Bionic sensing system based on single-chiral carbon nanotube sensing array and in-vitro detection method

Abstract

The invention relates to the technical field of biological detection, in particular to a bionic sensing system based on a single-chiral carbon nanotube sensing array and a detection method. The invention aims to solve the technical problems of poor universality, high cost and the like of the existing detection technology. The invention provides a bionic sensing system based on a single-chiral carbon nanotube sensing array. The bionic sensing system comprises the sensing array, a signal detection device and a data processing and analyzing module, wherein the sensing array is composed of a plurality of sensing units, and each sensing unit adopts a single-chiral carbon nanotube. According to the system, pure and independent signals generated by each single chiral sensing unit are collected and combined into a characteristic'signal fingerprint ', so that high-sensitivity analysis and distinguishing of different detection targets (complex biological fluid / various micromolecules) can be realized on the same platform without a preset specific recognition element (such as an antibody). The invention provides a novel in-vitro detection normal form which is simple in process, low in cost, accurate in result and high in universality.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of biosensing and nanotechnology, and particularly relates to a biomimetic sensing system based on single-chiral carbon nanotube sensing array and an in vitro detection method. BACKGROUND

[0002] In the field of biological sample analysis, traditional detection techniques mainly rely on specific recognition strategies, such as antigen-antibody binding or nucleic acid aptamer-target recognition. This "one-to-one" detection mode, although effective in some scenarios, faces increasingly prominent fundamental challenges: first, the complexity and high cost of multiplex detection: a single marker often lacks sufficient specificity to indicate complex biological states, but many complex biological samples (such as biological fluids) often require simultaneous detection of multiple biomarkers. Developing high-specificity recognition elements (such as monoclonal antibodies) for each marker and stably integrating them into the same detection platform is extremely complex in technology, and has a long development cycle and high cost; second, the contradiction between specificity and cross-reaction: ideal high specificity is difficult to achieve, and recognition elements often cross-react with non-target molecules in the sample, especially for highly similar physiological molecules such as amino acids, affecting the accuracy of detection; third, technical bottleneck for unknown target detection: when the detection target is not clear or potential unknown markers need to be screened from complex samples, the specific recognition strategy relying on pre-set targets is completely unsuitable, causing a huge gap in technology.

[0003] To overcome the limitations of specific detection, the present technology draws on the principles of biological olfaction and taste and develops a new technology paradigm called "biomimetic sensing system". The "sensing array + pattern recognition" technical architecture is used to realize intelligent analysis of the sample to be measured. The core idea is usually to use a sensing array composed of multiple sensing units with different response characteristics to interact with the measured substance to produce a multi-dimensional and characteristic "signal fingerprint". Subsequently, the "fingerprint map" is analyzed by machine learning and other pattern recognition algorithms, thereby realizing the classification or category recognition of the overall state of the sample.

[0004] DNA-modified carbon nanotubes (DNA-SWCNTs) are considered a highly promising sensing material due to their availability, excellent optical / electrical properties, unique sensitivity to different physicochemical environments, and long-term stability. However, existing publicly available technologies generally employ mixtures of carbon nanotubes with different chiralities to construct sensing arrays (Kim, M., Chen, C., Wang, P. et al., Detection of ovarian cancer via the spectral fingerprinting of quantum-defect-modified carbon nanotubes in serum by machine learning. Nat. Biomed. Eng. 6, 267–275 (2022); Zvi Yaari et al., A perception-based nanosensor platform to detect cancer biomarkers. Sci. Adv. 7, eabj0852 (2021)). This strategy has led to insurmountable bottlenecks in the practical application of current biomimetic sensing systems.

[0005] First, there is the issue of signal quality: Currently, commercially available mixtures containing various chiral carbon nanotubes are used. Due to the potential overlap of fluorescence spectra among different chiral carbon nanotubes, and the existence of complex competitive interactions between carbon nanotubes and between carbon nanotubes and analytes within the system, the final sensing signal becomes blurred, chaotic, and difficult to resolve. This severely reduces the sensitivity and resolution of the sensor. Second, there is the issue of functional limitations: Limited by the aforementioned poor signal quality, the detection capabilities of existing sensing systems based on mixed chiral DNA-SWCNTs are significantly limited. They can only distinguish between highly different sample types, achieving coarse binary classification. They cannot effectively identify subtle differences or gradual changes in characteristics between samples caused by minute variations in multiple components, which is common in biological samples. This severely restricts the application of sensing systems in precise detection and complex sample analysis.

[0006] Therefore, there is an urgent need in this field to develop a novel biomimetic sensing system that can fundamentally solve the signal quality problem, enabling it to achieve accurate, high-sensitivity, and high-resolution universal analysis of complex samples or highly similar samples of multiple categories, while also having the advantages of low cost and ease of operation. Summary of the Invention

[0007] To address the problems of blurred sensing signals and low resolution caused by the use of mixed chiral materials in existing carbon nanotube-based biomimetic sensing technologies, and the poor versatility and high cost of traditional detection technologies that rely on specific identification, this invention aims to provide a novel, highly sensitive, and high-resolution universal sensing technology solution to achieve rapid, accurate, and low-cost analysis of various in vitro samples.

[0008] The present invention provides a single-chiral carbon nanotube sensing array comprising at least two sensing units; the sensing units are DNA-single-chiral carbon nanotube complexes (DNA-scCNTs); the DNA-scCNTs of the at least two sensing units differ in at least one of the DNA sequences or the chirality of the carbon nanotubes.

[0009] More preferably, the sensor array consists of three or more sensing units.

[0010] Furthermore, the sensing array can be implemented in the following ways: (1) Liquid array: In one embodiment, the array is embodied as a multi-container device (e.g., a 96-well plate), wherein different DNA-scCNTs are respectively contained in different containers of the device in solution form; (2) Solid array: In another embodiment, different sensing units are fixed on different predetermined regions of the same substrate (e.g., silicon wafer, glass or flexible substrate) to form a chip-based sensing array, such as a microfluidic chip, a flexible sensing device or an electrochemical electrode array.

[0011] Furthermore, the DNA-scCNT complex can be obtained by blending a specific DNA sequence with carbon nanotubes, followed by ultrasonic treatment, and then separation by density gradient ultracentrifugation and gel chromatography in an aqueous two-phase system.

[0012] More preferably, the DNA-scCNT is separated and purified using an aqueous two-phase system.

[0013] The present invention also provides a biomimetic sensing system based on a single-chiral carbon nanotube sensing array, the sensing system comprising the following parts:

[0014] (1) Sensor array: A sensor array of any of the aforementioned forms;

[0015] (2) Detection device: used to detect the response signal generated by each sensing unit when the array comes into contact with the sample to be tested in vitro, and to combine these signals into a multi-channel signal group;

[0016] (3) Data processing and analysis module, used to receive the multi-channel signal group and analyze it through pattern recognition algorithm to obtain the category or status information of the sample to be tested.

[0017] Furthermore, the detection device can be an optical readout device (preferably a near-infrared fluorescence spectrometer), an electrochemical detection device, or a combination thereof; correspondingly, the response signal includes at least one of fluorescence intensity change, fluorescence wavelength shift, resistance change, or conductivity change.

[0018] Furthermore, the data processing and analysis module performs the following operations:

[0019] (1) Data preprocessing: Standardize the raw multi-channel signal (e.g., Z-fractional standardization) to eliminate the influence of dimensions;

[0020] (2) Feature extraction: Optionally, dimensionality reduction algorithms (e.g., principal component analysis, PCA) can be used to process the standardized data, analyze the data structure, and guide the selection of machine learning algorithms;

[0021] (3) Pattern recognition is performed using machine learning algorithms, including but not limited to support vector machine (SVM), artificial neural network (ANN), random forest (RF), decision tree (DT), and gradient boosting tree (XGBoost).

[0022] More preferably, the SVM algorithm can be used.

[0023] The present invention also provides an in vitro detection method based on the biomimetic sensing system, the detection method comprising the following steps:

[0024] (1) Provide system: Provide the aforementioned sensing system, wherein the data processing and analysis module of the system is configured with a pre-trained pattern recognition model;

[0025] (2) Sample contact: Make the sensing array of the system come into contact with the sample to be tested outside the body;

[0026] (3) Signal acquisition: Use a detection device to acquire the response signals of each sensing unit to obtain a multi-channel signal group;

[0027] (4) Analysis and identification: Using the data processing and analysis module, the signal group is analyzed through the pre-trained model to obtain the classification or state information of the sample to be tested.

[0028] Furthermore, the detection method can use a non-specific sensing and recognition strategy; the non-specificity is reflected in the fact that it can analyze samples without using preset, specific recognition elements (such as antibodies or aptamers) for specific target molecules, and usually does not require pretreatment steps such as separation and purification of complex samples.

[0029] Furthermore, the sample to be tested can be a complex biological fluid (such as blood, serum, urine, saliva) that has not been separated into its components, or a solution of structurally similar biomolecules (such as neurotransmitters, amino acids).

[0030] The present invention also provides a microfluidic chip on which the aforementioned sensor array is integrated.

[0031] The present invention also provides a flexible sensing device, wherein the aforementioned sensing array is integrated on a flexible substrate.

[0032] The present invention also provides an electrode array, wherein the electrode surface is modified or integrated with the aforementioned sensing array.

[0033] Compared with the prior art, the advantages and beneficial effects of the present invention are as follows:

[0034] (1) Fundamentally improves signal quality and detection sensitivity: By employing single-chiral carbon nanotubes in each sensing channel, this invention eliminates spectral overlap and signal interference caused by chiral mixing at the source. This enables the system to obtain a pure, clear, and high signal-to-noise ratio response signal, thereby distinguishing more subtle differences between samples and greatly improving the sensitivity and resolution of the analysis.

[0035] (2) Achieving true versatility and high flexibility: This invention adopts a "sensor array + pattern recognition" architecture, eliminating the need for complex design and synthesis of specific recognition elements. It enables the construction of "fingerprint maps" for a variety of heterogeneous samples, ranging from complex biological fluids to structurally similar small molecules, on a single platform. Furthermore, by combining different DNA sequences with different chiral carbon nanotubes in diverse ways, the scale and diversity of the sensor array can be easily expanded, rapidly adapting to various detection scenarios and needs.

[0036] (3) Significantly simplified process and reduced cost: Since there is no need to pre-design and prepare high-cost specific recognition agents for specific target molecules, and usually no need to perform complex separation and purification pretreatment on the sample, the method of the present invention is simpler and faster, significantly reducing manpower, time and economic costs.

[0037] (4) Improved accuracy in identifying complex samples: By combining high-quality “fingerprint” with advanced machine learning algorithms, this invention can effectively learn and analyze subtle difference patterns between samples, significantly improving the ability to identify and distinguish complex biological samples or molecules with highly similar structures. Attached Figure Description

[0038] Figure 1The main spectral properties of the six DNA-scCNT nanosensors used in Example 1 for detecting different serum samples are shown in Figure 1.

[0039] Figure 2 This is a standardized fluorescence response data and statistical analysis graph of each sensing unit of the DNA-scCNT sensing array in Example 1 after 24 hours of exposure to serum samples from different populations.

[0040] Figure 3 This is a standardized fluorescence response data heatmap of each sensing unit after the DNA-scCNT sensing array in Example 1 was exposed to serum samples from different populations for 24 hours.

[0041] Figure 4 This is a diagram illustrating the algorithm optimization process for applying three different machine learning models in Example 1.

[0042] Figure 5 The image shows the training results and validation diagram of the identification model for liver cancer, lung cancer, ovarian cancer, and healthy samples using the SVM algorithm in Example 1.

[0043] Figure 6 This is a sensitivity comparison chart of the DNA-scCNT sensing system in Example 1 for identifying three types of cancer serum samples and existing cancer serum sample detection methods based on tumor markers.

[0044] Figure 7 This is a performance evaluation of the DNA-scCNT sensing system for identifying early cancer models in Example 1, and a sensitivity comparison with tumor marker methods.

[0045] Figure 8 The fluorescence spectra of three DNA-scCNT nanosensors used in Example 2 for detecting different neurotransmitters in a DNA-single-chiral carbon nanotube sensing array are shown.

[0046] Figure 9 The fluorescence response curves and thermograms of each sensing unit in Example 2 are shown for the three DNA-scCNT sensing arrays (6,5), (9,1), and (8,3) after exposure to three neurotransmitters: 5-hydroxytryptamine (5-HT), dopamine (DA), and levodopa (L-DA).

[0047] Figure 10 The figure shows the training and cross-validation results of the recognition model for 5-HT, DA, and L-DA samples using the SVM algorithm in Example 2.

[0048] Figure 11The near-infrared fluorescence emission spectra of the four DNA-scCNT sensing units used for amino acid recognition in Example 3 of the present invention are used for purity and optical property characterization.

[0049] Figure 12 This is a response heatmap composed of the normalized fluorescence intensity change (I-I0) / I0 and wavelength shift (λ-λ0) of each sensing unit after the 4-channel sensing array is exposed to 18 different amino acids, according to Embodiment 3 of the present invention.

[0050] Figure 13 According to Embodiment 3 of the present invention, different machine learning algorithms (Support Vector Machine (SVM), Random Forest, etc.) are employed.

[0051] The graph compares the performance of RF, logistic regression (LR), multilayer perceptron (MLP), extreme gradient boosting tree (XGB), and decision tree (DT) in identifying 18 amino acids. The graph shows the F1 score and accuracy of each algorithm.

[0052] Figure 14 This is a cross-validation performance evaluation diagram of the optimized support vector machine (SVM) algorithm in the task of identifying 18 amino acids according to Embodiment 3 of the present invention.

[0053] Figure 15 This is a schematic diagram of a sensing system.

[0054] Figure 16 This is a flowchart of the detection method. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of the present invention clearer, the specific implementation of the present invention will be described in detail below with reference to the accompanying drawings and embodiments.

[0056] According to an embodiment of the present invention, the sensing array may be composed of at least two sensing units, and in a further embodiment, the sensing unit of the sensing array is a DNA-single-chiral carbon nanotube complex (DNA-scCNT).

[0057] Regarding "single-chiral carbon nanotubes": In this article, single-chiral carbon nanotubes refer to single-walled carbon nanotubes (scCNTs) with specific chirality (precisely defined by the coil vector (n,m)). In the ultraviolet-visible-near-infrared absorption spectrum, they exhibit clearly distinguishable characteristic peaks unique to this chirality, including primary (E11, near-infrared band) and secondary (E22, ultraviolet-visible band) peaks; in the near-infrared fluorescence spectrometer, they exhibit a single characteristic emission peak (E11) and a triexciton peak (Trion) near its (longer wavelength) wavelength.

[0058] Regarding "DNA": The DNA used to form a complex with scCNTs is an oligonucleotide. In a preferred embodiment of the invention, the length of the oligonucleotide can be 6 to 18 bases (6-18mer). There are no particular limitations on the sequence of the oligonucleotide; various sequences can be used to achieve different modulations of the scCNT surface properties, thereby producing diverse responses to different analytes. For example, GT-rich sequences (such as (GT)n), AT-rich sequences (such as (AT)n), and any other DNA sequences that can be used to separate single-chiral carbon nanotubes can be employed.

[0059] Regarding “array diversity”: In order to construct an effective sensing array, the at least two sensing units differ in composition, which may be reflected in: (1) the use of scCNTs with different chiralities (e.g., one unit uses (6,5) chiral carbon nanotubes and the other uses (7,5) chiral carbon nanotubes); or (2) the use of DNA sequences with different sequences (e.g., one unit uses (GT)6 to wrap (6,5) carbon nanotubes and the other uses (AT)15 to wrap (6,5) carbon nanotubes); or (3) both. Preferably, the sensing array consists of three or more different sensing units to provide richer sensing dimensions.

[0060] Regarding the preparation of DNA-scCNTs: In the embodiments of this invention, high-purity DNA-scCNT complexes can be isolated and purified from commercially available mixed chiral SWCNT raw materials using various methods well known to those skilled in the art. These methods include, but are not limited to, density gradient ultracentrifugation, gel chromatography, and aqueous two-phase extraction. Specific preparation and purification details will be described in detail in the subsequent detection embodiments.

[0061] Example 1: Identification of biological fluids (serum) by a biomimetic sensing system based on DNA-scCNT array

[0062] This embodiment will explain in detail how to prepare a sensor array containing six different sensing units, and how to use this sensor array and the corresponding biomimetic sensing system to achieve in vitro differentiation of serum samples from various sources.

[0063] 1. Preparation of high-purity DNA-scCNT sensing units

[0064] (1) Raw material stock solution: Take commercially available mixed single-walled carbon nanotube powder ( Sigma-Aldrich DNA was added to deionized water to prepare a carbon nanotube suspension with a concentration of 2 mg / mL. DNA sequences for isolating specific chiral carbon nanotubes were used, as shown in Table 1. DNA (synthesized by Shanghai Sangon Biotech) was dissolved to prepare a solution with a concentration of 10 μg / μL. 1M sodium chloride (NaCl) and phosphate buffer (NaPB) stock solutions were prepared.

[0065] (2) Preparation of crude DNA-SWCNT complex: In a 1.5 mL centrifuge tube, the carbon nanotube suspension and DNA solution were mixed with a volume ratio of V(carbon nanotube suspension):V(DNA solution) of 2:1. 200 μL of NaCl or NaPB was added to the mixture and deionized water was added to bring the total volume to 1 mL, resulting in a final salt concentration of 200 mM. The centrifuge tube was placed in an ice-water bath and sonicated using a probe-type sonicator at a power setting of 4 W for 60 minutes to obtain a uniformly dispersed crude DNA-SWCNT complex solution.

[0066] (3) Purification of single chiral DNA-scCNTs (aqueous two-phase extraction method)

[0067] Prepare an aqueous two-phase system for separating specific chiral DNA sequences by adding a top-phase enrichment polymer, polyethylene glycol (PEG, molecular weight range 4k-12kDa), and a bottom-phase enrichment polymer, dextran, along with specific types of salts (such as sodium phosphate or potassium phosphate). Table 1 details the components of the aqueous two-phase system used to separate six different target chiral DNA sequences.

[0068] Take 140 μL of the aqueous two-phase solution prepared for a specific chirality and add it to 60 μL of the previously prepared crude DNA-scCNT complex solution. Vortex the mixture thoroughly and centrifuge it for 1-5 minutes at room temperature using a benchtop centrifuge at 10,000-14,000 rpm and a rotor radius of 10 cm. After centrifugation, the solution will clearly separate into upper and lower phases. The DNA-scCNTs of the target chirality will be enriched in the upper phase (PEG phase). Carefully aspirate the clear upper phase, which is the high-purity single-chiral DNA-scCNT sensing unit solution. Repeat this process to prepare six different high-purity DNA-scCNT sensing units according to the corresponding DNA sequences and aqueous two-phase systems in Table 1, with coil vectors of (6,4), (6,5), (7,3), (7,5), (8,3), and (9,1), respectively.

[0069] Table 1

[0070]

[0071] 2. Quality Characterization of Sensing Units

[0072] To verify the chiral purity and spectral purity of the prepared DNA-scCNT sensing unit, we used a UV-Vis-NIR spectrophotometer and a NIR fluorescence spectrometer to characterize it.

[0073] Figure 1 Figures a(1)-(6) illustrate the near-infrared emission characteristic peaks of DNA-scCNTs with coil vectors (9,1), (8,3), (7,5), (7,3), (6,5), and (6,4). The chiral DNA-scCNT at (9,1) shows a major E11 emission peak at 928 nm, while a weaker secondary emission peak (Trion peak) is present near 1091 nm. The main emission peak of the chiral DNA-scCNT at (8,3) is located at 966 nm, and the secondary emission peak at 1133 nm is also present. (7,5) Chiral DNA-scCNT exhibits a strong primary emission peak at 1039 nm and a relatively weak secondary emission peak (Trion peak) near 1209 nm; (7,3) Chiral DNA-scCNT exhibits a single strong emission peak at 1014 nm; (6,5) The emission peak of chiral DNA-scCNT is located at 996 nm; (6,4) Chiral DNA-scCNT shows a characteristic emission peak at 888 nm.

[0074] like Figure 1 Figures b(1)-(6) illustrate the UV-Vis-NIR absorption characteristic peaks of DNA-scCNTs with curl vectors (9,1), (8,3), (7,5), (7,3), (6,5), and (6,4), respectively representing the UV-Vis-NIR absorption characteristic peaks of (9,1) at 926 nm (primary peak, E11) and 705 nm (secondary peak, E22); (8,3) at 966 nm (primary peak, E11) and 675 nm (secondary peak, E22); and (7,5) at 675 nm (secondary peak, E22). UV-Vis-NIR absorption characteristic peaks at 1036 nm (first-order peak, E11) and 654 nm (second-order peak, E22); UV-Vis-NIR absorption characteristic peaks at 1013 nm (first-order peak, E11) and 513 nm (second-order peak, E22) in (7,3); UV-Vis-NIR absorption characteristic peaks at 990 nm (first-order peak, E11) and 576 nm (second-order peak, E22) in (6,5); UV-Vis-NIR absorption characteristic peaks at 886 nm (first-order peak, E11) and 590 nm (second-order peak, E22) in (6,4).

[0075] The above spectral characterization results indicate that each DNA-scCNT sensing unit has a high degree of chiral purity and spectral purity. The characteristic absorption peaks and emission peak positions of each chiral carbon nanotube accurately correspond to their theoretical values, with sharp and symmetric peak shapes and no interference from the characteristic peaks of other chiral impurities. This result confirms the effectiveness of the preparation process and lays a material foundation for constructing a high-performance multi-channel sensing array. The unique and stable optical properties of each sensing unit will ensure that the sensing array can generate high-quality and reproducible fluorescence response signals.

[0076] 3. Construction of the sensing array

[0077] Take the six high-purity DNA-scCNT sensing unit solutions ((6,4), (6,5), (7,3), (7,5), (8,3), (9,1)) prepared and characterized in the previous steps. Dilute these six solutions with phosphate buffer (PBS, pH = 7.4) until the absorption intensity (OD value) at their respective first-order absorption peaks (E11) is adjusted to the range of 0.3 - 0.5. Take a 96-well black opaque flat bottom plate. In rows 2 to 7 of each column of this well plate, sequentially add 90 μL of the above six different diluted sensing unit solutions, one for each well. At this time, each column together constitutes a liquid sensing array described in the present invention, which can be used for subsequent in vitro analysis of serum samples.

[0078] 4. Verification of the in vitro discrimination performance of the sensing system

[0079] Prepare the following two groups of serum samples for performance verification at different levels: The first group (multi-category discrimination verification): liver cancer serum (category A), lung cancer patient serum (category B), ovarian cancer patient serum (category C), healthy human serum (sample D); The second group (micro-difference recognition verification): serum of early (<I stage) cancer patients (sample E), healthy human serum (sample F)

[0080] Incubate the above groups of samples with the sensing array for 24 hours (volume ratio 9:1, 25 °C, humidity 50%), and use a near-infrared fluorescence spectrometer to collect 12-dimensional fluorescence response data and transfer it to the data processing and analysis module. Data preprocessing includes: peak calibration, peak position calibration, calculation of relative changes (the relative change calculation is: fluorescence intensity change ΔI / I0, where ΔI = I - I0, I is the intensity after incubation, I0 is the control intensity after the same incubation time of each sensor unit with fetal bovine serum (FBS); and wavelength shift Δλ, where Δλ = λ - λ0, λ is the wavelength after incubation, λ0 is the control wavelength after the same incubation time with fetal bovine serum (FBS)), data normalization and feature construction; the feature construction is to combine the processed multi-channel relative fluorescence change data into a multi-dimensional feature vector.

[0081] To verify the potential of the sensor array to distinguish different samples, we first conducted a detailed statistical analysis of the response data (standardized ΔI / I0 and Δλ) of four groups of samples (samples A, B, C, and D) on each sensor channel. We first determined whether the data conformed to the normality of the distribution. If the normality test was passed, a homogeneity of variance test was performed. If the variances were homogeneous, one-way ANOVA was used for inter-group comparisons. If the data conformed to normality but the variances were not homogeneous, Welch's t test was performed. If the data did not conform to normality, the non-parametric Kruskal-Wallis (KW) test was performed. We then statistically analyzed the statistical significance of individual sensor units among the four categories, with p-values ​​for the four tests set as *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001.

[0082] Figure 2 The diagram illustrates the significant differences in chirality of nanosensors across four categories. The categories showing significant differences provide a basis for intuitively distinguishing between the different types. For example... Figure 2 As shown in Figure 1, multiple sensing units exhibited statistically significant differences in response to different groups of serum samples. For example, the (9,1) chiral sensing unit could significantly distinguish between the "liver cancer / lung cancer" serum sample group and the "liver cancer / ovarian cancer" sample group based on its fluorescence intensity response (ΔI / I0). This intuitively demonstrates that each sensing unit in the array contributes unique and effective information to the final discrimination task.

[0083] Figure 3 The figure shows a heatmap generated after processing ΔI / I0 and Δλ data, which can obtain the "fingerprint" of the overall sample for each category. The "fingerprint" is a multi-dimensional signal response that can reflect the overall chemical characteristics of the sample, generated by the interaction of various biomolecules in the sample with each sensing unit in the array to different degrees.

[0084] However, visual inspection alone cannot accurately distinguish the four types of samples because the relationships between features may contain non-linear patterns, making it impossible to observe their regularity intuitively. Principal component analysis (PCA) confirms this view. Figure 5 The first two principal components in Figure a account for only 49.68%, which means that there is a significant loss of information when performing dimensionality reduction on the features. Figure 5 Figure b further illustrates that there are numerous non-linear patterns among the features, with multiple sets of feature vectors being perpendicular to each other.

[0085] Several model training methods were constructed based on the above dataset, including Support Vector Machine (SVM), Random Forest (RF), and Artificial Neural Network (ANN). The results are as follows: Figure 4 and Figure 5As shown in c, the SVM model performs best, with both F1 score and AUC value at a high level. Further optimization of the algorithm's hyperparameters and evaluation of model complexity and efficiency are needed.

[0086] According to a first embodiment of the present invention, the present invention provides the influence of the number of sensing units on model performance: Figure 5 As shown in Figure d, the model's performance (F1 score) increases systematically with the increase in the number of sensor units. However, the improvement is minimal when the number of sensors increases from 5 to 6. Therefore, the model has a good balance between complexity and efficiency in the current recognition task.

[0087] After further hyperparameter optimization, Figure 5 The figure shows the model F1 scores obtained by the sensing system based on the SVM algorithm in distinguishing various serum samples: liver cancer 0.93, lung cancer 0.80, ovarian cancer 0.86, and healthy 0.93 (out of 1). Figure 5 The figure shows the area under the ROC curve (AUC): liver cancer 0.92, lung cancer 0.94, ovarian cancer 0.87, and healthy samples 0.97 (out of 1); indicating that the sensing system has good multi-class discrimination ability and generalization ability; in independent external validation datasets, the sensing system has excellent performance in identifying liver cancer, lung cancer, ovarian cancer, and healthy samples. Figure 5 g shows that the model F1 scores are 0.93 for liver cancer, 0.80 for lung cancer, 0.86 for ovarian cancer, and 0.93 for healthy individuals; Figure 5 h shows the area under the ROC curve (AUC): liver cancer 0.92, lung cancer 0.94, ovarian cancer 0.87, healthy cells 0.97, with an average precision of [missing value]. Figure 5 The percentage shown in i is 87.5%;

[0088] To further highlight the superiority of the technical solution of this invention, we directly compared its performance with detection schemes that rely on a single traditional biomarker. Figure 6 Figures a and b show that in these serum samples diagnosed with cancer (samples A, B, and C), a considerable number of patients still had tumor marker concentrations below the reference threshold, making it impossible to identify the samples by detecting the level of specific markers. This clearly reveals the inherent limitations and high risk of missed detection in traditional technical solutions that rely on a single marker when distinguishing complex samples. Figure 6 c demonstrates that the sensing system of the present invention can accurately distinguish samples that cannot be identified by conventional methods, and its overall discrimination accuracy is significantly improved compared with the marker scheme.

[0089] To further challenge and verify the performance of the sensing system in this invention, serum samples (sample E) from early-stage cancer patients with a clinical stage less than I and serum samples (sample F) from healthy individuals were subjected to identification tests, aiming to distinguish samples with more subtle differences in chemical composition.

[0090] Except for replacing the samples with a second set, the sensor array, sample incubation conditions, signal acquisition methods, and data processing and machine learning model (SVM) construction and evaluation procedures are basically the same as those described above. The purpose of this test is to construct a model that can distinguish between samples E and F, which have extremely high similarity, based on the acquired 12-dimensional fluorescence response data, and to evaluate its performance using rigorous 5-fold nested cross-validation.

[0091] Figure 7 The illustration shows the performance of the detection method provided in the first embodiment of the present invention in serum sample identification: Figure 7 Figure a shows a PCA plot, revealing minimal overlap in the distributions of early-stage cancer and healthy samples, indicating a potential non-linear boundary; as shown in the attached figure. Figure 7 As shown in b and 7c, the classification model achieved a sensitivity of 0.84 and an F1 score of 0.83, with a relatively balanced precision and recall, and low risk of false positives and false negatives. Figure 8 Figure d shows the sensitivity of the DNA-scCNT sensing system for early cancer sample identification, which is 0.84, a significant improvement compared to the sensitivity of 0.29 for the traditional marker scheme.

[0092] Example 2: Detection of neurotransmitter molecules based on DNA-scCNT array

[0093] 1. Fabrication of neurotransmitter sensing arrays

[0094] The preparation of the DNA-scCNT sensing unit in this embodiment follows the same basic process (including ultrasonic dispersion and aqueous two-phase extraction) as described in Example 1. Unlike Example 1, this embodiment uses a new set of DNA sequences to modify the surface of carbon nanotubes, constructing a ternary (containing three sensing units) sensing array specifically optimized for distinguishing neurotransmitter molecules. Due to the differences in DNA sequences, the fluorescence response characteristics of carbon nanotubes with the same chirality differ significantly from those in Example 1, demonstrating the important regulatory role of DNA sequences on sensor performance. Specific extraction schemes for the three chirities are shown in Table 2.

[0095] Table 2

[0096]

[0097] 2. Quality Characterization of Sensing Units

[0098] To verify the chiral purity of the newly prepared sensing unit in this embodiment, the purity of the obtained DNA-scCNT was characterized. Figure 9 The fluorescence emission spectra of the three DNA-scCNTs in Example 2 are shown. Spectral analysis results indicate that all three DNA-scCNTs exhibit a single E11 emission peak. The (6,5) chiral DNA-scCNT modified with TTATTAATTATT shows a single, sharp E11 emission peak at 996 nm, with a symmetrical peak shape and no interference from impurity peaks. The (8,3) chiral DNA-scCNT modified with GGGTTTTGGTTTTGG exhibits a bimodal characteristic, with the main emission peak (E11) located at 965 nm, and a weak secondary emission peak (Trion peak) near 1135 nm, which is a typical spectral characteristic of this chiral carbon nanotube. The (9,1) chiral DNA-scCNT modified with CCGCGGCGGCGCGG exhibits a bimodal emission characteristic, with the main emission peak located at 928 nm and a secondary emission peak (Trion peak) at 1091 nm, which is a typical spectral characteristic of this chiral carbon nanotube.

[0099] The above spectroscopic characterization results confirmed that all three DNA-scCNTs possess high chiral purity and are free from interference from impurity peaks of other chiral carbon nanotubes. The characteristic emission peaks of each sensing unit are clearly distinguishable, providing a reliable baseline reference for subsequent fluorescence response changes after interaction with the target molecule, and ensuring that the sensing array can generate high-quality, easily resolvable multichannel fluorescence signals.

[0100] 3. In vitro differentiation and verification of multiple neurotransmitter molecules

[0101] This experiment selected three neurotransmitter molecules with highly similar chemical structures as test samples: serotonin (5-HT), dopamine (DA), and L-DA (L-DA). These three molecules were prepared into concentrations ranging from 10... -9 M to 10 -2 A series of solutions of M. Different concentrations of the sample solution to be tested were added to the ternary sensing array constructed in the preceding steps, and fluorescence response signals were acquired.

[0102] As attached Figure 9 As shown in a to 9i, the three sensing units in the array exhibit unique, concentration-dependent differential responses to the three neurotransmitter molecules. For the TTATTAATTATT-(6,5) sensing unit ( Figure 9 a,d,g), 5-HT causes fluorescence quenching, while DA and L-DA cause fluorescence enhancement, providing a basis for distinguishing 5-HT; for the GGGTTTTGGTTTTGG-(8,3) sensing unit ( Figure 9b, e, h), all three molecules induced fluorescence enhancement, but the magnitude of the enhancement was in the following order: DA > L-DA > 5-HT; for the CCGCGGCGGCGG-(9,1) sensing unit ( Figure 10 The order of fluorescence enhancement amplitude becomes: L-DA > DA > 5-HT. These results clearly demonstrate that although the sensing units lack traditional specificity, they produce a unique and reproducible response pattern when interacting with different analytes. Combining the response signals of these three sensing units constructs a unique multidimensional "fingerprint" for each neurotransmitter molecule, enabling the effective identification of highly similar molecules using pattern recognition algorithms, whereas traditional detection methods struggle to effectively identify highly similar molecules.

[0103] Figure 10 j shows a two-dimensional heatmap generated after normalizing the relative fluorescence intensity changes (I-I0) / I0 and wavelength shifts λ-λ0 caused by different neurotransmitter molecules in the sensing array through a computer data processing module. This heatmap clearly shows the unique response pattern of each molecule in the multi-channel sensing system, forming a highly specific molecular fingerprint spectrum, which provides effective information for the accurate identification of molecules with similar structures.

[0104] To accurately decode and identify the aforementioned "fingerprint map," we employed the Support Vector Machine (SVM) algorithm to perform pattern recognition modeling on the collected data. To ensure the rigor and reliability of model performance evaluation, this embodiment also adopted a nested cross-validation strategy: the outer layer uses Leave-One-Out Cross-Validation (LOOCV) to evaluate the model's generalization performance, while the inner layer uses Bayesian Optimization (BayesOpt) to automatically tune SVM hyperparameters, including intelligent search and optimization of key parameters such as regularization parameter C, kernel function type (linear kernel, radial basis function kernel, polynomial kernel, sigmoid kernel), kernel function parameter γ, polynomial kernel degree, and class weights.

[0105] Figure 11 A shows the confusion matrix results of the SVM classification model after Bayesian optimization. The matrix clearly demonstrates the recognition performance of the three neurotransmitter molecules: 7 test samples each for 5-HT, DA, and L-DA, covering a concentration of 10... -9 M to 5.5×10 -4 M, all samples achieved 100% correct identification. The diagonal elements of the confusion matrix are all 7, and the off-diagonal elements are all 0, indicating that the model optimized by intelligent hyperparameters has the ability to accurately distinguish between the three structurally similar neurotransmitter molecules without misclassification.Figure 12 b presents the overall performance evaluation results of the model. The overall accuracy reached 1.0, and the average F1 score also reached 1.0.

[0106] The results of this embodiment fully demonstrate that the artificial sensing system and its in vitro analysis method of the present invention, by constructing an array composed of multiple DNA-scCNT sensing units with differentiated responses, can generate unique "fingerprints" for molecules with highly similar chemical structures (such as 5-HT, DA, L-DA, which differ only in their side group structures). Combined with advanced machine learning algorithms and rigorous validation strategies, the present invention successfully achieves accurate identification of molecules that are difficult to distinguish simultaneously using traditional methods. This decisively proves that the present invention, as a high-sensitivity, high-resolution, universal analytical platform, has significant technical advantages and application value in fields such as the precise detection of trace similar molecules.

[0107] Example 3: Detection and identification of 18 amino acids based on DNA-scCNT sensor array

[0108] This embodiment aims to further demonstrate the superior high-resolution capability and wide applicability of the sensing system described in this invention in distinguishing a range of structurally highly similar molecules. This embodiment constructs a DNA-scCNT sensing array containing four sensing units and achieves high-throughput rapid distinction of 18 standard amino acids without derivatization or chromatographic separation. Amino acids, as the basic building blocks of proteins, exhibit only subtle differences in their side-chain structures, making high-throughput and rapid distinction difficult to achieve using traditional detection methods without derivatization or chromatographic separation.

[0109] 1. Fabrication of amino acid sensing array

[0110] The preparation of the DNA-scCNT sensing unit in this embodiment follows the same basic procedure (e.g., aqueous two-phase extraction) as described in Example 1. In this embodiment, we constructed a sensing array containing four different sensing units. These four sensing units were obtained by using different DNA sequences or different carbon nanotube chirities, and the key parameters selected during their preparation are detailed in Table 3. For ease of subsequent description, we will refer to the (6,5) chiral sensing unit modified with the GGCCCGCGCCCG sequence as S7-(6,5).

[0111] Table 3

[0112]

[0113] 2. Quality Characterization of Sensing Units

[0114] The purity of the four DNA-scCNT sensing units was characterized using near-infrared fluorescence spectroscopy.Figure 13 Spectroscopic analysis results showed that: (7,5) chiral DNA-scCNT modified with CCGCCCCGCCCGCCC exhibited a bimodal characteristic, with its main emission peak (E11) located at 1039 nm and a weak secondary emission peak (Trion peak) near 1210 nm; (8,3) chiral DNA-scCNT modified with GGGTTTTGGTTTTGG also showed a bimodal characteristic, with the main emission peak located at 968 nm and the secondary emission peak located near 1140 nm; (6,5) chiral DNA-scCNT modified with TTATTAATTATT showed a single, sharp E11 emission peak at 996 nm, with a symmetrical peak shape and no obvious interference from impurity peaks; the main emission peak of (6,5) chiral DNA-scCNT modified with GGCCCGCGCCCG was located at 999 nm, also with a sharp and symmetrical peak shape.

[0115] The above spectral results confirm that each sensing unit prepared in this embodiment has clear and unique spectral characteristics and high chiral purity, meeting the requirements for constructing a high signal-to-noise ratio sensing array.

[0116] 3. Construction of sensor array and sample detection

[0117] Each of the above sensing units was diluted to achieve an intensity of 0.21–0.23 OD for the first-order absorption peak (E11) in the UV-Vis-NIR spectrum. A suitable amount of the working solution was then placed in a 96-well microplate to construct a 4-channel sensing array for subsequent amino acid recognition.

[0118] Weigh out 18 commercially available high-purity amino acids, including glycine (Gly), alanine (L-Alanine, L-Ala), arginine (L-Arginine, L-Arg), aspartic acid (L-Aspartic acid, L-Asp), cysteine ​​(L-Cysteine, L-Cys), and glutamic acid (L-Glutamic acid). The following amino acids are listed: L-Glu, L-Histidine, L-His, L-Isoleucine, L-Leucine, L-Lys, L-Methionine, L-Phenylalanine, L-Proline, L-Serine, L-Threonine, L-Trp, L-Tryptophan, L-Tyrosine, and L-Valine. To investigate the response capability of the sensor array at different concentrations and the robustness of the model, each amino acid was dissolved in phosphate buffer solution (PBS, 10 mM, pH 7.4) to prepare a series of test solutions with concentrations of 200 μM, 400 μM, 600 μM, 800 μM, 1000 μM and 1200 μM.

[0119] The 18 amino acids were added in series of concentrations to the microwells of a pre-prepared 4-channel sensor array, with a working solution to test solution volume ratio of 9:1 for each sensor unit, resulting in final concentrations of 20 μM, 40 μM, 60 μM, 80 μM, 100 μM, and 120 μM for each amino acid. The microplate was incubated at 25°C and 50% humidity for 30 minutes. After incubation, the fluorescence spectra of each microwell were sequentially acquired using a near-infrared fluorescence spectrometer, and the fluorescence curves were recorded. The results were then fed into a pre-defined algorithm to extract the intensity and peak wavelength information.

[0120] 4. In vitro differentiation verification of amino acids

[0121] The fluorescence signal of each sensing unit after incubation with pure PBS buffer solution (without amino acids) for the same time was used as the reference signal, with its fluorescence intensity denoted as I0 and peak wavelength denoted as λ0. The relative change rate of fluorescence intensity (I-I0) / I0 and wavelength shift λ-λ0 of each sensing unit after the addition of different amino acid solutions were calculated. For each amino acid sample, the (I-I0) / I0 and λ-λ0 values ​​generated on the four sensing units (a total of 8 data points) were combined into an 8-dimensional feature vector, which constitutes the response fingerprint spectrum of the amino acid on this sensing system.

[0122] Figure 14 This is a standardized heatmap (fingerprint) of the 8-dimensional response signals generated by 18 amino acids on a 4-channel sensing array. As shown in the figure, each amino acid induces a unique and reproducible response pattern in the sensing array. Notably, even isomers like leucine (L-Leu) and isoleucine (L-Ile), which have only minor differences in chemical structure, exhibit significant fingerprint patterns that can be identified by algorithms. For example, in the relative changes in fluorescence intensity acquired from a sensing unit composed of (6,5) chiral DNA-scCNTs, L-Leu causes fluorescence enhancement, while L-Ile causes significant fluorescence quenching. This result directly demonstrates the extremely high chemical resolution of this sensing system.

[0123] To select the optimal pattern recognition algorithm, this embodiment employs various machine learning algorithms (including but not limited to Support Vector Machine (SVM), Random Forest (RF), Logistic Regression (LR), Multilayer Perceptron (MLP), Extreme Gradient Boosting Tree (XGB), and Decision Tree (DT)) to train and evaluate the above dataset. Figure 14 This paper presents a performance comparison of different algorithms in the task of identifying 18 amino acids. The results show that the Support Vector Machine (SVM) algorithm performs best in both the F1 score and accuracy, with scores of 0.93 and 0.94 respectively, and is therefore selected as the algorithm for the final model construction.

[0124] Figure 14 The final selected SVM model is shown in the comprehensive performance evaluation after 3-fold nested cross-validation. ​ 'a' represents the confusion matrix of the model. In the matrix, out of a total of 108 samples, 106 samples are correctly classified into their corresponding categories (i.e., distributed along the main diagonal), with only 2 misclassified samples. ​ b represents the quantitative performance index of the model. The results show that the overall accuracy for classifying 18 amino acids reached 0.982, and the F1 score reached 0.981. It should be emphasized that this high accuracy was achieved under conditions where the concentration of the amino acids to be tested varied over a wide range from 20 μM to 120 μM.

[0125] This embodiment decisively demonstrates that the biomimetic sensing system and its in vitro analysis method of the present invention can effectively solve the problem of rapidly and synchronously distinguishing a series of structurally highly similar molecules (such as 18 amino acids), a challenge that traditional technologies struggle to address. Its high classification accuracy and excellent robustness to concentration changes showcase the significant technical advantages and broad application prospects of this invention as a high-resolution, versatile analytical platform.

[0126] Example 4: Sensor Array Chip Based on Solid-State Substrate and Its Application

[0127] This embodiment describes how to integrate the DNA-scCNT sensing array of the present invention onto a solid-state substrate to construct a chip-based, portable sensing device.

[0128] 1. Fabrication of solid-state sensor arrays

[0129] Using the method described in Example 1, several (e.g., six) different high-purity DNA-scCNT sensing unit solutions were prepared. In this example, polyethylene terephthalate (PET) was selected as the flexible substrate. The substrate surface was treated with an oxygen plasma cleaner to enhance the adhesion of subsequent sensing materials.

[0130] 2. Construction of the sensor array

[0131] First, a master mold containing a network of parallel microchannels is fabricated on a silicon wafer using standard photolithography. Then, liquid polydimethylsiloxane (PDMS) prepolymer is poured onto the mold, cured, and carefully peeled off to obtain a PDMS chip with a microchannel structure. The channel surfaces of this PDMS chip are irreversibly bonded to a pre-treated flexible PET substrate (e.g., directly bonded after oxygen plasma treatment), forming a series of closed, independent microfluidic channels on the PET substrate surface. Six different DNA-scCNT sensing unit solutions are slowly injected into different microfluidic channels using a syringe pump. The spotted PET substrate is incubated in a temperature and humidity controlled incubator (e.g., 25°C, 70% humidity) for 12 hours, allowing the droplets to slowly evaporate. The DNA-scCNT complex is firmly fixed to the substrate surface through physical adsorption and hydrogen bonding. All channels are rinsed with buffer to remove unbound sensing material. Finally, the upper PDMS chip can be selectively peeled off (if reversibly sealed) or retained as a channel for subsequent sample introduction. This creates an array of multiple parallel sensing strips on the PET substrate.

[0132] 3. Use of solid-state sensor chips

[0133] During detection, a drop of the sample solution to be tested (e.g., the serum diluent in Example 1) can be directly added to the array area of ​​the chip, completely covering all sensing points. After incubation for a certain period of time, the flexible chip is placed directly on the sample stage of an optical readout device with two-dimensional scanning capabilities (e.g., an array scanning fluorescence imager), and the fluorescence signal changes of each sensing point are collected sequentially or simultaneously. Subsequent data processing and pattern recognition procedures are the same as those described in Example 1.

[0134] This embodiment details a feasible method for chip-based implementation of the sensor array of the present invention, demonstrating that the present invention can be fabricated into portable and easy-to-use solid-state sensing devices (including flexible devices). This chip-based implementation greatly expands the application scenarios of the present invention, for example, it can be used in wearable sensing devices or point-of-care testing (POCT) devices.

[0135] Example 5: A biomimetic sensing system based on DNA-scCNT electrochemical detection

[0136] This embodiment describes how the sensing system of the present invention operates using an electrochemical detection method.

[0137] 1. Fabrication of electrochemical sensing arrays

[0138] A commercially available screen-printed carbon electrode (SPCE) array was selected as the substrate. Multiple independent working electrodes, a shared reference electrode (e.g., Ag / AgCl), and a shared counter electrode (e.g., a carbon electrode) were integrated onto this substrate. Using the method described in Example 1, various high-purity DNA-scCNT sensing unit solutions were prepared. Different DNA-scCNT sensing unit solutions (e.g., 5 μL per drop) were carefully dropped onto the surfaces of different working electrodes using a drop-coating method. After air drying at room temperature, the DNA-scCNT complex formed a sensing film through physical adsorption, covering the working electrode surface. Thus, an electrochemical electrode array integrating the sensing array of this invention was prepared.

[0139] 2. Detection of electrochemical signals

[0140] The electrochemical detection in this embodiment is based on the excellent conductivity of carbon nanotubes and their sensitivity to changes in the surrounding chemical environment. A DNA-scCNT complex is applied as a semiconductor film to the electrode surface. When analytes in the sample (e.g., neurotransmitters, amino acids, or proteins in serum) interact with this DNA-scCNT film (e.g., through electrostatic, hydrophobic, or π-π stacking interactions), the dielectric environment of the DNA-scCNT surface is altered. For example, certain analyte molecules with oxidizing or reducing properties can directly inject or extract electrons from the carbon nanotubes, thereby directly changing their conductivity (i.e., altering their resistance / conductance). The adsorption of analytes changes the local electric field around the carbon nanotubes, similar to the gate voltage change in a field-effect transistor (FET), thus modulating the current flowing through the carbon nanotube network. These minute changes in the conductivity of carbon nanotubes caused by molecular interactions can be precisely measured using an external electrochemical workstation.

[0141] The prepared electrochemical electrode array is connected to a multichannel electrochemical workstation (or potentiostat array), and the change in electrical signal can be tested using DC amperometry or electrochemical impedance spectroscopy (EIS). For DC amperometry, a constant, small DC bias voltage (e.g., 50 mV) is applied between each working electrode and the counter electrode. Before and after adding the sample to be tested, the steady-state current value (I) flowing through each working electrode is measured. The change in current (ΔI) or the change in resistance / conductivity calculated according to Ohm's law constitutes the response signal in this embodiment. For electrochemical impedance spectroscopy, a sample solution containing a small amount of electrolyte is first added to the surface of the electrode array. Then, electrochemical impedance spectroscopy (EIS) is used to measure the impedance spectrum of each sensing unit. By analyzing the impedance spectrum (e.g., directly comparing the impedance value change at a specific frequency), the response signal of each sensing unit can be obtained.

[0142] 3. Data Processing and Analysis

[0143] The response signal values ​​(e.g., ΔI / I0 or ΔR / R0) measured by each sensing unit before and after the addition of the sample are used as the signal of that channel. The response signals of multiple channels are combined into a multidimensional "electrochemical fingerprint". The subsequent data processing and pattern recognition process is the same as that described in Example 1.

[0144] This embodiment details how the present invention is combined with electrochemical detection technology to construct a biomimetic sensing system based on electrical signal response. This demonstrates that the core idea of ​​the present invention (i.e., generating fingerprint patterns using an array composed of multiple DNA-scCNTs) has broad platform applicability, not limited to optical detection, but also applicable to the field of electrochemical sensing, thus further expanding its application scope.

[0145] In summary, this invention provides a sensing system based on single-chiral carbon nanotubes and its in vitro analysis method. By configuring each sensing unit in the sensing array with high-purity DNA-scCNTs, the problem of low signal quality of mixed chiral carbon nanotubes used in the prior art is solved, ensuring that the acquired raw sensing signal has extremely high purity, signal-to-noise ratio, and resolution.

[0146] Based on this high-quality signal foundation, this invention further employs a core working mechanism of "sensor array + pattern recognition." It does not rely on any pre-set, one-to-one specific identification for particular molecules, but rather utilizes the multi-dimensional, unique response "fingerprint" generated after the entire array interacts with the sample to perform holistic identification. This mechanism frees this invention from the constraints of poor versatility and high cost of traditional detection technologies. Examples have fully demonstrated that, thanks to the effective combination of high-quality "fingerprint" and advanced machine learning algorithms, this system can not only accurately distinguish between various complex biological fluids but also reliably identify molecules with highly similar chemical structures or subtle chemical differences between samples.

[0147] Therefore, this invention provides a new paradigm for in vitro analysis that is simpler, lower in cost, and highly versatile. It typically eliminates the need for complex pretreatment steps such as separation and purification of samples, allowing for direct analysis and significantly shortening the detection process. It is important to understand that the technical solution of this invention has broad applicability; its application is not limited to the detection of biological fluids and small molecules, but is also applicable to other samples requiring high-resolution discrimination. It can analyze both untreated complex samples and differentiate processed samples. The core principle of this invention can be applied to detection based on optical signals, and can also be extended to detection based on electrochemical signals; its physical implementation can be in the form of a liquid array or fabricated as a solid-state chip, showing broad application prospects in numerous fields.

[0148] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. The various specific embodiments described in this specification are intended to elaborate on the technical solutions of the present invention in detail to demonstrate their feasibility and beneficial effects, and are not intended to constitute any form of limitation on the present invention. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A single chirality carbon nanotube sensing array, characterized in that, Comprising: at least two sensing units, which are DNA-single-chiral-carbon nanotube complexes DNA-scCNT, and the DNA-scCNT of the at least two sensing units differ in at least one of the DNA sequence or the chiral of the carbon nanotube.

2. The sensor array of claim 1, wherein, The at least two sensing units in the sensing array are respectively contained in different containers of a multi-container device in solution form to constitute a liquid array, or are fixed on different regions of the same substrate to constitute a solid array.

3. A biomimetic sensing system based on single-chiral carbon nanotube sensing array, characterized in that, Further comprising the following parts: a) the sensing array of claim 1; b) a detection device for detecting the response signals generated by each sensing unit in the sensing array when the sensing array is in contact with a sample to be tested, so as to constitute a multi-channel signal group; c) a data processing and analysis module for receiving and processing the multi-channel signal group and analyzing it through a pattern recognition algorithm to obtain the classification or state information of the sample to be tested.

4. The biomimetic sensing system of claim 3, wherein, The response signal comprises at least one of resistance change, conductance change, fluorescence intensity change or fluorescence wavelength shift; and the detection device comprises an electrochemical detection device or an optical detection device.

5. The biomimetic sensing system of claim 3, wherein, The pattern recognition algorithm is a machine learning algorithm, which comprises at least one of principal component analysis (PCA), support vector machine (SVM), artificial neural network (ANN), random forest (RF), decision tree (DT) or gradient boosting tree (XGBoost).

6. An in vitro detection method based on the sensor system of claim 3, characterized in that, Comprising the following steps: a) providing the sensing system of any one of claims 3 to 5, wherein a pre-trained pattern recognition model has been configured in the data processing and analysis module; b) contacting the sensing array of the system with a sample to be tested in vitro; c) using the detection device to detect and collect the response signals generated by each sensing unit in the sensing array to obtain a multi-channel signal group; d) using the data processing and analysis module to analyze the multi-channel signal group through the pre-trained pattern recognition model to obtain the classification or state information of the sample to be tested.

7. The detection method according to claim 6, characterized in that, A non-specific recognition strategy is used, which can realize the analysis of the sample to be tested without using pre-fixed specific recognition elements for specific target molecules; the sample to be tested is a complex biological fluid without component separation, or a solution of structurally similar biological molecules.

8. A microfluidic chip, characterized by, The flexible substrate is integrated with the sensing array of claim 1 or 2.

9. A flexible sensing device comprising a flexible substrate, and the flexible substrate is integrated with the sensing array of claim 1 or 2.

10. An electrode array, characterized by The electrode surface is modified or integrated with the sensing array of claim 1 or 2.

Images