease, pathogen identification, and cancer subtyping.

[0042] In some embodiments, the disease includes any one or more of the following: stomach cancer, breast cancer, endometrial cancer, cervical cancer, and ovarian cancer.

[0043] In another aspect, the present invention provides a complex matrix detection system based on nanopores, which comprises a solid quartz nanopore device, a sample end processing scheme, a measurement technology scheme, a measurement platform, and a data analysis module.

[0044] In some embodiments, the solid-state quartz nanoporous device has a diameter of 30-40 nm.

[0045] In some approaches, the sample processing scheme includes centrifuging and diluting clinical serum samples. Gradient centrifugation technology can be introduced to separate components with different molecular weights in the serum to meet various testing needs.

[0046] In some methods, the measurement technique includes direct detection of diluted samples and positive and negative potential detection methods. This technique can directly detect diluted serum samples without complicated extraction and purification operations. The combination of positive and negative potentials can achieve the capture of molecules with different charges in serum.

[0047] In some embodiments, the measurement platform incorporates a microfluidic channel sample cell, which improves the efficiency of nanopore capture by limiting the range of free diffusion of molecules, while also facilitating sample replacement; this measurement technique and platform can provide comprehensive acquisition of information on multiple types of molecules in the measured serum sample.

[0048] In some embodiments, the data analysis module includes a multidimensional feature extraction algorithm and a machine learning classification model, which extracts feature parameters from the sample nanopore signal in four dimensions: time domain, frequency domain, time-frequency domain, and nonlinear domain, and combines the machine learning model to achieve accurate classification.

[0049] Extracting features from multiple dimensions (time domain, frequency domain, time-frequency domain, and nonlinear domain) and combining them with machine learning for classification represents the best practice and core development trend in current single-molecule nanopore signal analysis, especially for complex biomedical applications (such as pathogen identification, cancer classification, and protein differentiation).

[0050] On another aspect, the present invention provides a non-targeted detection method based on nanopore technology, which uses the non-targeted detection device or the non-targeted detection system described above to perform detection, directly contacting the sample with the nanopore to obtain a nanopore detection spectrum, and classifying the sample by constructing a model after analysis and feature extraction.

[0051] This invention proposes a method for rapid analysis of complex serum matrices using nanopore pan-biomarker detection combined with a machine learning model. Nanopore technology possesses single-molecule sensitivity, enabling more precise acquisition of molecular-level information from serum samples. Simultaneously, by incorporating multiple types of molecules in the sample for detection and analysis, a comprehensive description of the original serum sample's state is obtained. Furthermore, by combining this with a machine learning model, accurate typing of different serum samples can be achieved.

[0052] Furthermore, the detection module uses an alternating positive and negative potential method to detect the sample.

[0053] On the other hand, the present invention provides a modular and portable complex matrix detection device. The core components of the detection device include a replaceable encapsulated solid-state nanoporous device or arrayed acquisition chip, a signal acquisition device integrating a microfluidic channel sample cell and a miniaturized circuit module, and an intelligent data analysis platform with multi-dimensional feature extraction and multi-classification functions. The portable detection device is advantageous for use in a variety of real-world scenarios other than laboratories.

[0054] In another aspect, the present invention provides the use of a device for preparing a nanopore for non-targeted detection of generalized biomarkers in samples.

[0055] On another aspect, the present invention provides an application of a non-targeted detection device or non-targeted detection system based on nanopores in various scenarios, including but not limited to biomedicine, agriculture, environment and other fields, as well as applications in various complex matrices such as serum, urine, bile, and cerebrospinal fluid. The detection system is suitable for the detection and differentiation of various complex matrices, and the detection method has the advantages of simple operation and low cost.

[0056] On another note, this invention provides an application of the concept of pan-biomarker detection using single-molecule nanopore technology in early disease diagnosis.

[0057] The present invention has the following beneficial effects:

[0058] 1. Truly achieve non-targeted detection: By synergistically integrating solid quartz nanopores, microfluidic channels and bidirectional potential bias system, the limitation of traditional detection methods relying on predefined markers is overcome;

[0059] 2. Direct detection capability: For complex matrices such as serum samples, non-targeted detection can be performed without complicated pretreatment, greatly simplifying the analysis process;

[0060] 3. More comprehensive detection results: The active transport and enrichment effect of microfluidic channels, combined with the unbiased driving of molecules with different charges by bidirectional potential, ensures that molecular groups with different charges, from small metabolites to large protein complexes, in serum can be fully captured, thereby obtaining an "electrical signal spectrum" that characterizes the overall molecular state of serum, providing an unprecedented comprehensive information dimension for disease screening.

[0061] 4. High sensitivity: Nanopore technology provides single-molecule sensitivity, enabling effective non-targeted detection of low-concentration molecular markers in the early stages of disease, and achieving accurate differentiation between different samples;

[0062] 5. The microfluidic channel design not only reduces the detection volume, but also concentrates low-abundance markers locally near the nanopores through an active enrichment mechanism, which greatly improves the capture rate and detection sensitivity.

[0063] 6. The two-way potentiometric coupling strategy endows nanopores with strong self-cleaning ability, which can effectively remove adsorbates inside the pores, ensuring the reliability and high throughput of long-term detection in complex serum matrices, and solving the core pain point of easy contamination of solid pores.

[0064] 7. The combined use of bidirectional potentials enables the same molecule to generate unique "bidirectional translocation signals" under voltages of different polarities. This set of signals simultaneously contains information about the molecule's size, charge properties, and dynamic interactions with the pores, constituting a high-dimensional molecular fingerprint;

[0065] 8. This multidimensional electrical characteristic data is far better able to distinguish different molecules than signals under a single polarity, providing a key and high-quality data source for subsequent use of machine learning algorithms to quickly and accurately classify and characterize serum status.

[0066] 9. Low cost and capability for large-scale testing;

[0067] 10. Scalability: The method proposed in this invention can be extended to the detection and analysis of many other complex matrices, and has wide applications in biomedicine, agriculture, environment and other fields. Attached Figure Description

[0068] Figure 1 This is a schematic diagram of a non-targeted detection device based on nanopores;

[0069] Figure 2 This is a schematic diagram of the nanopore preparation process;

[0070] Figure 3 The diagram shows the electrical properties of nanopores.

[0071] Figure 4 SEM characterization of the nanopore structure;

[0072] Figure 5 A flowchart illustrating the detection process using a non-targeted detection device based on nanopores.

[0073] Figure 6 Here is a flowchart of the data processing algorithm;

[0074] Figure 7 This is a spectrum of electrical signals from nanopores. Detailed Implementation

[0075] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be noted that the embodiments described below are intended to facilitate understanding of the present invention and are not intended to limit it in any way. The reagents used in this embodiment are all known products, and unless otherwise specified, they are all commercially available products.

[0076] Example 1: Non-targeted detection device based on nanopores

[0077] The nanopore-based non-targeted detection device provided in this embodiment is as follows: Figure 1 As shown, it includes a nanopore 1, a microfluidic sample cell 2, and an electrical control module 3. The microfluidic sample cell 2 is used to collect samples, which are then detected through the nanopore 1 under the control of the electrical control module 3, and the detection spectrum is collected.

[0078] Nanopore 1 is a solid-state quartz nanopore device with a pore size of 30-40 nm. The fabrication process of nanopore 1 is as follows: Nanopore 1 is fabricated using a laser-assisted pipette puller (Sutter Instrument, P-2000, USA) by pulling a quartz capillary 4 (Sutter Instrument QF100-50-7.5). Figure 2 A capillary tube 4 is aligned and secured in the groove, positioned on the puller's support, and then pulled using optimized two-wire pulling parameters to create two conical nanopores, employing a two-step pulling procedure:

[0079] Line1 HEAT: 650; FIL: 4; VEL: 30; DEL: 170; PUL: 70,

[0080] Line2 HEAT: 680; FIL: 3; VEL: 20; DEL: 145; PUL: 130.

[0081] The obtained nanopores 1 have a diameter of 30-40 nm, and their structure was verified using SEM. The geometry of the pores is significantly affected by the instrument's condition and environmental conditions (including laser intensity, temperature, and humidity). The experimental conditions in this embodiment were an ambient temperature of approximately 25°C and an ambient humidity of approximately 35%.

[0082] Solid-state quartz nanopores prepared by laser-assisted drawing were characterized. Before being used for complex sample testing, all nanopores underwent IV characterization tests in a buffer solution (1M KCl, 1×TE, pH≈8) without the sample being tested. The results are as follows: Figure 3 As shown. The nanoporous structure was characterized by SEM, and the results are as follows. Figure 4 As shown, the prepared nanopores possess excellent basic electrical properties and functionality, as well as superior electrical stability and reliability.

[0083] like Figure 1 As shown, the microfluidic sample cell 2 includes a glass tube 5 and a support frame 6. The inner diameter of the glass tube 5 is 1.5 mm to 2.5 mm, and it is placed above the support frame 6. The support frame 6 tilts the glass tube 5 at an angle of 10 to 30 degrees (preferably 15 degrees in this embodiment).

[0084] like Figure 1 The electrical control module 3 includes a working electrode 7, a reference electrode 8, a patch-clamp amplifier, and a digital-to-analog converter 9. The working electrode 7 is located inside the capillary 4 containing the nanopore 1, and the reference electrode 8 is located in the sample in the glass tube 5. The patch-clamp amplifier and the digital-to-analog converter 9 are used to control the bias voltage between the working electrode 7 and the reference electrode 8, and to acquire the nanopore detection spectrum.

[0085] Example 2: Non-targeted detection method based on nanopores

[0086] This embodiment uses the nanopore-based non-targeted detection device prepared in Example 1 for detection. The detection flowchart is shown below. Figure 5 The specific process is as follows:

[0087] (1) Sample pretreatment: Blood samples were collected using EDTA blood collection tubes. After standing and coagulation, the supernatant was collected and then centrifuged at 3000 rpm at room temperature for 10 min. The samples were then further diluted 10-fold with 1M KCl and 1×TE buffer solution (KCl is an electrolyte that provides an ionic environment for nanopore detection; 1×TE buffer solution provides a relatively stable buffer environment for nucleic acid, protein and other molecules in serum, with a pH of approximately 8). The 1M KCl and 1×TE buffer solution was prepared as follows: 745.5 mg of KCl powder was weighed using a balance and dissolved in 10 mL of 1×TE buffer solution.

[0088] (2) Electrical signal acquisition: The conical end of the nanopore was immersed in the glass tube (pore diameter 1.5 mm) of the microfluidic channel sample cell. At the same time, 1 M KCl and 1×TE buffer solution (injection volume approximately 10 μL) were injected into the quartz capillary using a Microfil microneedle. The sample cell contained diluted serum buffer. Ag / AgCl electrodes (AgCl side immersed in the solution, Ag side connected to the circuit) were placed in the quartz capillary and the glass tube of the sample cell, respectively. The working electrode was connected to the end of the quartz capillary, and the reference electrode was connected to the end of the sample cell. Bias voltages of +100 mV and -100 mV were applied successively, and the data were recorded for 5 min each time. The electrical signal spectrum was then acquired.

[0089] (3) Machine Learning Model: For electrical signals acquired under positive and negative bias, absolute value inversion is performed, all information is merged, wavelet transform is applied to the global signal sequence to remove baseline drift, then a constant-width sliding window is used to continuously slice the sequence and extract multi-dimensional features, including time domain, frequency domain, time-frequency domain, and nonlinear domain signal features, and then classification is performed through supervised machine learning. The data processing algorithm flow is as follows: Figure 6 As shown.

[0090] In this embodiment, blood samples were also collected from 46 healthy volunteers, 46 breast cancer patients, and 46 gastric cancer patients. Each test group sample was diluted 10-fold with 1M KCl and 1×TE buffer solution before non-targeted detection using the method described above. Representative detection patterns of the three types of blood samples are shown below. Figure 7 As shown, 0~300s and 300~600s are the test data under bias voltages of 100mV and -100mV, respectively.

[0091] A machine learning model was built using Python software. A multi-dimensional feature extraction strategy was employed to comprehensively characterize the complex characteristics of serum molecules passing through nanopore current signals across four levels: time domain, frequency domain, time-frequency domain, and nonlinear domain. In the time domain, peak detection (identifying local maxima in the signal whose amplitude exceeds a certain threshold) was performed by setting a threshold, and peak counts, peak mean, and peak standard deviation were extracted as local transient features of the signal. Simultaneously, statistical moments were calculated to characterize the overall signal distribution, including the overall signal mean, signal standard deviation, signal skewness, and signal kurtosis. In the frequency domain, Fast Fourier Transform was applied to extract core frequency parameters such as the signal's dominant frequency, energy, and spectral entropy, revealing its frequency composition. Furthermore, continuous wavelet transform was introduced for time-frequency analysis. Through multi-scale energy statistics and spectral entropy calculation, the dynamic distribution and complexity of signal energy in the two-dimensional time-frequency plane were analyzed. Finally, at the nonlinear dynamics level, a complementary feature system was constructed: the Hjorth parameter rapidly quantifies the time-domain energy, frequency variation, and regularity of the signal from a statistical perspective; while the fractal dimension based on the Higuchi algorithm measures the self-similarity and complexity of the signal waveform from a geometric perspective.

[0092] A classification model was built using 10-fold cross-validation. First, 138 blood samples (46 healthy, 46 breast cancer, and 46 gastric cancer) were randomly and stratified into 10 folds of similar size, ensuring the class proportions in each fold were consistent with the original dataset. Then, 10 iterations were performed. In each iteration, one fold was reserved as the test set, and the remaining nine folds were combined into the training set to train an independent classification model. The performance of this model was evaluated using the test set. After the iterations, the results of the 10 tests were summarized, and metrics such as average accuracy were calculated as robust estimates of the model's generalization ability. Finally, based on the optimal parameters determined through this validation, a final model for actual prediction was retrained using all 138 samples. This method makes full use of the sample with limited data and provides a reliable and unbiased performance evaluation.

[0093] In addition, models were constructed using different feature combinations to examine the impact of models constructed using different feature combinations on sample classification. The results are shown in Table 1.

[0094] Table 1. Impact of different feature combinations on detection results

[0095]

[0096] According to the table above, when time-domain statistical features, Fourier spectrum features, power spectrum features, wavelet features, and nonlinear features are used together to construct a machine learning model, the model achieves the best performance and the highest accuracy for sample classification.

[0097] The machine learning model built using this example classified the above 138 blood samples with an accuracy of 94%. Among them, the sensitivity for identifying breast cancer patients reached 80% and the specificity reached 96%, and the sensitivity for identifying gastric cancer patients reached 89% and the specificity reached 90%.

[0098] The results were validated in another batch of the same number of validation samples, and the accuracy reached 94.8%. Among them, the sensitivity for judging breast cancer patients reached 84% and the specificity reached 96%, and the sensitivity for judging gastric cancer patients reached 90% and the specificity reached 91%.

[0099] Example 3: The Influence of Microfluidic Channels on Detection

[0100] This embodiment uses the detection equipment provided in Example 1 and conducts experiments according to the experimental method in Example 2. A control group was established for the microfluidic channel; the microfluidic channel was removed, and a traditional large-volume sample cell (a cylindrical container with a diameter of 8 mm, a height of 5 mm, and a volume of approximately 200 μL) was used. The results for different glass tube diameters were also compared. Thirty healthy blood samples, 30 samples each of gastric cancer blood samples, and 30 samples of breast cancer blood samples were used. Using a nanopore detection device with other conditions kept consistent, the samples were detected for 5 minutes under both positive and negative bias voltages. The average value was taken, and a machine learning model was constructed based on the detection patterns to examine its accuracy for sample classification. The results are shown in Table 2 below.

[0101] Table 2. The Influence of Microfluidic Channels on Detection

[0102]

[0103] As shown in Table 2, microfluidic channels increased the molecular information capture amount of each serum sample by more than 2 times, and the classification accuracy was also improved by more than 20% compared to direct addition. Therefore, the synergistic integration of microfluidic channels and nanopore sensors is key to achieving high-performance, high-reliability serum biomarker detection. Meanwhile, microfluidic channels prepared with glass tubes of different pore sizes also directly affect the detection results and the classification accuracy of the constructed model. This is likely because the pore size directly affects the intensity of the capillary effect, further influencing the diffusion distribution of serum molecules in the tube. Glass tubes with smaller diameters produce a more significant local enrichment effect on serum molecules, and the average distance of molecule transport to the nanopore detection range is shorter, resulting in higher molecular capture efficiency. However, the diameter of the glass tube cannot be too small. Since the outer diameter of the nanopore probe is 1 mm, the glass tube diameter should not be less than this value. Furthermore, when the glass tube diameter is less than 1.5 mm, the effect on improving the molecular information capture amount and model classification accuracy is no longer significant, and it also increases the risk of damage during the insertion of the nanopore probe into the microfluidic channel. Therefore, the optimal aperture of the glass tube is 1.5~2.5mm, with 1.5mm being the most preferred.

[0104] Example 4: The effect of alternating positive and negative potentials

[0105] This embodiment uses the detection equipment provided in Embodiment 1 and conducts experiments according to the experimental method in Embodiment 2. Three groups are established for the driving voltage mode: a single positive voltage group, a single negative voltage group, and a bidirectional alternating voltage group. Thirty healthy blood samples, 30 blood samples from gastric cancer, and 30 blood samples from breast cancer are used. A nanopore detection device with other conditions kept consistent is used. Detection is performed for 5 minutes under both positive and negative voltages. Individual positive bias data, individual negative bias data, and combined positive and negative bias data are collected to construct a machine learning model. The accuracy of this model for sample classification is examined. The results are shown in Table 3 below.

[0106] Table 3. Effect of bidirectional potential bias

[0107]

[0108] As shown in Table 3, the number of effective translocation events captured by bidirectional potentials, the signal-to-noise ratio, and the classification accuracy are all significantly higher than those of unidirectional potentials. Unidirectional potentials perform "partial sampling" of serum molecular communities, while bidirectional potentials achieve "panoramic capture," ensuring the integrity of diagnostic information and significantly improving the classification accuracy of the constructed model.

[0109] Example 5: Screening of Nanopore Diameter

[0110] This embodiment uses the detection device provided in Example 1 and conducts experiments according to the experimental method in Example 2. For the nanopore diameter, three groups were selected: 20-30 nm, 30-40 nm, and 40-50 nm. Five healthy blood samples were used, and the nanopore detection device with other conditions kept consistent was employed. Each sample was tested for 5 minutes under the same parameters, and the average value was taken. The results are shown in Table 4 below.

[0111] Table 4. Effect of nanopore diameter on detection

[0112]

[0113] As shown in Table 4, fewer nanopore signals are detected when a smaller pore size is used; when a larger pore size is used, nanopore blockage is more likely to occur during the detection process. Therefore, when the nanopore diameter is 30-40 nm, a wide range of molecular sizes can be captured while maintaining a good signal-to-noise ratio and acceptable stability.

[0114] Example 6: Selection of Buffer Solution

[0115] This embodiment uses the detection device provided in Example 1 and conducts experiments according to the experimental method in Example 2. For the serum buffer, multiple treatment groups were designed according to Table 5: 100mM KCl + 1×TE, 500mM KCl + 1×TE, 1 M KCl + 1×TE, 2M KCl + 1×TE, 1 M LiCl + 1×TE, 1 M NaCl + 1×TE, 1 M KCl + 1×HEPES, and 1 M KCl + 1×PBS. Thirty samples were randomly selected from each of the three blood samples described in Example 1 for each group. All three treatment groups used the same nanopore and were tested for 5 minutes under the same parameters. The results are shown in Table 5 below.

[0116] Table 5. Effect of bidirectional potential bias

[0117]

[0118] As shown in Table 5, the 1 M KCl + 1×TE buffer exhibits significant advantages over other buffers, substantially improving the accuracy of non-target sample classification. This is likely because 1 M KCl provides strong electrophoretic force and efficient molecular transport, while 1×TE prevents molecular aggregation and pore adsorption caused by divalent ions, thus reducing clogging frequency. Through synergistic action, both achieve the highest molecular transport efficiency and optimal pore stability, generating the largest number and highest quality of effective translocation events, thereby enhancing the classification of non-target samples.

[0119] While the present invention has been disclosed above, it is not limited thereto. Any person skilled in the art can make various modifications and alterations without departing from the spirit and scope of the invention; therefore, the scope of protection of the present invention should be determined by the scope defined in the claims.

Claims

1. A non-targeted detection system based on nanopores, characterized in that, The system includes a detection module and an analysis module. The detection module includes a non-targeted detection device for sample detection. The non-targeted detection device includes nanopores and a microfluidic sample cell. The microfluidic sample cell is used to collect samples and includes a glass tube. The nanopores are placed in the glass tube for sample detection. The pore size of the nanopores is 20-50 nm. The microfluidic sample cell includes a glass tube with an inner diameter of 1.5 mm to 2.5 mm and its support frame. The detection module uses an alternating positive and negative potential method to detect the samples. The analysis module is used to analyze the nanopore detection spectrum obtained by the detection module, extract features and build a model to classify the samples; the classification of the samples includes any one or more of the following: differentiation between disease and health, differentiation between different types of disease, pathogen identification, and cancer subtyping.

2. The non-targeted detection system as described in claim 1, characterized in that, The support frame allows the glass tube to be placed at an angle of 10 to 30 degrees; it also includes an electrical control module, which includes a working electrode, a reference electrode, a patch-clamp amplifier, and a digital-to-analog converter.

3. The non-targeted detection system as described in claim 1, characterized in that, It also includes a sample diluent, which includes any one or more of KCl, NaCl, LiCl, TE buffer, Tris-HCl buffer, HEPES buffer, and phosphate buffer.

4. The non-targeted detection system as described in claim 1, characterized in that, The features include feature parameters extracted from four dimensions: time domain, frequency domain, time-frequency domain, and nonlinear domain; the samples include any one or more of serum, urine, tears, bile, cerebrospinal fluid, saliva, and body fluids.

Images