Biosensors and Nanotheranostics

1. Introduction

Electrochemistry has long been one of the most practical ways to turn a molecular recognition event into a number. Its usual form, however, is an ensemble measurement: the current that reaches the potentiostat represents the averaged response of millions or even billions of nanoparticles (NPs) or biomolecules reacting at the same interface (Anderson & Zhang, 2016). That averaging is useful for extracting macroscopic thermodynamic and kinetic parameters, but it comes at a cost. Local differences in particle size, surface morphology, catalytic reactivity and interfacial charge-transfer kinetics are smoothed into a single mean value, and the information they carry is lost (Anderson & Zhang, 2016; Wang & Tao, 2015). On densely loaded electrodes the problem is compounded, because the diffusion layers of neighbouring particles overlap and make it difficult to evaluate charge transfer at any one site (Anderson & Zhang, 2016).

Single-entity electrochemistry (SEE) was developed to address precisely this limitation. Under labels such as nano-impact electrochemistry (NIE) and single-particle collision electrochemistry (SPCE), the approach interrogates individual nanostructures, viruses, cells and even single biomolecular binding events in real time (Ganesh et al., 2025; Wang & Tao, 2015). In its most common configuration, freely diffusing entities collide one at a time with an ultramicroelectrode (UME) or nanoelectrode held at a controlled potential, and each collision is registered as a discrete current transient (Cheng & Compton, 2014; Wang & Tao, 2015). The shape of that transient is informative. Conductive or electrocatalytic particles that touch the electrode briefly generate "hit-n-run" spikes, whereas particles that remain adsorbed produce "hit-n-stand" staircase steps (Wang & Tao, 2015). Non-conductive entities, by contrast, obstruct the local flux of a redox mediator and appear as step-down blockades in the steady-state current (Anderson & Zhang, 2016; Wang & Tao, 2015). Analysed collectively, the frequency, amplitude and duration of these events report on particle size, concentration, diffusion coefficient and single-particle catalytic turnover (Anderson & Zhang, 2016; Sokolov et al., 2017).

The diagnostic appeal of this capability is considerable. Clinicians increasingly need to quantify low-abundance biomarkers, including circulating microRNAs, viral nucleic acids and early cancer proteins, at concentrations that challenge conventional assays, and ideally without thermal cycling or laborious sample pre-treatment (Arab et al., 2027; Schoch et al., 2026). Functional nanomaterials such as gold and platinum nanoparticles (AuNPs, PtNPs), metal-organic frameworks (MOFs), two-dimensional MXenes and graphitic carbon nitride (g-C3N4) have become central to this effort, serving as nanoreporters, electrocatalysts and high-capacity carriers that strengthen electrochemical transduction (Arab et al., 2027; Natarajan et al., 2024). Yet sensitivity in single-entity biosensing has historically been capped by stoichiometry: in a classic 1:1 format, one target molecule can produce at most one particle impact (Bai et al., 2020). The development of one-to-many (1:n) amplification, in which a single microRNA-21 (miRNA-21) molecule triggers the release of approximately 4 × 107 catalytic PtNPs and supports a detection limit of 47 aM, showed that this ceiling could be lifted (Bai et al., 2020).

A parallel development has come from CRISPR-Cas biology. Once Cas12a, Cas13a or related effectors recognise a target sequence through their CRISPR RNA (crRNA), they acquire an indiscriminate trans-cleavage (collateral) activity that can be read out against reporter probes on an electrode (Dong et al., 2024; Schoch et al., 2026). When these reactions are organised on structured nanointerfaces, for example tetrahedral DNA nanostructures (TDNs) or conductive Faraday cage-type (FCT) scaffolds, steric hindrance is reduced and electron transfer becomes more efficient (Arab et al., 2027; Dong et al., 2024). Enzyme-free toehold-mediated strand displacement reactions (TSDR) and rolling circle amplification (RCA) have extended these schemes further, allowing PAM-free detection of SARS-CoV-2 genes at attomolar levels (Dong et al., 2024; Shi et al., 2023).

Progress in the laboratory has not yet translated into routine clinical use, and the reasons are partly physical and partly practical (Kanokkanchana & Tschulik, 2022; Schoch et al., 2026). The high ionic strength of biological buffers tends to aggregate nanoparticles and lower collision frequency (Bai et al., 2020). Non-uniform diffusion at the edge of a disk UME broadens and sometimes splits the distribution of impact amplitudes, while picoampere transients lasting less than a millisecond are distorted by the low-pass filters of standard instrumentation (Kanokkanchana & Tschulik, 2022). In serum or cell lysate, proteins and lipids foul the electrode, slow electron transfer and raise background noise (Lee et al., 2024; Schoch et al., 2026). Preserving enzyme conformation during immobilisation and reproducing micro- and nanoelectrode arrays across batches remain unresolved manufacturing questions (Schoch et al., 2026).

Against this background, the present review has four objectives. First, it sets out a mechanistic foundation for single-entity transduction, comparing particle collision at UMEs, single-nanoparticle immobilisation at nanoelectrodes and in-electrode FCT architectures (Anderson & Zhang, 2016; Arab et al., 2027; Wang & Tao, 2015). Second, it evaluates nanomaterial-assisted and enzymatic amplification, including nanocatalysis, 1:n target recycling and CRISPR-Cas collateral cleavage on three-dimensional DNA nanostructures (Bai et al., 2020; Dong et al., 2024; Schoch et al., 2026). Third, it surveys diagnostic applications spanning microRNAs, viral nucleic acids (SARS-CoV-2 and monkeypox virus), protein biomarkers such as neutrophil gelatinase-associated lipocalin (NGAL) and carcinoembryonic antigen (CEA), and toxic mercuric ions (Hg2+) (Dong et al., 2024; Ganesh et al., 2025; Natarajan et al., 2024; Shi et al., 2023). Fourth, it identifies the engineering bottlenecks that stand between these platforms and the clinic, with particular attention to equivalent-circuit (SPICE) modelling, antifouling surface chemistry and the emerging role of artificial intelligence and machine learning (AI/ML) in signal processing and microfluidic integration (Kanokkanchana & Tschulik, 2022; Schoch et al., 2026).

2. Single-Entity Electrochemistry and CRISPR-Cas Nanointerfaces for Ultrasensitive Biosensing in Complex Matrices

2.1 From Ensemble Averages to Single-Entity Measurement

Traditional electrochemical detection records the collective response of an enormous number of nanoparticles or electroactive molecules reacting simultaneously across an electrode surface (Anderson & Zhang, 2016; Wang & Tao, 2015). Such macroscale measurements remain indispensable for thermodynamic and kinetic characterisation, but they mask heterogeneity between individual particles in catalytic activity, morphology and interfacial charge transfer (Anderson & Zhang, 2016). Dense nanoparticle arrays add a further complication, since overlapping diffusion profiles make it hard to attribute a measured current to any particular particle (Anderson & Zhang, 2016). Single-entity electrochemistry, encompassing single-particle impact electrochemistry (SPIE) and nano-impact electrochemistry, emerged as a means of interrogating individual nanoscale entities, cells, viruses and single biomolecular binding events in real time (Cheng & Compton, 2014; Ganesh et al., 2025; Wang & Tao, 2015).

2.2 Stochastic Collision Dynamics and Transduction Modes

A typical impact experiment immerses a UME or nanoelectrode in a dilute suspension of electroactive or catalytic entities, which diffuse freely and collide with the electrode one at a time (Cheng & Compton, 2014; Sokolov et al., 2017). The resulting current transients depend on how the particle interacts with the surface (Figure 1). Catalytic or conductive nanoparticles that strike an otherwise inert electrode under an applied potential generate Faradaic spikes by locally catalysing the oxidation or reduction of a solution redox species, the electro-oxidation of hydrazine being a widely used example (Anderson & Zhang, 2016; Bai et al., 2020). Brief contact produces "hit-n-run" spikes, whereas permanent adsorption yields "hit-n-stand" staircase steps (Wang & Tao, 2015) (Figure 1A, B). When non-conductive entities such as polystyrene beads or liposomes collide with an active electrode, they transiently obstruct the mass transport of the redox mediator and produce characteristic step-down blockades (Anderson & Zhang, 2016; Wang & Tao, 2015) (Figure 1C). Statistical analysis of event frequency, amplitude and duration then yields particle size distributions, number concentrations, diffusion coefficients and single-particle catalytic turnover rates (Sokolov et al., 2017). Representative single-entity platforms and their transduction mechanisms are compiled in Table 1.

2.3 Spatial Transport Heterogeneity and Signal Fidelity

The elegance of the collision experiment is tempered by heterogeneity in local mass transport. Standard disk UMEs have a non-uniform primary current density, with both electric field and diffusive flux markedly enhanced at the disk perimeter relative to its centre (Kanokkanchana & Tschulik, 2022; Sokolov et al., 2017) (Figure 1E). Particles that land near the edge therefore produce larger spikes or blockades than those landing near the centre, giving bimodal amplitude distributions that complicate size deconvolution (Kanokkanchana & Tschulik, 2022). Several electrode designs have been introduced to counter this effect, including recessed Pt ring UMEs that maintain a uniform radial current density, carbon-coated nanopipettes that promote complete particle oxidation within a confined geometry, and interfaces between two immiscible electrolyte solutions (ITIES) (Anderson & Zhang, 2016; Wang & Tao, 2015). Instrumental artefacts are a separate concern. Fast-scan cyclic voltammetry (FSCV) at sweep rates of the order of 400 V/s has been used to record complete voltammograms of individual colliding particles (Anderson & Zhang, 2016), and SPICE equivalent-circuit simulations now allow capacitive charging, resistive and diffusional contributions to be separated, so that filter-induced amplitude attenuation and phase shifts can be corrected (Kanokkanchana & Tschulik, 2022).

2.4 Functional Nanomaterials and Three-Dimensional Conductive Scaffolds

Figure 1. Transduction modes and transport heterogeneity in single-entity collision electrochemistry. (A–C) Simulated, illustrative current–time traces for the three principal collision signatures: hit-n-run transient spikes produced by brief contact of catalytic or conductive nanoparticles with an inert ultramicroelectrode (A); hit-n-stand staircase steps arising from permanent particle adsorption (B); and step-down blockades caused by insulating entities that obstruct redox-mediator flux at an active electrode (C). (D) Schematic of the corresponding particle–electrode interactions, including electrocatalytic hydrazine (N2H4) oxidation at PtNPs. (E) Qualitative radial flux profile across a disk UME, showing edge-enhanced transport that produces bimodal amplitude distributions, compared with the near-uniform flux targeted by recessed ring geometries. Traces are schematic and not reproduced experimental data. Based on Anderson and Zhang (2016), Wang and Tao (2015), Sokolov et al. (2017) and Kanokkanchana and Tschulik (2022).

Figure 2. Nanomaterial-enabled signal amplification: one-to-many release and in-electrode Faraday cage-type architecture. (A) Workflow of the one-to-many (1:n) single-entity biosensor, in which hybridisation of target miRNA-21 to satellite magnetic nanoparticle–DNA–PtNP conjugates allows duplex-specific nuclease (DSN) to digest the DNA linker and release the target for further cycles; approximately 4 × 10⁷ PtNPs are liberated per target, separated magnetically and counted by catalytic collisions at a gold UME (LOD 47 aM). (B) Comparison of a conventional on-electrode assay, in which labels at the distal end of long recognition complexes suffer exponential electron-tunnelling decay, with a Faraday cage-type design, in which a conductive two-dimensional scaffold (MXene, rGO) encloses the complex, extends the outer Helmholtz plane and places all labels within the electron-transfer zone. Based on Bai et al. (2020), Arab et al. (2027), Nie et al. (2022) and Chen et al. (2021).

Engineered nanomaterials address what is, in practice, the central constraint on electrochemical biosensing: the need to detect scarce biomarkers without extensive sample preparation (Arab et al., 2027; Natarajan et al., 2024). AuNPs, PtNPs, MOFs, MXenes and carbon nanostructures function variously as nanoreporters, high-capacity nanocarriers and high-turnover electrocatalysts (Arab et al., 2027; Natarajan et al., 2024) (Table 3). DNA-templated bimetallic silver-platinum nanoclusters (DNA-Ag/Pt NCs), formed on single-stranded DNA scaffolds, illustrate the catalytic route particularly well. They mimic peroxidase, catalysing the reduction of hydrogen peroxide (H2O2) coupled with oxidation of 3,3′,5,5′-tetramethylbenzidine (TMB) to its blue oxidised form (Mu et al., 2023; Shateri et al., 2026). Platinum deposition onto the silver core quenches the intrinsic fluorescence of the silver cluster while raising peroxidase-like activity well above that of the monometallic counterparts, producing a tag that can be read optically or electrochemically (Mu et al., 2023; Shateri et al., 2026).

Three-dimensional conductive interfaces pursue a complementary goal, namely faster electron transfer and lower background capacitance (Arab et al., 2027). In-electrode Faraday cage-type architectures embed probe-label complexes within porous, highly conductive scaffolds such as Ag-doped Ti3AlC2 MXene nanosheets or single-atom Au/MXene heterostructures (Arab et al., 2027; Nie et al., 2022). Because the reporters sit inside the conductive network rather than at the distal end of a long recognition complex, the outer Helmholtz plane (OHP) is effectively extended, electron-tunnelling distances are shortened and a larger fraction of labels participates in the signal (Figure 2B). This design has supported near amplification-free microRNA detection (Arab et al., 2027; Nie et al., 2022) and, in MXene-based electrochemiluminescence (ECL) immunosensors, added a catalytic contribution to the geometric one (Chen et al., 2021). Three-dimensional porous graphene networks offer a related benefit by preventing π–π restacking of two-dimensional sheets and preserving accessible surface area (Natarajan et al., 2024).

2.5 One-to-Many Signal Amplification

A conceptual turning point in single-entity biosensing was the shift from 1:1 to one-to-many (1:n) signal transduction (Bai et al., 2020). In a standard single-collision bioassay, each target can generate no more than one collision or cleavage event, which places a hard limit on sensitivity. Bai et al. (2020) removed this limit with satellite magnetic nanoparticle-DNA-Pt nanoparticle (MN-DNA-PtNP) conjugates coupled to duplex-specific nuclease (DSN) target recycling (Figure 2A). Hybridisation of a single miRNA-21 strand allows DSN to digest the DNA linker in the resulting duplex, after which the target is freed to hybridise again. Repeated cycles release approximately 4 × 107 catalytic PtNPs per target molecule; these are separated magnetically from the intact conjugates and counted through their catalytic hydrazine-oxidation collisions at a gold UME, giving a detection limit of 47 aM (Bai et al., 2020) (Table 1).

2.6 Programmable CRISPR-Cas Systems and Nanointerface Cleavage Kinetics

CRISPR and its associated Cas endonucleases, including Cas12a, Cas13a and Cas9, have changed molecular diagnostics by uniting programmable sequence recognition with enzymatic amplification (Schoch et al., 2026). Guided by a customisable crRNA, the Cas protein binds its target nucleic acid to form an active ternary ribonucleoprotein complex (Dong et al., 2024; Schoch et al., 2026). For class 2, type V effectors such as Cas12a, target recognition switches on non-specific single-stranded DNA (ssDNA) trans-cleavage that degrades surrounding reporters with high turnover (Dong et al., 2024; Schoch et al., 2026), an activity that has also been resolved at the single-molecule level for SARS-CoV-2 gene detection (Ji & Koh, 2023). The principal CRISPR-integrated platforms reviewed here are summarised in Table 2.

When this chemistry is moved onto an electrode, nanoscale spatial organisation becomes decisive (Dong et al., 2024; Zhao et al., 2025). Conventional electrochemical CRISPR sensors immobilise linear ssDNA or two-dimensional hairpin probes directly on planar electrodes, where flexible strands become entangled, collapse locally and crowd the surface, restricting enzyme access and slowing trans-cleavage (Dong et al., 2024) (Figure 3A). Tetrahedral DNA nanostructures, self-assembled from four complementary oligonucleotides, were introduced as rigid three-dimensional scaffolds that anchor to the electrode with nanometre-scale spacing (Dong et al., 2024) (Figure 3B). The rigid framework prevents probe intertwining, resists non-specific protein adsorption and presents pendant ssDNA reporters to solution, accelerating Cas12a cleavage and enabling detection of microRNAs and circulating tumour DNA (ctDNA) at sub-femtomolar levels (Dong et al., 2024).

Table 1. Single-entity electrochemistry (SEE) and nanoparticle collision platforms for ultrasensitive sensing. The table compares representative single-entity platforms according to the nanoparticle tag, electrode architecture, target and matrix, analytical performance, and the mechanism by which an individual collision or adsorption event is converted into a measurable current. Entries span catalytic spike counting after one-to-many (1:n) target recycling, in situ formation and cathodic impact of Hg2Cl2 nanoparticles, fast-scan cyclic voltammetry of single Pt particles, recessed-ring geometries designed to remove edge effects, and hit-n-run versus hit-n-stand transients. Detection limits are reported as given in the original studies. Abbreviations: UME, ultramicroelectrode; MN, magnetic nanoparticle; DSN, duplex-specific nuclease; CFE, carbon fibre microelectrode; FSCV, fast-scan cyclic voltammetry; LOD, limit of detection.

Table 2. CRISPR-Cas integrated electrochemical and optical biosensors. The table summarises CRISPR-Cas biosensors according to effector system, target and disease context, sensing interface, upstream amplification cascade, analytical performance and the key mechanistic feature that distinguishes each design. Platforms include Cas12a acting on tetrahedral DNA nanostructures, PAM-free toehold-mediated strand displacement for SARS-CoV-2, Cas13a-triggered hybridisation chain reaction with electrochemiluminescent readout, magnetic-bead separation in serum, digital microfluidic automation and nanozyme-based colorimetric detection of a cancer protein. Optical entries are included where their interfacial or nanomaterial design informs electrochemical implementation. Abbreviations: TDN, tetrahedral DNA nanostructure; E-RCA, endonuclease-mediated rolling circle amplification; TSDR, toehold-mediated strand displacement reaction; HCR, hybridisation chain reaction; SDR, strand displacement reaction; RPA, recombinase polymerase amplification; DMF, digital microfluidics; EWOD, electrowetting-on-dielectric.

Magnetic bead (MB) nanointerfaces offer a different handle on the same problem. By varying the root (spacer), stem and loop lengths of MB-anchored hairpin reporters, Zhao et al. (2025) showed that both the cleavage rate and the cleavage site can be tuned (Figure 3C). Colocalising hairpins on the bead surface raises local reactant density and accelerates cleavage relative to unconfined linear ssDNA in solution (Zhao et al., 2025). Because Cas12a binds single-stranded loops more readily than double-stranded stems, adjusting stem stability allows cleavage to release defined single-stranded initiator strands, which can then drive hybridisation chain reactions (HCR) on the bead without thermal denaturation, linking CRISPR recognition to dynamic DNA nanotechnology (Xu et al., 2023; Zhao et al., 2025).

2.7 PAM-Free Activation, Isothermal Circuits and Non-Nucleic-Acid Targets

Native Cas12a requires a protospacer adjacent motif (PAM), typically 5′-TTTN-3′, on double-stranded DNA targets, which restricts the regions that can be selected within short viral or mutated sequences (Hu et al., 2026; Shi et al., 2023). PAM-free activation strategies circumvent this requirement by constructing toehold-mediated duplexes (toe-dsDNA) or artificial bubble structures in the seed region, allowing the guide RNA to hybridise with the target strand directly (Hu et al., 2026; Shi et al., 2023). Coupling PAM-free TSDR with Cas12a (PfTSDR-CRISPR) creates a dual isothermal cascade in which target RNA drives recycling displacement cycles that generate bubble-containing dsDNA activators; these in turn trigger cleavage of methylene blue-modified hairpin probes (MB-HP) on a gold electrode (Shi et al., 2023). Because the displacement region is physically separated from the Cas12a/gRNA recognition domain, the same gRNA serves several SARS-CoV-2 gene fragments (RdRp, ORF1ab and N) with a detection limit of 40 aM (Shi et al., 2023) (Table 2).

Enzyme-driven signal converters have extended CRISPR sensing beyond nucleic acids to cofactors and proteins (Schoch et al., 2026). Ligation-driven designs use E. coli or T4 DNA ligase to seal a single-stranded gap only in the presence of NAD+ or ATP, generating an intact dsDNA activator (Schoch et al., 2026). Aptamer-mediated switches have been used for prostate-specific antigen (PSA) (Taghdisi et al., 2022), and combining RCA with Cas12a cleavage of DNA-Ag/Pt NCs has enabled visual detection of CEA (Mu et al., 2023). Electrochemical platforms for NGAL, an acute kidney injury (AKI) biomarker, have likewise drawn on nanostructured transducers and dual-signal readouts (Liang et al., 2022; Natarajan et al., 2024).

2.8 Diagnostic Translation in Complex Matrices and Microfluidic Integration

Integrated single-entity and CRISPR-assisted sensors have now been tested in a range of complex biological and environmental matrices (Table 4). Immunomagnetic separation, antifouling nanocomposites and bead-mediated enrichment have allowed analytes to be isolated from diluted human serum, urine, food and environmental water (Duan et al., 2023; Lee et al., 2024; Zhang et al., 2023). Reported demonstrations include miRNA-21 quantification in urine from AKI patients (Xu et al., 2023), detection of Salmonella Typhimurium and Staphylococcus aureus in food and clinical fluids (Duan et al., 2023; Lu et al., 2023), monitoring of Hg2+ through single-particle impact reduction of in situ-formed Hg2Cl2 (Ganesh et al., 2025) and viral gene profiling (Lee et al., 2024; Shi et al., 2023).

Translation to the point of care nevertheless remains incomplete (Schoch et al., 2026). In whole blood or undiluted plasma, protein adsorption and high ionic strength reduce diffusion, coat the electrode and suppress Faradaic current (Lee et al., 2024; Schoch et al., 2026). Zwitterionic antifouling nanocomposites, biomimetic interfaces and integrated pre-treatment modules are the main responses so far (Arab et al., 2027; Lee et al., 2024). Digital microfluidic (DMF) chips that combine recombinase polymerase amplification (RPA) with Cas12a cleavage automate lysis, enrichment, mixing and readout within a closed device (Lu et al., 2023), and AI/ML methods are beginning to be proposed for baseline correction, deconvolution of stochastic impact traces and pattern recognition across multiplexed arrays (Kanokkanchana & Tschulik, 2022; Schoch et al., 2026). Figure 4 summarises how these elements might be assembled into a single diagnostic workflow.

3. Methods

3.1 Review Design and Reporting Framework

This work was designed as a structured narrative review with a systematic search component. Its reporting follows the principles of transparency, explicit eligibility criteria and reproducible data extraction recommended for literature

Figure 3. Nanoscale spatial organisation of CRISPR-Cas12a reporter interfaces. (A) Linear one- or two-dimensional ssDNA probes immobilised directly on a planar electrode become crowded and entangled, creating steric hindrance that restricts Cas12a access and slows trans-cleavage. (B) Rigid tetrahedral DNA nanostructures anchor pendant reporters at defined nanometre-scale spacing, exposing them to solution and accelerating Cas12a-mediated cleavage of methylene blue-labelled probes. (C) Hairpin reporters colocalised on magnetic beads, in which root, stem and loop dimensions tune cleavage rate, cleavage site and Cas12a affinity; regulated cleavage releases single-stranded initiators that can trigger downstream hybridisation chain reaction (HCR). Schematic; not to scale. Based on Dong et al. (2024), Zhao et al. (2025) and Xu et al. (2023).

Figure 4. Translational workflow and stage-specific bottlenecks for point-of-care single-entity and CRISPR-electrochemical diagnostics. The schematic assembles the strategies reviewed into a sample-to-answer sequence: a clinical or environmental sample (serum, urine, saliva, milk or wastewater) undergoes matrix management through antifouling zwitterionic coatings or magnetic and immunomagnetic separation; target signals are amplified by PAM-free strand displacement, RCA, RPA, HCR, CRISPR trans-cleavage or one-to-many nanoparticle release; transduction proceeds by nano-impact counting, voltammetric or impedance methods, ECL or Faraday cage-type scaffolds; and readout is automated by digital microfluidics with SPICE-informed or machine-learning signal deconvolution. The principal bottleneck addressed at each stage is shown beneath it. Based on Lu et al. (2023), Lee et al. (2024), Duan et al. (2023), Kanokkanchana and Tschulik (2022) and Schoch et al. (2026).

reviews indexed in PubMed, and a PRISMA-style record of identification, screening and inclusion was maintained throughout. A formal meta-analysis was not attempted, because the included studies differ in target analyte, transduction principle, matrix and units of detection, and pooling their detection limits would not yield a meaningful summary estimate. The review protocol was not prospectively registered.

3.2 Information Sources and Search Strategy

Records were identified from PubMed/MEDLINE, Web of Science Core Collection and Scopus, supplemented by backward and forward citation tracking of key reviews and primary studies (Anderson & Zhang, 2016; Cheng & Compton, 2014; Schoch et al., 2026; Wang & Tao, 2015) and by a search of the SSRN preprint server. The search covered publications from January 2014, which corresponds to the consolidation of nano-impact methods as an analytical field (Cheng & Compton, 2014), to [final search date to be inserted by the authors], including articles in press. Search strings combined three concept blocks with Boolean operators: (1) single-entity terms ("single-entity electrochemistry" OR "nano-impact" OR "particle collision" OR "single nanoparticle" OR "stochastic electrochemistry" OR "ultramicroelectrode"); (2) amplification and recognition terms ("CRISPR" OR "Cas12a" OR "Cas13a" OR "strand displacement" OR "rolling circle amplification" OR "hybridization chain reaction" OR "Faraday cage" OR "nanozyme" OR "MXene" OR "tetrahedral DNA"); and (3) application terms ("biosensor" OR "microRNA" OR "nucleic acid" OR "biomarker" OR "diagnosis" OR "mercury"). Controlled vocabulary (MeSH terms such as Biosensing Techniques, Electrochemical Techniques, CRISPR-Cas Systems and Nanoparticles) was used in PubMed and adapted for each database. 

3.3 Eligibility Criteria

Studies were eligible if they (i) reported an electrochemical, electrochemiluminescent or closely related nanomaterial-based transduction scheme operating at the single-entity level or using nanostructured interfaces relevant to single-entity amplification; (ii) described a recognition or amplification mechanism in enough detail for its logic to be reconstructed; and (iii) reported at least one quantitative analytical figure of merit, such as limit of detection (LOD), linear range, recovery or relative standard deviation (RSD). Optical CRISPR assays were admitted only when their nanomaterial or interfacial design informed electrochemical implementation, as with peroxidase-mimicking DNA-Ag/Pt nanoclusters (Mu et al., 2023) or cruciform DNA aptasensors (Taghdisi et al., 2022). Peer-reviewed articles, critical reviews and preprints in English were considered. Conference abstracts were retained only where they reported distinct single-molecule data (Ji & Koh, 2023). Studies without analytical validation, purely theoretical modelling with no experimental reference, and duplicate reports of the same dataset were excluded.

3.4 Study Selection Process

All retrieved records were exported to a reference manager and de-duplicated by DOI and title. Two reviewers independently screened titles and abstracts against the eligibility criteria, followed by full-text assessment of potentially eligible articles. Disagreements were resolved by discussion and, where necessary, by a third reviewer. The final evidence base comprises the 29 sources listed in the reference list.

3.5 Data Extraction and Variable Definitions

A standardised extraction form, piloted on five studies before full use, captured the following fields for each primary study: sensing system or technique; nanoparticle or nanomaterial tag, including size where reported; electrode or interfacial architecture; target analyte and sample matrix; recognition element (antibody, aptamer, crRNA or molecularly imprinted polymer); amplification cascade; transduction readout (amperometric impact counting, differential pulse voltammetry, electrochemical impedance spectroscopy, ECL or colorimetry); LOD and linear range; recovery and RSD; assay time and reagent volume; and the reference method used for clinical validation (for example ELISA or qRT-PCR). For single-entity studies, extraction additionally recorded the collision mode (spike, staircase or blocking), the electrode material and geometry, and the applied potential and redox mediator, consistent with the reporting conventions described by Sokolov et al. (2017). Where signal distortion was discussed, filter settings and any equivalent-circuit correction were recorded (Kanokkanchana & Tschulik, 2022). Extraction was carried out by one reviewer and checked in full by a second; values were transcribed exactly as reported, and the original units were retained alongside any converted values.

3.6 Harmonization of Analytical Metrics

To allow comparison across studies, molar detection limits were expressed on a common scale (aM, fM, pM or nM), with mass-based protein limits (pg/mL or ng/mL) and bacterial counts (CFU/mL) reported in their original units because conversion would require assumptions about molecular weight or cell state. LOD values were accepted as defined by each study, which in most cases used a signal three standard deviations above the blank; studies applying other definitions were flagged. Detection limits obtained in buffer and in real matrices were recorded separately, since the difference between them is itself an indicator of matrix susceptibility (Zhang et al., 2023).

3.7 Quality Appraisal

Because no validated risk-of-bias instrument exists for analytical biosensor studies, each primary study was appraised against a seven-item checklist adapted for this review: (1) clear description of electrode fabrication; (2) reporting of replicate number and RSD; (3) inclusion of specificity or interference testing; (4) testing in at least one real or spiked matrix; (5) comparison against an accepted reference method; (6) reporting of stability or reproducibility across electrodes or batches; and (7) availability of raw or supporting data. Studies were not excluded on quality grounds, but appraisal results informed the weight given to each study in the synthesis and are discussed where they bear on interpretation.

3.8 Evidence Synthesis and Construction of Tables and Figures

Extracted data were synthesised thematically. Studies were grouped into four domains that structure Tables 1 to 4: single-entity collision platforms (Table 1), CRISPR-Cas integrated sensors (Table 2), functional nanomaterial and interfacial architectures (Table 3), and applications in clinical and real-world matrices (Table 4). Within each domain, findings were compared by mechanism, analytical performance and matrix, and recurrent design principles were identified. Figures 1 to 4 are original schematic illustrations prepared by the authors in Python (Matplotlib) on the basis of the mechanisms described in the cited studies; current traces in Figure 1 are simulated for illustration and are not reproduced experimental data.

4. Sensitivity Gains, Interfacial Design Rules and Matrix Performance Across Single-Entity and CRISPR-Electrochemical Platforms

4.1 Overview of the Evidence Base

Read together, the studies point in a fairly consistent direction. The field appears to be moving away from static ensemble measurements at macro-electrodes and toward dynamic single-entity readouts and three-dimensional, CRISPR-functionalised interfaces (Anderson & Zhang, 2016; Schoch et al., 2026). What the evidence seems to show, more specifically, is that sensitivity gains rarely come from a single innovation. They tend to arise where nanoscale spatial control, programmable enzymatic cleavage and an efficient amplification cascade are combined, and it is in those combinations that attomolar quantification has been reported in clinical and environmental samples (Bai et al., 2020; Dong et al., 2024; Shi et al., 2023) (Tables 1, 2 and 4).

4.2 Amplification Efficiency in Single-Entity Platforms

Early impact assays faced what might be called a stoichiometric ceiling, since one target could produce, at best, one collision (Bai et al., 2020). The 1:n strategy of Bai et al. (2020) is the clearest example in the evidence base of that ceiling being broken (Table 1; Figure 2A). In their design, binding of a single miRNA-21 molecule to a satellite MN-DNA-PtNP conjugate initiates repeated DSN-catalysed cleavage of the DNA linkers, so that roughly 4 × 107 PtNPs, each about 4.5 nm in diameter, are released for every target. After magnetic separation, these particles collide with a gold UME and catalyse hydrazine oxidation, producing discrete amperometric spikes. The reported LOD was 47 aM, with a linear range of 50 aM to 5 nM (Bai et al., 2020). The authors described this as an improvement of almost six orders of magnitude over a conventional 1:1 impact assay; whether that figure holds across other targets has not, as far as we can tell, been tested independently.

4.3 Heavy-Metal Sensing and Kinetic Resolution of Impact Events

A rather different use of single-entity counting appears in environmental analysis. Ganesh et al. (2025) described a one-step multi-potential-step approach in which trace Hg2+ is converted in situ into Hg2Cl2 nanoparticles that are then reduced cathodically on collision with a carbon UME (Table 1). Spike frequency rose exponentially with Hg2+ concentration, and sub-nanomolar detection was achieved without chemical reducing agents or electrode functionalisation (Ganesh et al., 2025). The work is currently available as a preprint, which is worth keeping in mind. On the measurement side, FSCV at 400 V/s has made it possible to extract full voltammograms from individual Pt nanoparticles of 4 to 80 nm (Anderson & Zhang, 2016), and recessed ring UMEs appear to remove much of the edge-related spread in step amplitudes (Table 1; Figure 1E). Kanokkanchana and Tschulik (2022) add an important caution, however: some of the asymmetry and attenuation seen in published transients may originate in the measurement circuit rather than in the chemistry.

4.4 Comparative Performance of CRISPR-Cas Nanointerfaces

Among CRISPR-integrated sensors (Table 2), interfacial architecture emerges as a strong, if not always decisive, determinant of performance. Sensors that anchor flexible ssDNA or hairpin probes directly to planar electrodes repeatedly report restricted enzyme access and probe crowding (Dong et al., 2024). The TDN platform of Dong et al. (2024), assembled on AuNP-modified glassy carbon electrodes and coupled with endonuclease-mediated RCA (E-RCA), offers perhaps the most direct comparison. Rigid tetrahedra hold methylene blue-labelled probes at defined spacings; once E-RCA generates ssDNA activators, Cas12a cleaves the vertex probes and the methylene blue signal falls. The resulting LOD for miRNA-155 was 0.5 fM across a 10 fM to 1 nM range (Dong et al., 2024) (Figure 3B).

The magnetic bead studies tell a complementary story. Zhao et al. (2025) found that hairpin reporters on beads were cleaved faster, and with better signal-to-background ratios, than unconfined linear ssDNA. Stem and root lengths governed cleavage site and stability, whereas the loop domain influenced Cas12a binding affinity (Figure 3C). Regulated cleavage released single-stranded initiators that could drive HCR, a coupling that Xu et al. (2023) exploited with Cas13a to reach an LOD of 0.53 fM for miRNA-21 (Table 2). It is tempting to conclude that three-dimensional confinement is generally superior. The comparison, though, involves different targets, enzymes and readouts, so that inference should be held loosely until head-to-head studies are available.

4.5 PAM-Free and Non-Nucleic-Acid CRISPR Formats

PAM-free designs address a practical rather than a sensitivity problem. Shi et al. (2023) engineered a six-nucleotide artificial bubble into the seed region of the output dsDNA, physically separating the target displacement region from the gRNA recognition domain. This allowed one crRNA to serve three SARS-CoV-2 fragments (RdRp, ORF1ab and N) with an LOD of 40 aM, a range of 100 aM to 10 pM and an RSD below 4.3% (Shi et al., 2023) (Table 2). Hu et al. (2026) extended the toe-dsDNA concept to DNA methylation, reporting roughly five-fold better discrimination of single-CpG methylation than a conventional PAM-containing duplex.

For targets other than nucleic acids, the evidence is thinner but suggestive. Ligation-driven converters produce intact dsDNA activators only in the presence of NAD+ or ATP (Schoch et al., 2026). Taghdisi et al. (2022) designed a PSA aptasensor in which target binding prevents the aptamer from hybridising with crRNA, leaving Cas12a inactive and allowing a PicoGreen-loaded cruciform DNA structure to form. Mu et al. (2023) linked CEA recognition to RCA-generated Cas12a activators that degrade DNA-Ag/Pt NCs and suppress their peroxidase activity, reaching 0.85 ng/mL in clinical serum (Table 2).

4.6 Nanomaterial Architectures and Signal Enhancement

The nanomaterial studies (Table 3) suggest that structure and conductivity govern both charge-transfer efficiency and label loading. FCT biosensors are the most distinctive development here (Arab et al., 2027). In a conventional on-electrode sandwich assay, redox or ECL labels sit at the outer edge of a long immune or nucleic-acid complex, where electron tunnelling decays exponentially with distance (Figure 2B). FCT designs instead use conductive two-dimensional materials such as Ti3C2Tx MXene, Ag-doped Ti3AlC2 MAX/MXene and reduced graphene oxide, co-functionalised with receptors and labels, so that the captured complex is enclosed within a low-resistance pathway (Arab et al., 2027). Single-atom Au/MXene heterostructures supported microRNA detection by this route (Nie et al., 2022). In the Cry1Ab immunosensor of Chen et al. (2021), MXene nanosheets carrying 3,4,9,10-perylenetetracarboxylic acid and secondary antibodies seemed to do two things at once: connecting the labels electrically to the electrode and catalysing O2 reduction to reactive oxygen species. Zhou et al. (2023) combined FSCV and ECL in a dual-mode cytosensor built from black phosphorus nanosheets decorated with copper nanoparticles and MXene@Fe3O4, which distinguished

A549 and MCF-7 cells.

Bimetallic nanozymes outperform their monometallic counterparts in the reports available (Mu et al., 2023; Shateri et al., 2026). Platinum deposits onto a silver core of about 2.5 nm, quenching fluorescence while sharply accelerating H2O2 decomposition to hydroxyl radicals. The Michaelis constant (Km) for TMB was 0.061 mM, about seven-fold lower than that of horseradish peroxidase, with a Vmax of 10.6 × 10−8 mol L−1 s−1 (Mu et al., 2023). For AKI, three-dimensional composites including Prussian blue-decorated g-C3N4 (PB-g-C3N4), AuNP-decorated graphene foam and Au/Cu-MOF/single-walled carbon nanohorns (SWNHs) have pushed NGAL detection lower (Natarajan et al., 2024). The dual-signal Au/Cu-MOF/SWNH sandwich sensor, which combined square-wave voltammetry of the Cu2+/Cu+ couple with amperometric H2O2 reduction, reported an LOD of 0.0074 pg/mL (Liang et al., 2022; Natarajan et al., 2024). Label-free formats are also represented: nanoporous gold electrodes quantified VEGF165 through modulation of [Fe(CN)6]3−/4− signals (Pourbahram et al., 2025), and dual-template molecularly imprinted polymers achieved selective protein recognition in the presence of gamma-globulin (Jalili & Jalalvand, 2023).

4.7 Performance in Clinical and Real-World Matrices

The matrix studies (Table 4) are where the picture becomes more mixed. Direct analysis of serum, urine or saliva brings non-specific protein adsorption, ionic-strength-driven aggregation and elevated background (Lee et al., 2024; Schoch et al., 2026). Zhang et al. (2023) used streptavidin magnetic beads to remove biotinylated reaction products from 1% human serum, leaving free trigger strands to activate Cas12a. Their LOD for miRNA-21 rose from 2.42 fM in buffer to 12.12 fM in serum, a five-fold loss that is modest but not trivial (Table 2). Xu et al. (2023), working with urine from AKI patients, removed unbound Ru(phen)32+ magnetically before ECL measurement and reported recoveries of 96.9% to 105%, RSDs of 1.43% to 4.57% and close agreement with qRT-PCR across disease stages.

Pathogen detection without DNA extraction has also been shown. The CATCHER platform of Duan et al. (2023) used immunomagnetic capture of Salmonella Typhimurium in chicken and milk, reaching 79 CFU/mL in buffer and 631 CFU/mL in chicken within about two hours. Lu et al. (2023) automated lysis, RPA and Cas12a cleavage on an electrowetting-on-dielectric DMF chip, detecting S. aureus in milk and urine at 32 CFU/mL in 55 minutes with 6.6 µL of reagent. Lee et al. (2024) reported amplification-free detection of monkeypox virus DNA in undiluted serum on zwitterionic antifouling electrodes, with minimal signal drift. Taken as a whole, these results suggest that matrix management, rather than intrinsic sensitivity, is now the more frequent limiting factor (Figure 4).

5. Interface Design, Matrix Effects, and Clinical Validation Gaps in Single-Entity CRISPR Biosensing

5.1 Amplification, Interface Design, and Real-Matrix Performance in Single-Entity CRISPR Biosensing

Three conclusions seem reasonably secure. First, the stoichiometric limit that once constrained impact-based biosensing is not fundamental; it can be lifted by amplification cascades that convert one recognition event into many collisions (Bai et al., 2020) (Table 1; Figure 2A). Second, interfacial architecture, whether TDNs, hairpin-decorated beads or conductive FCT scaffolds, matters at least as much as the choice of enzyme or nanomaterial (Arab et al., 2027; Dong et al., 2024; Zhao et al., 2025) (Tables 2 and 3; Figures 2B and 3). Third, and perhaps least comfortably, performance in real matrices lags behind performance in buffer, sometimes by a substantial margin (Duan et al., 2023; Zhang et al., 2023) (Table 4). Each of these points deserves some qualification.

5.2 Is Attomolar Sensitivity the Right Benchmark?

The detection limits in Tables 1 and 2 are impressive, ranging from 40 aM for SARS-CoV-2 RNA (Shi et al., 2023) to 47 aM for miRNA-21 (Bai et al., 2020) and 0.5 fM for miRNA-155 (Dong et al., 2024). It is less obvious, however, that pushing these limits further is where effort is best directed. At 47 aM, a 10 µL sample contains only a few hundred target molecules, and sampling statistics begin to dominate the uncertainty well before the transducer does. For many clinical decisions, a reliable femtomolar assay that works in undiluted serum may well be more useful than an attomolar one that works only in buffer. Such a perspective does not diminish the 1:n concept; rather, it suggests that the amplification headroom it provides could be spent on robustness, shorter assays or simpler hardware rather than on lower detection limits alone.

5.3 Interfacial Architecture as a Design Variable

The findings on interfacial organisation point to something like a set of design rules, though they remain provisional. Rigid scaffolds that fix probe spacing appear to accelerate trans-cleavage (Dong et al., 2024); bead-confined hairpins allow cleavage site and rate to be tuned (Zhao et al., 2025); and conductive enclosures bring otherwise distant labels into the electron-transfer zone (Arab et al., 2027; Chen et al., 2021; Nie et al., 2022) (Figures 2B and 3). What is missing is a quantitative link between these structural variables and the observable signal. The studies vary target, enzyme, electrode material and readout simultaneously, so the contribution of architecture alone is hard to isolate. Systematic studies that hold everything else constant would help considerably, as would reporting of probe surface density, which is often absent.

5.4 Matrix Effects, Biofouling and the Gap to Clinical Samples

The matrix data in Table 4 are encouraging, but they should be read with some care. Several studies used diluted serum (Zhang et al., 2023), spiked rather than naturally infected samples (Duan et al., 2023) or simulated swabs (Shi et al., 2023). Only a minority compared their results against a clinical reference method in patient samples, Xu et al. (2023) with qRT-PCR and the NGAL sensors with ELISA (Natarajan et al., 2024) being notable exceptions. For single-entity measurements in particular, high ionic strength remains a basic obstacle, because it promotes nanoparticle aggregation and lowers collision frequency (Bai et al., 2020). Magnetic separation appears to be the most broadly effective response, since it physically removes the reporter from the matrix before detection (Bai et al., 2020; Xu et al., 2023; Zhang et al., 2023). Antifouling zwitterionic coatings, as used by Lee et al. (2024), address the problem at the electrode instead. It is not yet clear which of these strategies will scale better, and it is plausible that both will be needed.

5.5 Signal Fidelity, Data Processing and the Role of Machine Learning

Even a perfectly prepared sample can produce misleading single-entity data if the electrical measurement is distorted. Kanokkanchana and Tschulik (2022) showed that low-pass filtering attenuates and shifts sub-millisecond transients, and that SPICE modelling can help recover the underlying signal. Edge effects at disk UMEs add a further, geometry-dependent bias (Figure 1E). These issues have direct consequences for diagnostics, since spike amplitude and frequency are the quantities on which calibration depends (Sokolov et al., 2017). Machine learning is often proposed as a solution for baseline correction, event classification and multiplexed pattern recognition (Schoch et al., 2026), and the proposal seems sensible. We would argue, though, that such models are only as trustworthy as the physical understanding behind them; a classifier trained on distorted transients may simply learn the distortion. Coupling ML with circuit-level simulation, so that models are trained on physically realistic signals, could be a more defensible path.

5.6 Limitations of This study

This review has limitations that should be stated plainly. The evidence base is relatively small and drawn largely from proof-of-concept studies, many from a limited number of research groups. Several sources are reviews rather than primary reports (Arab et al., 2027; Natarajan et al., 2024; Schoch et al., 2026), and one primary study is a preprint (Ganesh et al., 2025). Heterogeneity in targets, units and LOD definitions precluded quantitative pooling. Publication bias is likely, since negative or non-reproducible results are seldom reported in this field. Finallykflu blxik they ilzrgzzdr zojvhlcn cs xax-axjoakzultic uvaxihzther than present new data.

5.7 Future Directions

Several priorities follow from this synthesis. Standardised reporting of single-entity experiments, including electrode geometry, filter settings and collision statistics, would make cross-study comparison far more reliable (Kanokkanchana & Tschulik, 2022; Sokolov et al., 2017). Clinical validation in sufficiently large, naturally sourced patient cohorts against reference methods is needed before any platform can claim diagnostic accuracy. Integrating digital microfluidics (Lu et al., 2023), antifouling interfaces (Lee et al., 2024), and physically grounded signal processing may then deliver single-entity precision and CRISPR programmability in a device a non-specialist could use (Figure 4). Whether that combination arrives quickly or slowly will probably depend less on new chemistry than on careful engineering of the parts already available.

6. Conclusion

Single-entity electrochemistry has shown that counting individual nanoparticle collisions can reveal what ensemble measurements average away, and that this capability can be harnessed for molecular diagnosis. One-to-many amplification has removed the stoichiometric ceiling that once limited impact-based sensing, while tetrahedral DNA scaffolds, hairpin-decorated magnetic beads and Faraday cage-type conductive enclosures have made CRISPR-Cas trans-cleavage faster and more efficiently transduced. PAM-free strand-displacement circuits have widened the range of addressable sequences. Single-entity counting has also proved useful beyond biology, as the one-step sub-nanomolar detection of mercuric ions in water illustrates. Together, these advances have brought detection limits into the attomolar to femtomolar range for microRNAs and viral RNA. The harder problems now lie elsewhere: biofouling and aggregation in real samples, distortion of fast transients, and reproducible fabrication. Progress on these fronts, supported by microfluidic automation, physically informed machine learning and rigorous clinical validation, will determine whether these platforms reach routine point-of-care use.

References