Orthogonal nanopores cross-validation for multiplex single-molecule profiling

Abstract

Multiplex biomolecule profiling provides comprehensive molecular insights essential for precision health diagnosis. Yet, achieving simultaneous analysis of diverse biomolecule types, such as DNAs, RNAs, peptides, oligosaccharides, and metabolites, remains challenging due to their fundamentally diversity including size, charge, structure, and chemical composition. We present a nanopore multiplex sensor (NMS) based on a bioelectronic microchip that enables concurrent profiling of multiple biomolecules in a single experiment, which was previously unattainable. Cross-validation using multiple types of orthogonal nanopores with complementary recognition abilities offered the simultaneous detection and relative quantification of five major types of biomolecules relevant to non-small-cell lung cancer. Our work establishes a robust and generalizable nanopore platform for rapid and parallel biomarker sensing, providing an important step toward addressing the vital challenge of single-molecule multiomics.

---

Introduction

Precision medicine and understanding of complex biological systems require systematic investigation across a diverse set of biomolecules.^1,2 The disease progression network is characterized by broad changes in biomolecular composition and abundance at the single-molecule level.^3,4 To capture such complexity, multiplex technologies have been developed to integrate different layers of biological information because reliance on a single type of biomolecule often yields an incomplete or even misleading view of biological states.^5,6 Single-molecule multiplex biomolecule sensing reveals molecular heterogeneity and temporal molecular dynamics through direct and complementary analysis of individual biomolecules. Conventional multiplex techniques, such as mass spectrometry, sequencing, and fluorescence, can detect more than one biomolecular type,^7–9 yet face three main critical bottlenecks. First, simultaneous single-molecule sensing of diverse biomolecules remains challenging. Nucleic acids, proteins, and glycans differ greatly in physicochemical properties and detection efficiency, making it difficult to integrate preparation, separation, and detection of a sample within a single analytical workflow. As a result, different biomolecular types are often measured separately, which limits comprehensive and simultaneous profiling. Moreover, traditional techniques generally offer averaged information, thereby obscuring molecular dynamics behaviors. Second, complex labeling processes are required. Isotopic, chemical or fluorescent labeling is often required for accurate quantification and multiplexing. However, these labeling and probe-based strategies introduce additional cost and operational complexity, potentially increasing false-positive bias and perturbation of native biomolecular states. Additionally, multiplex capacity is often influenced by factors, such as signal overlap, label compatibility, and the limited number of distinguishable labels, presenting challenges for further scaling the simultaneous measurement of diverse molecules. The third bottleneck is a lack of analytical efficiency. Existing platforms rely on multi-step, labor-intensive workflows with limited parallel capacity, usually requiring hours to days from sample preparation to data analysis. Further improvements in workflow integration and analytical throughput are desirable for rapid, high-throughput multiplex biomolecule profiling at single-molecule level. Therefore, these challenges highlight the urgent need for developing a generalized multiplex platform that can sense diverse biomarkers in a rapid, label-free, and scalable manner at the single-molecule level.

Nanopore is a label-free single-molecule sensing technique based on electrochemistry. It can capture the specific and transient ionic current signatures of individual biomolecules with high sensitivity and resolution.^10,11 To date, biological nanopore offers high sensitivity and selectivity governed by its intrinsic physical and chemical properties of the sensing interface, such as pore size, geometry, and surface charge.^12–14 These characteristics enable the individual sensing of diverse biomolecules, including DNAs, RNAs, peptides, oligosaccharides, and small-molecule metabolites.^15–20 Such advances establish the nanopore as a promising solution for label-free multiplex single-molecule sensing with temporal dynamics. However, each nanopore exhibits optimal sensing performance within a specific range of molecular dimensions and properties. Hence, detecting a broad range of biomolecules could benefit from the integration of multiple types of orthogonal nanopores with complementary sensing capabilities for cross-validation. Moreover, further advances in instrument integration and compatibility would facilitate wider practical development. Current nanopore setups generally employ external electrodes coupled to amplifier, potentially introducing additional noise.^21,22 Further, the relatively large volume per channel may raise sample consumption and reduce operational efficiency. Several reported methods using microchips have incorporated multi-channel recording capabilities,^23,24 while typically relying on a single type of nanopore for each biomolecule type sensing.

To achieve single-molecule multiple biomolecules sensing, herein, we developed a nanopore multiplex sensor (NMS) (Fig. 1). The NMS integrated diverse orthogonal nanopores into a multi-chamber microchip with built-in film electrodes and an ultralow-current amplifier array, enabling label-free, high signal-to-noise ratio (SNR), high-throughput and high-sensitivity sensing. The parallel recording of ionic current captured distinct signals from each sample across multiple nanopore types, enabling the sensing of a wide range of biomolecules. Adopting machine-learning algorithms and linear fitting models, this cross-validation approach enabled robust qualitative identification and relative quantification of multiple biomolecules spanning heterogeneous sizes, weights and charges, such as circulating tumor DNA (ctDNA) segments, microRNA (miRNA) segments, tumor-associated protein fragments and their glycosylation, and small-molecule metabolites. Implementation of the parallel cross-validation strategy in NMS establishes an important step toward rapid multiplex sensing.

Fig. 1 Single-molecule nanopore multiplex sensor (schematic). The bioelectronic nanopore microchip features a three-layer design consisting of a substrate, electrode and bilayer-supporting layer, which is directly connected with an ultra-low-current amplifier array via Au wires patterned on the microchip for parallel measurements. Spatially separated sensing chambers integrate Ag/AgCl electrode-pair arrays and microwell arrays, with each chamber accommodating a distinct type of nanopore to enable simultaneous multiplex single-molecule sensing. Mixture samples, such as ctDNA segments, miRNA segments, peptides, oligosaccharides, and small-molecule metabolites, are introduced into a single microchip to generate parallel ionic current signals. These signals are processed through machine-learning classification combined with multi-nanopore cross-validated linear fitting models, enabling rapid identification and relative quantification of multiplex biomolecules within minutes.

Results and discussion

Nanopore multiplex sensor design

The bioelectronic microchip adopted a three-layer architecture according to our previous study.²⁵ It comprised, from bottom to up, a smooth SiO₂ substrate with a low dielectric constant, patterned Ag/AgCl electrode arrays and Au wires, and a hydrophobic supporting layer for a stable membrane array (see the Materials and Methods section in SI). Specifically, to accommodate multiple types of nanopores on a single run, the low-noise bioelectronic microchip featured six spatially partitioned chambers, each functionalized with a specific nanopore type for simultaneous sensing of various biomolecules. Each chamber integrated four independent Ag/AgCl_trans microelectrodes and one Ag/AgCl_cis electrode that enhanced experimental throughput and reproducibility (Fig. S1).

Integrated electrode pairs were prepared by electron beam evaporation (EBE) at the bottom of each chamber, which eliminated the need for external hanging electrodes and thereby improved integration. Subsequently, a microwell array was patterned by a photolithographic SU-8 coating serving as the supporting layer for bilayer formation (Fig. S2). Ag/AgCl_cis as the ground electrode was further electroplated via chronoamperometry on the EBE layer to increase thickness and robustness (Fig. S3). After Ag deposition and chlorination, the spacing between the electrode and supporting layer decreased by ∼3 µm, corresponding to the formation of the Ag/AgCl layer (Fig. 2a). The open-circuit potential (OCP) remained stable during measurements, confirming the good stability and reusability of the integrated electrodes under our experimental conditions (Fig. 2b). Minor potential offsets could be corrected via software-based voltage compensation before single-channel recording. The microchip was directly connected with a homemade multichannel ultra-low-current amplifier, where the shortened wiring effectively reduced noise, enabling simultaneous and independently controlled current recording. As shown in Fig. 2c, the integrated platform exhibited low baseline current noise, with a standard deviation (STD) of 0.5 ± 0.1 pA after nanopore insertion, substantially lower than that of conventional setups (∼2.0 pA at 5 kHz low pass filter (LPF)).²⁶ This high SNR enabled clear identification of ionic current changes as low as 1.5 pA (with the threshold of three times the baseline STD). Benefiting from its compact chamber volume of 100–200 µL, the platform reduced sample consumption to the nanogram level, further extending its applicability to low-volume and low-abundance biological samples. Therefore, this microchip configuration enabled high-SNR, reliable and parallel multiplex biomolecule detection.

Fig. 2 Stability characterization of the nanopore multiplex sensor. (a) The depths between the supporting layer and the electrode layer. The inset scatter plots represent profilometer traces from a single measurement, showing the height profile from the SU-8 surface down to the Ag/AgCl_trans (EBE) of 14.7 ± 0.1 µm or Ag/AgCl_cis (electroplating) of 11.7 ± 0.2 µm, and back to the SU-8 surface. Error bars were calculated from three independent measurements. (b) OCP of Ag/AgCl electrodes on a single microchip. Blue: OCP between the integrated Ag/AgCl_cis electrode and a commercial Ag/AgCl electrode as a reference electrode (RE). Green: OCP measured after re-oxidation of the Ag/AgCl_cis electrode following one round of nanopore experiment vs. RE. Pink: OCP of the integrated Ag/AgCl_trans microelectrode vs. Ag/AgCl_cis electrode in the one chamber. (c) STD values of baseline current. One random channel in each chamber on the one microchip was selected for current noise testing. Blue: the noise of unloaded channels with only microchip connected to the amplifier array of 0.3 ± 0.1 pA. Pink: the noise level of open-pore current (I₀) with WT AeL loaded with 0.5 ± 0.1 pA at +100 mV. The inserted scatter plots showed typical baseline current under unload and nanopore-inserted membrane conditions. Minor noise variations among chambers are likely attributable to slight variations introduced during channel fabrication. (d) Statistical analyses of Poly(dA)₄ detection with WT AeL across six channels in six distinct chambers on one microchip. Green: fitted distributions of duration time (t) of 4.6 ± 0.7 ms. Pink: fitted distributions of residual current depth (I/I₀) of 0.45 ± 0.01. Here, I represents the blockade residual current. Data were recorded with a sampling rate of 100 kHz and filtering at 5 kHz.

To assess the performance of the platform, Poly(dA)₄ (5′-AAAA-3′) was detected as a model analyte with the wild-type (WT) Aerolysin (AeL) nanopore across all chambers, owing to its well-established reproducibility and stability for single-molecule sensing.^26,27 Here, the recorded current blockades and event durations remained comparable across channels and microchips, agreeing with the single-channel platform results (Fig. 2d and S4–S6). Furthermore, Poly(dA)₄ sensing with WT AeL in one chamber could be served as an internal reference for microchip calibration, facilitating robust normalization and accurate quantification.

Orthogonal nanopores for complementary detection

Each type of nanopore is typically suited to a specific range of molecular dimensions and chemical properties. Hence, a single nanopore can barely detect broad types of biomolecule efficiently. We introduced a series of orthogonal protein nanopores with complementary sensing capabilities. The engineered AeL and outer membrane protein F (OmpF) nanopores were selected owing to their diverse sensitivity and selectivity for small biomolecules.²⁸ Specifically, the K238Q (KQ) AeL nanopore,²⁹ with strong electrostatic interactions, was employed for detecting charged single-stranded DNAs (ssDNAs) and short RNAs; the T232K/K238Q (TK/KQ) mutant AeL,³⁰ with strengthened pore-analyte interactions, was used to improve the SNR for peptides and small-molecule metabolites; the WT OmpF nanopore,³¹ with narrow constriction and stereoselectivity, was used for oligosaccharides detection. All nanopores were prepared in our laboratory according to previously reported protocols (Fig. S7). Therefore, integrating these nanopores enabled complementary molecular recognition and improved the multiplex sensing capability of the platform.

Using non-small-cell lung cancer (NSCLC)-related biomolecules as a representative model,^32,33 we included: the oligonucleotide from WT epidermal growth factor receptor (EGFR) and its mutant T790M ctDNA segments to reflect drug resistance;^34,35 miR-21 and miR-92a segments related to early diagnosis and oncogenicity;^36–38 peptide fragments of programmed cell death ligand 1 (PD-L1) and its phosphorylation at Tyr (p-PD-L1);^39–41 as well as the N-linked glycosylation polylactosamine (DiLacNAc) from PD-L1 to indicate immune regulation and evasion;^42–45 and flavin adenine dinucleotide (FAD) representing metabolic dysfunction.^46,47 The sequences of the biomolecules are listed in Table S1. Simultaneous detection of EGFR, miR-21, PD-L1 and its glycosylation, as well as FAD in body fluids, enabled complementary characterization of tumor progression, immune regulation, and metabolic dysfunction, thereby highlighting the potential of multiplex nanopore sensing for lung cancer profiling.

As shown in Fig. 3 and S8, each type of nanopore was evaluated through all these analytes separately, and their sensing performance was assessed through capture frequency (f), t, and full width of half maximum (FWHM) of I/I₀. The KQ AeL was responsive to oligonucleotides, miRNAs, peptides and small-molecule metabolites, but it showed the highest discrimination ability for WT EGFR ctDNA and its T790M mutant segment with narrower FWHM (Fig. S9). Due to the higher capture rate and longer duration time of the analyte, the TK/KQ AeL nanopore could effectively differentiate miR-21 and miR-92a segments, peptide fragments from PD-L1 with or without tyrosine phosphorylation, as well as small metabolic molecules such as FAD. WT OmpF reliably detected DiLacNAc as well as other oligosaccharides as described in previous work,³¹ like human milk oligosaccharide (HMO)⁴⁸ motifs (Fig. S10), whereas other nanopores in this study yielded few events.

Fig. 3 Multiplex biomolecule detection with orthogonal nanopores. (a) Sensing characteristics, including f (red), t (green) and FWHM of I//I₀ (blue), for the detection of five types of biomolecules with four nanopores, respectively. Diagonal lines indicate few signals produced preventing statistical analyses. Darker colors indicate the stronger sensing capability of a given nanopore towards a specific biomolecule. All color scales are presented on a logarithm scale to facilitate comparison. (b) The current trace, typical events, scatter plots of the I/I₀ versus t, and Gaussian fits of the histograms of I/I₀. From left to right and top to bottom: detection of 2 µM Poly(dA)₄ with WT AeL at an applied voltage of +100 mV with 1000 events. Detection of a mixture of 2 µM WT EGFR and 2 µM T790M EGFR segments with KQ AeL at +140 mV (2000 events) showed larger blockage current for T790M EGFR segments. Detection of a mixture of 5 µM miR-21 and 5 µM miR-92a segments using TK/KQ AeL at +100 mV (1000 events) showed two peaks from miR-92a exhibiting longer t and larger I/I₀ compared with miR-21. Detection of a mixture of 2 µM PD-L1 peptide and 2 µM p-PD-L1 with TK/KQ AeL at +140 mV (2000 events) showed longer t for p-PD-L1 compared with PD-L1. Detection of 1 mM DiLacNAc with WT OmpF at −100 mV with 400 events. Detection of 2 µM FAD with TK/KQ AeL at +80 mV with 1000 events. Data were recorded in 1.0 M KCl, 10 mM Tris, pH 8.0, with a sampling rate of 100 kHz and filtering at 5 kHz using a homemade multichannel microchip and instrument.

Multi-nanopore cross-validation for multiplex biomolecule sensing

Collectively, these current signatures from multiple nanopores under identical on-chip experimental conditions constitute a consistent, high-quality dataset suitable for comparative sensing while reducing measurement discrepancies arising from different sensing approaches. Taking advantage of these reliable measurements, we implemented a machine-learning (ML) strategy^49,50 based on a Random Forest model for rigorous, high-throughput and automated profiling of current signals (Fig. 4a and S11). The event features of I/I₀ and t were highly reproducible; therefore, these two parameters were selected for model training and validation. To reduce interference, the Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN) algorithm was applied as a pre-clustering denoising processing. After HDBSCAN-based filtering, each 3 min recording was evaluated based on the number of retained events. Recordings containing fewer than 10 events were considered insufficient for reliable statistical learning and were excluded from subsequent machine learning analysis to prevent overfitting and ensure model robustness. The cross-validation confusion matrices and learning curves for the four nanopores are shown in Fig. S12 and S13. Then, the ML models were used in the identification of a mixture of WT EGFR, miR-21, PD-L1, DiLacNAc, and FAD using multiple orthogonal nanopores.

Fig. 4 Cross-validation using parallel orthogonal nanopores for multiplex biomolecule profiling in mixtures. (a) NMS for relative quantification using multiple nanopores for cross-validation (schematic). Machine learning with the linear fitting model enables identification and concentration prediction of each analyte. (b) Scatter plots classified by ML models were obtained from a mixture sample containing 1 µM WT EGFR, 2.5 µM miR-21, 1 µM PD-L1, 0.5 mM DiLacNAc, and 1 µM FAD detection with WT AeL, KQ AeL, TK/KQ AeL at +100 mV, and WT OmpF at −100 mV for 3 min, respectively. (c) Concentration prediction error for a mixture containing five types of analytes: 1 µM EGFR, 2.5 µM miR-21, 1 µM PD-L1, 0.5 mM DiLacNAc and 1 µM FAD. Relative error is equal to the absolute value obtained by subtracting the true value from the predicted value, and then dividing by the true value. Gray represents the incapability of nanopore sensing for this type of analyte. Blue bars highlight the lowest error achieved by the multi-nanopore cross-validation method for five types of biomolecules. Data were recorded with a sampling rate of 100 kHz and filtering at 5 kHz using a homemade microchip and amplifier array.

However, using a single type of nanopore makes it challenging to simultaneously identify all types of analytes because of differences in selectivity, leading to false positives and false negatives in multiplex sensing. Herein, a multiple-nanopore-based linear fitting model enabled cross-validated quantification using concentration-dependent event counts from each nanopore. As demonstrated in previous work,^{10,49,51–53} the capture frequency of single-molecule translocation events shows a linear relationship with analyte concentration,

N = α × N^[C] (1)

where N is the number of machine learning-predicted events for the mixture sample, α is the fitted coefficient, and N^[C] is the classified number for a pure sample at a known concentration ([C]). Therefore, for each analyte (i), the predicted concentration C_i of each nanopore was directly calculated from the number of ML-predicted events using:

C_i = α × [C_i] (2)

where α = N/N^[C]. Using this relationship, all analytes in each nanopore yields independent linear eqn (3),where N_i is the classified number of each analyte in the complex sample, α_i is the fitted concentration coefficient of each analyte, is the number of each pure analyte at a known concentration ([C_i]), which was classified as analyte i, and n is the type of analyte in total. Eqn (3) describes the multivariate linear equations formed by a single nanopore for the sensing of multiple analytes, with each equation corresponding to the classified counts of each analyte in the mixture. By jointly solving these overdetermined equations across all nanopores, the concentrations of analytes in the mixture were predicted. To reduce prediction errors, we weighted the data of each nanopore by its classification accuracy in the trained Random Forest model. Given that the statistical event frequency remains constant at a fixed analyte concentration, data collected over 3 min were analyzed using weighted least squares (WLS) fitting for reliable prediction (Table S2–S4). To validate this approach, it was first applied to simulated datasets mimicking complex samples. As shown in Fig. S14, the predicted concentrations closely matched the real values across mixtures with different concentration ratios, confirming the feasibility and reliability of the model. We then extended this method to experimental data, which also showed good agreement with the nominal concentrations. For example, in mixtures containing 2 or 3 analytes, the cross-validation model could identify the components well and predict their concentrations (Fig. S15, S16, Table S5 and S6). A single nanopore can give response to several molecule types, but its varified sensing ability across diverse analytes often leads to prediction deviations in ML (Fig. S16). When 5 analytes were mixed for validation, the multiple-nanopore cross-validation method consistently yielded lower relative errors (Fig. 4b and c, Table S7). Therefore, our method could employ the complementary strengths of each nanopore to improve the accuracy and robustness of molecular identification and quantification.

To evaluate real-sample compatibility, the ML model was applied further to identify a mixture of WT EGFR, miR-21, PD-L1, DiLacNAc, and FAD in the presence of 1% fetal bovine serum (FBS), enabling evaluation of its stability within a complex and nonspecific molecular background. Each nanopore accurately recognized its preferred targets despite the complex background, demonstrating the potential of NMS for the sensing of clinical samples (Fig. S17–S19). However, at high analyte concentrations, the nanopore becomes saturated by multiple molecule translocations,^54,55 which decreases sensitivity to coexisting molecules (Fig. S20 and S21). Conversely, low-abundance analytes may yield too few events submerged in background noise to enable reliable quantification.⁵⁶ To address these challenges, we defined the upper detection limit of concentration as a minimum resolvable event interval time of 10 ms (Table S8 and S9). The lower detection limit was set at twice the background events frequency over a 3 min window (Table S9). The prediction accuracy gradually improved with increasing numbers of integrated nanopores (Fig. S22), suggesting that incorporation of more orthogonal nanopores could further improve the prediction performance. To further increase throughput and experimental efficiency, we implemented a microchip-based automated system compatible with liquid handling automation, enabling standardized membrane assembly and parallelized workflows. As shown in Fig. S23, for validation, six chambers produced consistent and reproducible measurements. By integrating the quantification model with the automated platform, NMS paves the way for efficient, reproducible and high-throughput multiplex single-molecule sensing.

Conclusions

We developed a nanopore multiplex sensor that integrates a series of orthogonal nanopores into a low-noise bioelectronic microchip platform for multiplex biomolecule sensing. Through the combination of ML classification with multi-nanopore-driven linear fitting model, NMS achieved rapid identification and relative quantification of diverse analytes in a mixture. This platform reduced the need for labor-intensive standard curve generation and cross-approach calibration while enhancing accuracy through multi-nanopore cross-validation. The NMS exhibited high integrity, stability, accuracy, and efficiency, enabling simultaneous, time-resolved multiplex biomolecule sensing from a single, low-volume sample. Such an integrated orthogonal nanopore platform could improve data comparability and measurement throughput by minimizing variability arising from separate measurements and batch-to-batch differences. Furthermore, the capability for concurrent analyses of biomarkers involved in molecular crosstalk might facilitate the capture of temporally correlated molecular variations, providing more comprehensive insights into disease progression, molecular regulation, and therapeutic response.

The employed orthogonal nanopores of AeL and OmpF have inherent benefits in detecting short biomolecules due to narrow nanopore size, offering high sensitivity for single-nucleotide discrimination, post-translational modification detection, and small-molecule sensing. Extending the applicability of the NMS to larger targets, such as full-length DNA, proteins and glycans, will require enzymatic digestion or the adoption of larger nanopores such as ClyA and FhuA. Detection of low-abundance biomarkers in real samples will benefit from integration with upstream enrichment and separation strategies directly on-chip. While validated results support system compatibility, further engineering and intelligent feedback control are needed to enhance throughput, enable independent modular control, and facilitate integration with clinical sample processing pipelines. Further improvements in prediction accuracy could be achieved through advanced ML models with expanded multidimensional signal feature extraction and optimized concentration prediction algorithms. By integrating more orthogonal nanopores with complementary sensing characteristics, NMS could serve as a scalable framework for DNA/RNA/peptide sequencing, single-omics analysis, and the concurrent sensing of diverse biomolecular types. Together, advances in orthogonal nanopore engineering, microchip packaging, integrated circuits, and intelligent data analysis algorithms will further enhance the performance of NMS for multiplex biomarker sensing, potentially contributing to the development of next-generation precision molecular diagnosis and comprehensive single-molecule multiomics analysis.

Author contributions

Y.-L. Y. and Y.-T. L. conceived the original idea of the research and supervised the project. L.-L. Z. designed the platform. L.-L. Z. and Y.-H. W. conducted experiments. L.-L. Z. and F. G. performed machine learning and decision-making algorithm. C.-B. Z. designed the multichannel ultra-low-noise amplifier circuit. L.-L. Z. processed the data and wrote the manuscript. L. W. synthesized oligosaccharides. All authors discussed the results and commented on the manuscript.

Conflicts of interest

There are no conflicts of interest to declare.

Data availability

The data supporting the conclusions reached from our study have been included as part of the supplementary information (SI). Supplementary information: materials, methods, and additional measurements. See DOI: https://doi.org/10.1039/d6sc03524h.

Acknowledgements

The authors thank Dr Meng-Yin Li, Jun-Ge Li, and Yan Gao from Nanjing University for aerolysin protein production and helpful discussion. This work was financially supported by National Key Research and Development Program of China (2022YFA1304604), National Natural Science Foundation of China (22334006, 22474055 and 22027806), Scientific Instrument Developing Project of the Chinese Academy of Sciences (PTYQ2024YZ0008), Jiangsu Province “Double First-Class” Initiative (0205-1480601101) and the Fundamental Research Funds for the Central Universities (020514380356).

References

1. J. Wen, Nat. Biomed. Eng., 2025, 9, 1386–1389.
2. L. C. Kida and C. A. Lareau, Nat. Methods, 2025, 22, 1135–1136.
3. M.-Y. Li, J. Jiang, J.-G. Li, H. Niu, Y.-L. Ying, R. Tian and Y.-T. Long, Nat. Methods, 2024, 22, 241–253.
4. J. Tu, C. D. Flynn, J. Yeom, Z. Wu, S. O. Kelley and W. Gao, Nat. Nanotechnol., 2025, 20, 1388–1404.
5. A. Baysoy, Z. Bai, R. Satija and R. Fan, Nat. Rev. Mol. Cell Biol., 2023, 24, 695–713.
6. J. Yin, Y. Zheng, Z. Huang, W. Zhou, Y. Yuan, P. Cai, Y. Bai, S. Yang, Y. Gao, S. Duan, Y. Wang, Z. Xu, W. Zhang, X. Zhang, Y. Wei, Y. Huang, Y. Liu, W. Wang, T. Yang, Z. Zhang, X. Chen, X. Zhang, J. Lv, F. Li, Y. Zhang, G. Zeng, X. Wang, W. Ma, G. Hou, S. Hao, C. Liu, Y. Lai, P. Liu, B. Wang, Y. Li, W. Zhang, P. Gao, J. Xie, M. A. Esteban, Y. Gu, X. Liu, J. Ji, T. Qi, B. Liu, H. Wang, Y. Zhao, X. Yang, X. Wang, R. Chen, J. Yang, Y. Yin, J. Wang, Y. Cao, X. Xu, L. Liu, X. Jin and C. Liu, Science, 2026, 391, eadt3130.
7. Y. Zheng, Y. Liu, J. Yang, L. Dong, R. Zhang, S. Tian, Y. Yu, L. Ren, W. Hou, F. Zhu, Y. Mai, J. Han, L. Zhang, H. Jiang, L. Lin, J. Lou, R. Li, J. Lin, H. Liu, Z. Kong, D. Wang, F. Dai, D. Bao, Z. Cao, Q. Chen, Q. Chen, X. Chen, Y. Gao, H. Jiang, B. Li, B. Li, J. Li, R. Liu, T. Qing, E. Shang, J. Shang, S. Sun, H. Wang, X. Wang, N. Zhang, P. Zhang, R. Zhang, S. Zhu, A. Scherer, J. Wang, J. Wang, Y. Huo, G. Liu, C. Cao, L. Shao, J. Xu, H. Hong, W. Xiao, X. Liang, D. Lu, L. Jin, W. Tong, C. Ding, J. Li, X. Fang and L. Shi, Nat. Biotechnol., 2024, 42, 1133–1149.
8. X. Wang, H. Huang, S. Jiang, J. Kang, D. Li, K. Wang, S. Xie, C. Tong, C. Liu, G. Hu, H. Li, C. Li, L. Yang, Y. Ding, S.-T. Li, F. Wang, J. U. Lohmann, Z. Liang and X. Gu, Nature, 2025, 644, 722–730.
9. T. Vu, A. Vallmitjana, J. Gu, K. La, Q. Xu, J. Flores, J. Zimak, J. Shiu, L. Hosohama, J. Wu, C. Douglas, M. L. Waterman, A. Ganesan, P. N. Hedde, E. Gratton and W. Zhao, Nat. Commun., 2022, 13, 169.
10. H. Bayley and P. S. Cremer, Nature, 2001, 413, 226–230.
11. A. Dorey and S. Howorka, Nat. Chem., 2024, 16, 314–334.
12. Y.-L. Ying, Z.-L. Hu, S. Zhang, Y. Qing, A. Fragasso, G. Maglia, A. Meller, H. Bayley, C. Dekker and Y.-T. Long, Nat. Nanotechnol., 2022, 17, 1136–1146.
13. S. Pal, A. Naik, A. Rao, B. Chakraborty and M. M. Varma, ACS Appl. Nano Mater., 2022, 5, 8804–8810.
14. L. Ratinho, N. Meyer, S. Greive, B. Cressiot and J. Pelta, Nat. Commun., 2025, 16, 3211.
15. M. Zhang, C. Tang, Z. Wang, S. Chen, D. Zhang, K. Li, K. Sun, C. Zhao, Y. Wang, M. Xu, L. Dai, G. Lu, H. Shi, H. Ren, L. Chen and J. Geng, Nat. Methods, 2024, 21, 609–618.
16. X. Wei, A. Choudhary, L. Y. Wang, L. Yang, M. J. Uline, M. Tagliazucchi, Q. Wang, D. Bedrov and C. Liu, Sci. Adv., 2024, 10, eadp8134.
17. Y. Yi, P. Song, Z. Li, J. Ju, G. Sun, Q. Ren, K. Zhou, L. Liu and H.-C. Wu, Nat. Nanotechnol., 2025, 20, 1079–1086.
18. G. Yao, B. Xia, F. Wei, J. Wang, Y. Yang, S. Ma, W. Ke, T. Li, X. Cheng, L. Wen, Y.-T. Long and Z. Gao, J. Am. Chem. Soc., 2025, 147, 1721–1731.
19. Q. Liu, Y. Ouyang, Y. Wang, S. Zhou, Y. Zhan and L. Wang, Adv. Healthcare Mater., 2025, 14, 2405058.
20. X. Zhang, X. Wang, Y. Zhang, D. Liu, X. Wei, Y. Shao, Z. Li and G. Huang, Angew. Chem., Int. Ed., 2026, 65, e23298.
21. A. J. W. Hartel, S. Shekar, P. Ong, I. Schroeder, G. Thiel and K. L. Shepard, Anal. Chim. Acta, 2019, 1061, 13–27.
22. A. Fragasso, S. Schmid and C. Dekker, ACS Nano, 2020, 14, 1338–1349.
23. G. Baaken, M. Sondermann, C. Schlemmer, J. Ruhe and J. C. Behrends, Lab Chip, 2008, 8, 938–944.
24. T. Tonooka, T. Osaki, K. Sato, R. Kawano and S. Takeuchi, Sens. Actuators, B Chem., 2021, 334, 129643.
25. L.-L. Zhang, C.-B. Zhong, T.-J. Huang, L.-M. Zhang, F. Yan and Y.-L. Ying, Chem. Sci., 2024, 15, 8355–8362.
26. L.-L. Zhang, C.-B. Zhong, J.-G. Li, H.-Y. Niu, Y.-L. Ying and Y.-T. Long, J. Electroanal. Chem., 2022, 915, 116266.
27. Z.-L. Hu, Y.-L. Ying, M.-Z. Huo, X.-F. Kong, X.-D. Yu, J.-R. Zhang and Y.-T. Long, J. Chem. Educ., 2020, 97, 4345–4354.
28. L. Iesu, M. Sai, V. Torbeev, B. Kieffer, J. Pelta and B. Cressiot, Chem. Sci., 2025, 16, 9730–9738.
29. J. Wang, S.-C. Liu, Z.-L. Hu, Y.-L. Ying and Y.-T. Long, J. Am. Chem. Soc., 2025, 147, 1781–1791.
30. Y.-H. Fu, X. Li, L. Ma, Y.-J. Wan, L.-M. Zhang, Y.-L. Ying and Y.-T. Long, J. Phys. Chem. C, 2024, 128, 1110–1115.
31. F. Gao, J. H. Wang, H. Ma, B. Xia, L. Wen, Y. T. Long and Y. L. Ying, Angew. Chem., Int. Ed., 2025, 64, e202422118.
32. M. Boeri, C. Verri, D. Conte, L. Roz, P. Modena, F. Facchinetti, E. Calabrò, C. M. Croce, U. Pastorino and G. Sozzi, Proc. Natl. Acad. Sci. U. S. A, 2011, 108, 3713–3718.
33. K.-Z. Chen, F. Lou, F. Yang, J.-B. Zhang, H. Ye, W. Chen, T. Guan, M.-Y. Zhao, X.-X. Su, R. Shi, L. Jones, X. F. Huang, S.-Y. Chen and J. Wang, Sci. Rep., 2016, 6, 31985.
34. A. Leal, N. C. T. van Grieken, D. N. Palsgrove, J. Phallen, J. E. Medina, C. Hruban, M. A. M. Broeckaert, V. Anagnostou, V. Adleff, D. C. Bruhm, J. V. Canzoniero, J. Fiksel, M. Nordsmark, F. A. R. M. Warmerdam, H. M. W. Verheul, D. J. van Spronsen, L. V. Beerepoot, M. M. Geenen, J. E. A. Portielje, E. P. M. Jansen, J. van Sandick, E. Meershoek-Klein Kranenbarg, H. W. M. van Laarhoven, D. L. van der Peet, C. J. H. van de Velde, M. Verheij, R. Fijneman, R. B. Scharpf, G. A. Meijer, A. Cats and V. E. Velculescu, Nat. Commun., 2020, 11, 525.
35. J. Lei, Y. Huang, W. Zhong, D. Xiao and C. Zhou, Anal. Chem., 2020, 92, 8867–8873.
36. Y. Chen, F. Yang, L. Zubovic, T. Pavelitz, W. Yang, K. Godin, M. Walker, S. Zheng, P. Macchi and G. Varani, Nat. Chem. Biol., 2016, 12, 717–723.
37. C. Lu, Z. Shan, J. Hong and L. Yang, Int. J. Oncol., 2017, 51, 235–244.
38. R. Tavallaie, J. McCarroll, M. Le Grand, N. Ariotti, W. Schuhmann, E. Bakker, R. D. Tilley, D. B. Hibbert, M. Kavallaris and J. J. Gooding, Nat. Nanotechnol., 2018, 13, 1066–1071.
39. Q. Zhang, R. Salzler, A. Dore, J. Yang, D. Ma, W. C. Olson and Y. Liu, J. Proteome Res., 2018, 17, 3932–3940.
40. L.-C. Chan, C.-W. Li, W. Xia, J.-M. Hsu, H.-H. Lee, J.-H. Cha, H.-L. Wang, W.-H. Yang, E.-Y. Yen, W.-C. Chang, Z. Zha, S.-O. Lim, Y.-J. Lai, C. Liu, J. Liu, Q. Dong, Y. Yang, L. Sun, Y. Wei, L. Nie, J. L. Hsu, H. Li, Q. Ye, M. M. Hassan, H. M. Amin, A. O. Kaseb, X. Lin, S.-C. Wang and M.-C. Hung, J. Clin. Invest., 2019, 129, 3324–3338.
41. X. Wang, X. Yang, C. Zhang, Y. Wang, T. Cheng, L. Duan, Z. Tong, S. Tan, H. Zhang, P. E. Saw, Y. Gu, J. Wang, Y. Zhang, L. Shang, Y. Liu, S. Jiang, B. Yan, R. Li, Y. Yang, J. Yu, Y. Chen, G. F. Gao, Q. Ye and S. Gao, Proc. Natl. Acad. Sci. U. S. A, 2020, 117, 6640–6650.
42. M. Miyake, N. Kohno, E. D. Nudelman and S.-i. Hakomori, Cancer Res., 1989, 49, 5689–5695.
43. H.-H. Lee, Y.-N. Wang, W. Xia, C.-H. Chen, K.-M. Rau, L. Ye, Y. Wei, C.-K. Chou, S.-C. Wang, M. Yan, C.-Y. Tu, T.-C. Hsia, S.-F. Chiang, K. S. C. Chao, I. I. Wistuba, J. L. Hsu, G. N. Hortobagyi and M.-C. Hung, Cancer Cell, 2019, 36, 168–178.
44. J. Benicky, M. Sanda, Z. Brnakova Kennedy, O. C. Grant, R. J. Woods, A. Zwart and R. Goldman, J. Proteome Res., 2021, 20, 485–497.
45. Z. Duan, R. Shi, B. Gao and J. Cai, J. Transl. Med., 2024, 22, 705.
46. M.-Y. Li, Y.-Q. Wang, Y.-L. Ying and Y.-T. Long, Chem. Sci., 2019, 10, 10400–10404.
47. K. Awasthi, F. L. Chang, P. Y. Hsieh, H. Y. Hsu and N. Ohta, J. Biophotonics., 2020, 13, e201960210.
48. Y. Zhu, G. Luo, L. Wan, J. Meng, S. Y. Lee and W. Mu, Crit. Rev. Biotechnol., 2021, 42, 578–596.
49. M. Li, J. Wang, C. Zhang, X. Zhao, Y. Xiong, Y. Cao, D. Wang, X. Li, X. Liang and G. Qing, ACS Nano, 2024, 18, 25155–25169.
50. R. Akita, A. Lysenko, K. A. Boroevich, T. Yokota, D. Kawai, R. Iizuka, T. Tsunoda and S. Uemura, Chem. Sci., 2025, 16, 18607–18615.
51. P. Waduge, R. Hu, P. Bandarkar, H. Yamazaki, B. Cressiot, Q. Zhao, P. C. Whitford and M. Wanunu, ACS Nano, 2017, 11, 5706–5716.
52. N. S. Galenkamp, M. Soskine, J. Hermans, C. Wloka and G. Maglia, Nat. Commun., 2018, 9, 4085.
53. J. Jiang, M.-Y. Li, X.-Y. Wu, Y.-L. Ying, H.-X. Han and Y.-T. Long, Nat. Chem., 2023, 15, 578–586.
54. C. Plesa, D. Verschueren, S. Pud, J. van der Torre, J. W. Ruitenberg, M. J. Witteveen, M. P. Jonsson, A. Y. Grosberg, Y. Rabin and C. Dekker, Nat. Nanotechnol., 2016, 11, 1093–1097.
55. R. A. Lazenby, F. C. Macazo, R. F. Wormsbecher and R. J. White, Anal. Chem., 2018, 90, 903–911.
56. K. Chuah, Y. Wu, S. R. C. Vivekchand, K. Gaus, P. J. Reece, A. P. Micolich and J. J. Gooding, Nat. Commun., 2019, 10, 2109.

---