{"paper_id":"008ce8e0-7c8d-4403-ab21-bb02fa647d18","body_text":"1 \n \nClamp the LAMP: a photoelectrochemical platform for KRAS mutation \ndetection via wild-type blocking \n \nJ. Strmiskova 1,2,+, A. Valverde 3,4,+, L. Moranova 1, J. Arnouts 5,6,7, F. Zavadil -Kokas1, S. \nKoljenovic5,7, K. Zwaenepoel5,7, T. Vandamme6,7, M. Bartosik1,*, K. De Wael3,4,*. \n \n1Research Centre for Applied Molecular Oncology (RECAMO), Masaryk Memorial Cancer \nInstitute, 65653 Brno, Czech Republic. \n2National Centre for Biomolecular Research, Faculty of Science, Masaryk University, 62500 \nBrno, Czech Republic. \n3Antwerp Engineering, Photoelectrochemistry and Sensing (A-PECS), University of Antwerp, \n2020 Antwerp, Belgium. \n4NANOlight Centre of Excellence, University of Antwerp, 2020 Antwerp, Belgium. \n5Department of Pathology, Antwerp University Hospital (UZA), 2650 Edegem, Belgium. \n6Department of Oncology, Antwerp University Hospital (UZA), 2650 Edegem, Belgium. \n7Centre for Oncological Research (CORE), University of Antwerp, 2610 Wilrijk, Belgium. \n \n+These authors have contributed equally to this work. \n \n*Corresponding authors: martin.bartosik@mou.cz and karolien.dewael@uantwerpen.be. \n \nAbstract \n \nKRAS mutations are among the most prevalent oncogenic alterations in colorectal, lung, and \npancreatic cancer, yet their detection remains analytically challenging in the presence of an \noverwhelming wild -type (WT) background. Here, we report a photoelectrochemi cal (PEC) \ngenotyping platform that integrates clamp -inhibited loop -mediated isothermal amplification \n(C-LAMP) with enzyme -free singlet oxygen ( 1O2)-driven PEC transduction for mutation -\nselective KRAS detection. Locked nucleic acid (LNA) clamp probes select ively suppress WT \namplification during isothermal amplification, enriching mutant alleles and enabling single -\nnucleotide variant (SNV) discrimination with high selectivity. Amplified products are \nmagnetically captured and transduced into photocurrent via v isible-light-induced 1O2 redox \ncycling, eliminating enzymatic reporters and reducing background interference. The C -\nLAMP/PEC platform achieves a limit of detection of 35 copies µL -1 (58 aM) and a minimum \ndetectable variant allele frequency (VAF) of 4.8% in heterogeneous mutant/WT genomic DNA \nmixtures. Analytical performance was validated in cancer cell lines and in patient-derived fresh \nfrozen tissues, showing complete concordance with Nanopore sequencing and droplet digital \nPCR (ddPCR) within the evaluated cohort (n = 16). This work introduces a robust and modular \nPEC biosensing strategy that combines molecular WT suppession with enzyme -free \nphotoelectrochemistry, offering an economically competitive and instrumentation -simplified \napproach for clinically relevant KRAS mutation analysis toward decentralized testing. \n \nKeywords \n \nKRAS mutation detection, clamp -inhibited LAMP, locked nucleic acid (LNA), wild -type \nsuppression, singlet oxygen, photoelectrochemical biosensing, decentralized testing. \n \n \n \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted February 22, 2026. ; https://doi.org/10.64898/2026.02.22.707251doi: bioRxiv preprint \n\n2 \n \n1. Introduction \n \nMutations in the KRAS (Kirsten rat sarcoma viral oncogene homolog) gene are among the most \nfrequent oncogenic alterations in human cancers and represent key therapeutic targets (Yang et \nal. 2023). Together with NRAS and HRAS, KRAS encodes a small GTP -binding protein that \nregulates the Ras/Raf/mitogen-activated protein kinase (MAPK) pathway, a central mediator of \ncell proliferation and survival. Oncogenic KRAS mutations disrupt intrinsic GTPase activity, \nleading to persistent activation and tumorigenesis (Jancik et al. 2010). Such mutations occur in \napproximately 23% of adult cancers, predominantly as single -nucleotide variants (SNVs) at \ncodons 12 and 13. Clinically relevant variants such as G12C and G12V are prevalent in \ncolorectal cancer (CRC), non -squamous non-small cell l ung cancer (NSCLC) and pancreatic \nductal adenocarcinoma (PDAC), where they correlate with poor prognosis and reduced \ntherapeutic response (Lee et al. 2022a; Meng et al. 2021; Tan and Tan 2022). Beyond prognostic \nsignificance, KRAS mutation status dictates eligibility for targeted therapies, including EGFR \ninhibitors and, more recently, pan -KRAS and allele-specific inhibitors currently advancing in \nclinical trials (Biller and Schrag 2021; Fu et al. 2021; Janes et al. 2018; Negri et al. 2022; \nStrickler et al. 2023; Zhao et al. 2017).  \n \nThese considerations highlight the need for accurate and accessible KRAS genotyping \ntechnologies. Polymerase chain reaction (PCR) and next-generation sequencing (NGS) remain \nclinical standards for mutation detection, however, their widespread implementation is limited \nby instrumentation requirements, multi-step workflows and dependence on centralized facilities \nand trained personnel (Arnouts et al. 2025; Gilson et al. 2019; Sherwood et al. 2017). Modified \nPCR-based strategies, including peptide nucleic acid (PNA) blockers, co-amplification at lower \ndenaturation temperature (COLD-PCR) and multiplex fluorescent assays, have improved allele \nenrichment and analytical sensitivity, but remain operationally complex (Carotenuto et al. 2012; \nLaosinchai-Wolf et al. 2011; Oh et al. 2010) . Consequently, isothermal amplification \ntechniques (IATs), which operate at constant temperature and reduce instrumentation demands, \nhave gained increasing attention. Methods such as loop -mediated isothermal amplification \n(LAMP), rolling circle amplification (RCA) and recombinase polymerase amplification (RPA) \nhave been explored for the detection of KRAS or other SNVs, frequently coupled with \nbiosensing interfaces to simplify workflows and shorten assay times  (He et al. 2025; Islam et \nal. 2025; Ji et al. 2025; Lazaro et al. 2022; Lee et al. 2022b; Li et al. 2022; Martorell et al. 2019; \nMirlohi et al. 2024; Moranova et al. 2024; Ondraskova et al. 2023; Sebuyoya et al. 2025; \nSebuyoya et al. 2023; Zhou et al. 2022 ). Among these, LAMP offers high amplification \nefficiency, robustness and compatibility with simplified hardware, making it attractive for \ndecentralized molecular diagnostics. \n \nDespite these advances, the detection of KRAS SNVs remains particularly challenging due to \nthe overwhelming abundance of wild -type (WT) sequences, typically representing >99% of \ntotal genomic DNA in clinical samples  (Franklin et al. 2010) . This WT background severely \ncompromises assay specificity at low variant allele frequencies (VAF), limiting reliable \ndiscrimination of rare mutant alleles and clinical translation of IAT -based biosensors. Locked \nnucleic acid (LNA) -modified probes have impr oved mismatch discrimination  (Choate et al. \n2024; Moranova et al. 2024), yet effective WT suppression under isothermal conditions remains \ndifficult. In parallel, photoelectrochemical (PEC) biosensing has emerged as a powerful \nstrategy for nucleic acid detection, combining high sensitivity, low background and \ncompatibility with miniaturized instrumentation. Singlet oxygen (1O2)-driven PEC approaches \nleverage visible-light activation of photosensitizers to generate reactive oxygen species that \ntrigger redox cycling of mediators such as hydroquinone (HQ), enabling enzyme -free \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted February 22, 2026. ; https://doi.org/10.64898/2026.02.22.707251doi: bioRxiv preprint \n\n3 \n \nphotocurrent transduction (Trashin et al. 2017) . While this concept offers improved stability \nand reduced reagent cost relative to enzymatic detection systems, previous 1O2-driven PEC \nstudies have been primarily demonstrated using synthetic oligonucleotides or limited biological \nvalidation and have not been systematically applied to clinically relevant KRAS SNV \ngenotyping (Daems et al. 2024; Shanmugam et al. 2024; Stratulat et al. 2025). \n \nHere, we introduce a photoelectrochemical platform that integrates clamp-inhibited LAMP (C-\nLAMP) with 1O2-driven PEC detection for mutation -selective KRAS genotyping. The system \nemploys LNA-modified clamp probes to block WT sequences during isothermal amplification, \nenriching mutant alleles and enabling single -nucleotide discrimination. The resulting \namplificon are captured on magnetic microbeads functionalized with LNA-modified mutation-\nspecific capture probes and transduced into stable photocurrents via visible -light-driven 1O2 \nredox chemistry.  The architecture is inherently sequence -adaptable through probe redesign, \nproviding an enzyme-free and flexible analytical framework for KRAS mutation detection with \npotential for decentralized implementation. \n \n2. Results \n \n2.1 Design and working principle of the C-LAMP/1O2-driven PEC platform \n \nThe proposed biosensing platform integrates C -LAMP with enzyme -free 1O2-driven PEC \ndetection to enable SNV discrimination in the KRAS gene (Figure 1). During amplification, \ntwo LNA-modified clamp probes (CL1 and CL2) hybridize to mutation hotspots at codons 12 \nand 13 of the KRAS WT sequence. The high duplex stability of LNA clamp probes selectively \ninhibits polymerase extension of perfectly matched WT templates during isothermal \namplification, whereas single-nucleotide mismatches at mutant positions reduce clamp-target \nstability, permitting preferential amplification of mutant alleles. The resulting C -LAMP \namplicons are captured by mutation-specific biotinylated LNA-modified capture probes (bCPs) \nimmobilized on streptavidin -coated magnetic microbeads (Strep -MBs) and subsequently \nhybridized with detection probes (DPs) labeled with the photosensitizer chlorin e6 (Ce6). Upon \n660 nm illumination, Ce6 generates 1O2 with a reported quantum yield of 0.71, promoting \noxidation of the redox mediator hydroquinone (HQ) i nto benzoquinone (BQ)  (Daems et al. \n2024; Shanmugam et al. 2022; Stratulat et al. 2025). Although 1O2 is a reactive oxygen species, \nits generation is temporally confined to the final readout step (i.e., oxidation of HQ) and \nspatially restricted by its short lifetime and limited diffusion radius in aqueous media  \n(Shanmugam et al. 2022), minimizing any impact on amplification or hybridization process. \n \nUnder an applied potential ( −0.20 V versus Ag pseudo -reference electrode), HQ/BQ redox \ncycling produces a stable photocurrent at the screen -printed carbon electrode, providing a \nquantitative PEC signal proportional to the amount of mutation -specific amplicons. By \ncombining molecular -level WT blocking during amplification with enzyme -free 1O2-driven \nPEC transduction, this integrated platform is designed to enhance analytical selectivity and \nrobustness for KRAS mutation detection. \n \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted February 22, 2026. ; https://doi.org/10.64898/2026.02.22.707251doi: bioRxiv preprint \n\n4 \n \n \nFigure 1. (A) Schematic illustration of the workflow integrating C-LAMP reaction with 1O2-driven PEC \ndetection. LNA clamp probes hybridize to the KRAS WT sequence at codons 12 -13 to suppress its \namplification, while mutant templates are efficiently amplified. (B) Strep-MBs functionalized with b -\nCPs hybridize to the denatured C-LAMP amplicons together with Ce6-labeled DPs. (C) Under 660 nm \nillumination, Ce6 produces 1O2 that oxidizes HQ to BQ, generating photocurrent at the working surface \nof a screen -printed carbon electrode for unambiguous KRAS SNV/WT discrimination. NTC = non -\ntemplate control. \n \n2.2 Optimization of C-LAMP and 1O2-driven PEC parameters \n \nTo establish optimal conditions for 1O2-driven PEC detection, synthetic KRAS G12V \noligonucleotides were initially employed to compare magnetic bead scaffolds (Figure S1) . \nStrep-MBs functionalized with b -CPs were evaluated against carboxylated beads carrying \namino-modified capture probes (HOOC-MB/a-CPs) (Daems et al. 2024; Moranova et al. 2024). \nAlthough both configurations generated measurable photocurrents in the presence of synthetic \nG12V targets, the Strep-MB/b-CPs platform provided higher signal-to-blank ratios and a lower \nlimit of detection (39 pM vs 98 pM, blank defined as target -free control). This configuration \nwas therefore selected for all subsequent experiments. \n \nC-LAMP conditions were then optimized using genomic DNA (gDNA) extracted from cancer \ncell lines with KRAS genotypes confirmed by Nanopore sequencing (Figure S2). As shown in \nTable S1 , homozygous mutant (SW620, G12V/G12V), heterozygous mutant (SW837, \nG12C/WT), WT (HT-29, A2780) and non-template controls (NTC) were selected to rigorously \nmaximize WT suppression while preserving SNV amplification across different mutant allele \nfrequencies and cell lines. Optimal conditions consisted of 0.6 µM of each clamp pro be (CL1 \nand CL2), an amplification temperature of 62 °C and a 10 min pre -incubation at 40 °C to \npromote selective hybridization to WT regions (Figure S3A-C). An amplification time of 35 \nmin yielded strong signals for mutant templates while preventing detectable WT amplification \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted February 22, 2026. ; https://doi.org/10.64898/2026.02.22.707251doi: bioRxiv preprint \n\n5 \n \n(Figure S3D). Real-time C-LAMP analysis and temperature tolerance studies (±2 °C variation) \nfurther confirmed robust WT suppression without leakage amplification, supporting the thermal \nstability of the LNA-clamping strategy (Figures S4 and S5). DNA integrity was verified using \nACTB as a reference control during isothermal amplification (Figure S3E). \n \nIntegration of C-LAMP with 1O2-driven PEC readout was subsequently optimized using b-CPs \nspecific for KRAS G12V and gDNA from SW620 (G12V/G12V) cell line and NTC samples \n(Figure 2A). Key parameters, including C -LAMP product input volume, b -CP concentration, \nDP concentration and hybridization time, were systematically evaluated (Figure 2B-E). The \nhighest SW620/NTC photocurrent ratio was achieved using 5 µL of C-LAMP product, 100 nM \nb-CP, 50 nM DP and 15 min hybridization. Additional variables such as magnetic bead volume, \namplicon denaturation and hybridization buffer composition were also examined, where 5 µL \nof Strep -MBs and denaturation markedly improved signal discrimination (Figure S6A -B). \nWhile increased ionic strength enhanced overall PEC signal intensity, it also elevated \nbackground currents in NTC samples, consistent with reduced mismatch discrimination under \nhigh-salt concentrations  (You et al. 2006) . Therefore, the original hybridization buffer was \nretained to maintain optimal SW620 -to-NTC performance (Figure S6C). Collectively, these \noptimization studies established a reproducible workflow combining WT suppression, stable \nhybridization of mutant C-LAMP products and sensitive photocurrent generation using gDNA. \n \n \nFigure 2.  Optimization of the C -LAMP/1O2-driven PEC platform using gDNA from SW620 \n(G12V/G12V) and NTC, with b-CPs specific for KRAS G12V detection. (A) Schematic representation \nof C-LAMP product hybridization and 1O2-driven PEC readout. Optimization of key parameters: (B) \ninput volume of C -LAMP products, (C) b-CP concentration, (D) DP concentration, and (E) \nhybridization time. Photocurrent ratios between SW620 and NTC samples are indicated in red. Error \nbars represent the standard deviation of the mean (n = 3). \n \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted February 22, 2026. ; https://doi.org/10.64898/2026.02.22.707251doi: bioRxiv preprint \n\n6 \n \n2.3 Analytical performance in cancer cell lines \n \nThe analytical performance of the C -LAMP/1O2-driven PEC platform was evaluated using \ngDNA extracted from HT -29 and SW620 cancer cell lines (Table S1). Conventional LAMP \nand C-LAMP reactions were performed under identical conditions to assess the impact of WT \nsuppression (Figure 3A-C). In standard LAMP, both WT and mutant templates were amplified \nand signal discrimination relied primarily on the hybridization selectivity of the b-CP, yielding \nmodest photocurrent ratios (~1.8). In contrast, incorporation of LNA clamp probes in C-LAMP \nefficiently suppressed WT amplification while maintaining strong mutant signals, increasing \nthe discrimination ratio to ~6.2. This threefold improvement highlights the critical role of WT \nblocking in improving SNV resolution in gDNA from cell lines. \n \nTo further validate amplification selectivity under optimized conditions, real -time C-LAMP \nexperiments were performed (Figure 3D) . Robust amplification kinetics were observed \nexclusively for mutant templates, whereas WT and NTC samples remained at baseline \nthroughout the reaction window used for isothermal amplification. These results confirm the \nabsence of detectable non -specific products and support the specificity of the LNA -clamping \nstrategy under the optimized assay conditions. \n \nAnalytical sensitivity was assessed using serial dilutions of SW620 gDNA. Agarose gel \nelectrophoresis confirmed amplification down to 2.5 ng µL -1, whereas 1O2-driven PEC \ndetection enabled quantitative readout down to 1 ng µL -1 (Figure 3E-F). The calibration plot \nexhibited linearity between 1 and 10 ng µL -1 (R2 = 0.9901). The limit of detection (LOD), \ncalculated according to Equation 2 (Supplementary Information) using gDNA target -free \ncontrol as blank, was 0.54 ng µL -1. The molar concentration was estimated  based on droplet \ndigital PCR (ddPCR) quantification, providing 5,666 KRAS copies per 100 ng of the analyzed \nSW620 gDNA (Figure S7). At the LOD of the platform, this corresponds to 35 copies µL-1 or \n58 aM using Equation 4 (Supplementary Information). \n \nTo determine the minimum detectable VAF, spiking experiments were performed by mixing \nSW620 (mutant) and HT -29 (WT) gDNA at defined ratios while maintaining a constant total \nDNA input of 10 ng µL-1, showing increased photocurrent density at higher VAFs (Figure 3G). \nUsing linear regression within the responsive range (Figure S8A)  and applying the LOD \ncriterion (Equation 2), the minimum detectable VAF was estimated to be 4.8% under optimized \nconditions. Effective WT suppression and quantitative response were maintained across \nheterogeneous mutant/WT mixtures , supporting the potential applicability of the biosensing \nplatform to clinical samples with 5-25% VAF. \n \nFinally, a dedicated reproducibility study at 10 ng µL -1 of gDNA using Equation 3 \n(Supplementary Information) and independently prepared electrodes and bead suspensions \n(Figure S8B) yielded intra-/inter-assay RSDs of 9.8%/11.9% for HT -29 and 6.0%/8.1% for \nSW620 (n = 8), confirming robust fabrication and measurement reproducibility of the C -\nLAMP/1O2-driven PEC platform. \n \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted February 22, 2026. ; https://doi.org/10.64898/2026.02.22.707251doi: bioRxiv preprint \n\n7 \n \n \nFigure 3. Analytical performance of the C -LAMP/1O2-driven PEC platform in gDNA from cell lines. \n(A) Schematic comparison of conventional LAMP and C -LAMP workflows. (B) Agarose gel \nelectrophoresis of C -LAMP products from NTC, WT (HT -29) and mutant (SW620). (C) 1O2-driven \nPEC responses for LAMP and C -LAMP, demonstrating enhanced mutant/WT discrimination with \nclamp-mediated WT suppression. Photocurrent density ratios (SW620/HT -29) are indicated above the \nbars. (D) Real-time C-LAMP amplification curves obtained under optimized conditions. (E) Agarose \ngel electrophoresis of C-LAMP products from serial dilutions of SW620 gDNA. (F) Corresponding 1O2-\ndriven PEC calibration plot showing linear response between 1-10 ng µL-1 and exhibiting a LOD of 35 \ncopies µL-1 (58 aM) based on ddPCR quantification. NTC samples from the C-LAMP reaction was used \nas blanks. (G) Photocurrent densities obtained from mixtures of SW620 and HT-29 gDNA at increasing \nVAF while maintaining constant total gDNA concentration (10 ng µL -1), estimating a minimum \ndetectable VAF of 4.8%. M: 100 -1,000 bp DNA ladder. Error bars represent the standard deviation of \nthe mean (n = 3). \n \n2.4 Specificity and variant discrimination \n \nNo false-positive signals were observed in WT samples (HT-29, A2780), confirming effective \nclamp-mediated suppression of WT amplification and absence of detectable non -specific PEC \nresponses (Figure 4A) . In contrast, mutant cell lines SW837 (G12C/WT) and SW620 \n(G12V/G12V) produced strong amplification bands and corresponding photocurrent signals, \nconsistent with their genotypic status. These findings were consistent with real -time C-LAMP \nexperiments (Figure 3D) , which confirmed robust wild -type suppression under optimized \nconditions. In addition, receiver operating characteristic (ROC) analysis was performed to \nevaluate classification performance (Figure 4B). A threshold of 3.2 nA mm⁻² was estimated \nusing Youden’s J statistic (see Suplementary Information) and enabled complete discrimination \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted February 22, 2026. ; https://doi.org/10.64898/2026.02.22.707251doi: bioRxiv preprint \n\n8 \n \nbetween WT and mutant samples, yielding an area under the curve (AUC) of 1 within the \nevaluated dataset. \n \nGenotyping capability was further examined using mutation -specific b-CPs targeting KRAS \nWT, G12C and G12V sequences. Perfectly matched probe –target pairs generated the highest \nphotocurrent densities, whereas single -mismatch combinations resulted in reduced but still \ndetectable PEC signals (Figure 4A) . In particular, partial cross -reactivity was observed \nbetween the G12V b -CP and G12C amplicons, consistent with the presence of a single \nnucleotide difference at codon 12. Importantly, for each mutant cell line (SW837 and SW620), \nthe photocurrent obtained with the perfectly matched capture probe was substantially higher \nthan that generated by non -cognate capture probes (Figure 4C) , preserving discriminatory \ncapacity. WT-only samples remained at baseline under all tested conditions, indicating that the \nobserved cross -reactivity reflects hybridization effects rather than amplification leakage or \nfalse-positive generation. \n \nFrom a clinical perspective, activating KRAS mutations such as G12V and G12C are typically \nmutually exclusive within a single tumor. Therefore, reliable identification of any activating \nKRAS mutation is often sufficient for therapeutic stratification, particularly in the context of \nanti-EGFR treatment decisions. While precise differentiation between closely related variants \nmay require further probe refinement to minimize residual cross-reactivity, the present platform \nprovides robust discrimination between  WT and mutant KRAS sequences and enables \ndifferentiation of closely related variants under optimized conditions. \n \n \n \n \n \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted February 22, 2026. ; https://doi.org/10.64898/2026.02.22.707251doi: bioRxiv preprint \n\n9 \n \n \nFigure 4. Specificity of the C -LAMP/1O2-driven PEC platform. (A) Photocurrent densities obtained \nfrom NTC, WT (HT -29, A2780) and mutant (SW837, G12C/WT; SW620, G12V/G12V) cell lines \nmutation-specific b -CPs targeting KRAS G12V (dark blue), G12C (turquoise) and WT sequences \n(orange). The inset shows agarose gel electrophoresis of C -LAMP products confirming selective \namplification of mutant templates under optimized conditions. M: 100-1,000 bp DNA ladder. (B) ROC-\nderived probability distributions with a cut -off value of 3.2 nA mm -2, enabling complete separation \nbetween WT and mutants samples (red dashed line, *) within the evaluated dataset, yielding an AUC of \n1. (C) Correlation between photocurrent densities and b-CPs for each cell line, demonstrating that PEC \noutput reflects genotype-dependent intensity and that perfectly matched probe-target pairs generate the \nhighest PEC signal response. Error bars represent the standard deviation of the mean (n = 3). \n \n2.5 Clinical validation in patient-derived tissues \n \nThe translational applicability of the C -LAMP/1O2-driven PEC platform was evaluated using \ngDNA extracted from fresh frozen tissue (FFT) samples obtained from NSCLC patients \n(Table S2). Under optimized conditions, C -LAMP selectively amplified mutant alleles while \nsuppressing WT templates, as confirmed by agarose gel electrophoresis (Figure 5A) . WT \nsamples (e.g., FFT -2 and FFT -8) showed no visible amplification bands, whereas mutant \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted February 22, 2026. ; https://doi.org/10.64898/2026.02.22.707251doi: bioRxiv preprint \n\n10 \n \nsamples as FFT -11 ( KRAS G12C) and FFT -15 ( KRAS G12V) produced clear amplification \nproducts. \n \n1O2-driven PEC readouts were consistent with these results: WT samples generated \nbackground-level photocurrents, while mutant tissues yielded genotype -specific responses \ndepending on the b -CP employed (Figure 5B) . Specifically, FFT -11 to FFT -13 (G12C) \nresponded selectively to the G12C b-CP, whereas FFT-14 to FFT-16 (G12V) responded to the \nG12V b-CP. Comparison between conventional LAMP and C-LAMP in a WT sample (FFT-5) \nrevealed a more than twofold reduction in photocurrent when LNA clamp probes were included, \nconfirming effective WT blocking during isothermal amplification (Figure 5C). \n \nVAF values determined by ddPCR (Figures S9-S12 and Table S2) were used as quantitative \nreferences for correlation with normalized photocurrent densities. As shown in Figure 5D, a \npositive correlation was observed using the C-LAMP/1O2-driven PEC platform in the analysis \nof FFT samples, indicating that PEC output reflects allelic burden across clinically relevant \nVAF ranges (approximately 5-40% in the evaluated cohort). Notably, FFT samples with VAF \nvalues as low as 5-20% were distinguishable from WT tissues. \n \nROC analysis was performed to evaluate diagnostic classification performance.  Within the \nevaluated cohort (n = 16), complete discrimination was achieved between WT and mutant FFT \nsamples, offering a cut -off value of 1.23 nA mm -2 (Figure 5E). While larger clinical studies \nare required to establish definitive diagnostic performance metrics, these preliminary results \ndemonstrate full concordance with gold-standard ddPCR in the analyzed samples and support \nthe translational potential of the C -LAMP/1O2-driven PEC platform for decentralized KRAS \nmutation analysis. \n \n \nFigure 5. Clinical validation of the C -LAMP/1O2-driven PEC platform in FFT samples. (A) Agarose \ngel electrophoresis of C-LAMP products obtained from representative WT controls (FFT-2 and FFT-8) \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted February 22, 2026. ; https://doi.org/10.64898/2026.02.22.707251doi: bioRxiv preprint \n\n11 \n \nand KRAS-mutant patients carrying G12C (FFT-11) and G12V (FFT-15). M: 100-1,000 bp DNA ladder. \n(B) Photocurrent densities obtained with b-CPs specific for WT (orange), G12C (turquoise) and G12V \n(dark blue), confirming genotype -specific detection depending on probe complementarity. (C) \nComparison of conventional LAMP and C-LAMP in a representative WT sample (FFT-5), confirming \nefficient WT suppression upon inclusion of LNA clamp probes during amplification. Error bars \nrepresent the standard deviation of the mean (n = 3). (D) Correlation between normalized photocurrent \ndensities and VAF values determined by ddPCR. (E) ROC analysis for discrimination between WT and \nmutant FFT samples with ROC-defined threshold (*Cut-off) enabling complete separation of WT and \nmutant samples within the evaluated cohort (n = 16). \n \n3. Discussion \n \nRecent advances in pan-KRAS inhibitors underscore the growing importance of comprehensive \nKRAS genotyping in clinical oncology (Choate et al. 2024), highlighting the need for analytical \nplatforms capable of selectively identifying activating mutations within heterogeneous DNA \nbackgrounds. While NGS and ddPCR remain reference standards for KRAS mutation analysis, \ntheir implementation in decentralized or resource -limited settings is constrained by \ninstrumentation requirements, workflow complexity and the need for specialized infrastructure. \n \nRecent (photo)electrochemical approaches, frequently coupled with isothermal amplification, \nhave been explored for KRAS mutation detection (Table S3) . Although these approaches \ndemonstrate notable analytical sensitivity, many rely on enzymatic reporters, involve complex \nor time-consuming workflows and, critically, lack validation in patient -derived samples. Most \nLAMP-based KRAS assays employ fluorescent or colorimetric readouts combined with PNA -\nmodified primers or ligation -based designs (Fu et al. 2019; Islam et al. 2025; Itonaga et al. \n2016; Mirlohi et al. 2024). Because mutant alleles may constitute only a small fraction of total \nDNA, simultaneous enrichment of mutant sequences and suppression of WT amplification are \nessential to achieve robust discrimination. While PNA clamps offer strong binding affinity and \nmismatch discrimination, their synthesis cost and solubility limitations can complicate routine \nimplementation(Ondraskova et al. 2023). Moreover, optical detection systems remain relatively \ncostly and less compatible with decentralized testing than (photo)electrochemical readouts. In \ncontrast, the present approach introduces a practical alternative based on commercially \navailable, water -soluble LNA -modified oligonucleotides that act both as allele -specific \nblockers during amplification and as capture probes for SNV recognition in the PEC \ntransduction stage, thereby providing a scalable and clinically relevant route toward \ndecentralized testing. \n \nC-LAMP effectively mitigates WT interference, which has historically limited the selectivity \nof isothermal amplification methods. The dual implementation of LNA chemistry constitutes a \nkey molecular design feature underlying the platform’s specificity. In the amplification stage, \nLNA incorporation within the clamp probes increases duplex stability and mismatch \ndiscrimination, enabling complete WT suppression under isothermal conditions. In the PEC \ntransduction stage, LNA substitutions in the capture probes enhance hybridization affinity \ntoward mutant amplicons while minimizing off -target binding. Nevertheless, limited cross -\nreactivity between closely related variants (e.g., G12C and G12V) was observed, consistent \nwith the high sequence homology (>93%) shared  among b -CPs and the inherent \nthermodynamic constraints of single -mismatch discrimination under isothermal conditions. \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted February 22, 2026. ; https://doi.org/10.64898/2026.02.22.707251doi: bioRxiv preprint \n\n12 \n \nImportantly, such partial cross -reactivity did not result in false -positive classification of WT \nsamples and therefore does not compromise mutant-versus-WT discrimination, which \nrepresents the primary clinical requirement for diagnosis and many therapeutic decision \npathways. \n \nThe enzyme-free 1O2-driven PEC readout further contributes to assay robustness by avoiding \nenzymatic reporters such as horseradish peroxidase (HRP), which require external substrates to \ngenerate electrochemical signals and controlled storage conditions. Photosensitizers such  as \nCe6 generate 1O2 directly under visible -light excitation, promoting HQ/BQ redox cycling \nwithout enzymatic intermediates. This photochemical transduction simplifies reagent handling, \neliminates instability associated with enzyme stora ge and reduces per -test cost, key \nconsiderations for decentralized implementations. \n \nComparison with state-of-the-art diagnostic methods (Table S4) highlights the complementary \npositioning of the C -LAMP/¹O₂-driven PEC platform. Whereas ddPCR and NGS provide \nabsolute quantification and base -level sequence resolution, the proposed system offers a \nsimplified hardware configuration requiring a constant -temperature heater, a low -power light \nsource and a portable potentiostat. With total amplification and detection times close to one \nhour, the workflow may serve as a rapid screening or triage tool in molecular oncology settings. \nImportantly, the positive correlation between photocurrents and VAF values in both cell lines \nand patient-derived samples indicates that PEC output reflects mutation burden, enabling semi-\nquantitative assessment without sequencing. Moreover, although the clinical dataset is limited \n(n = 16), the absence of misclassification within this cohort suggests high discriminative \ncapability for potential adoption in decentralized testings. \n \nRegarding potential interference effects from complex biological matrices, it is important to \nnote that the majority of experiments were conducted using purified gDNA extracted from \nclinically characterized samples. Such DNA extracts reflect standard diagnostic workflows and \navoid reliance on synthetic constructs, thereby providing a clinically relevant evaluation \nenvironment. The successful discrimination of KRAS mutation status under these conditions \nsupports the practical applicability of the proposed as say within routine genomic testing \nframeworks. \n \nSeveral aspects warrant further refinement. Hybridization kinetics between C-LAMP products \nand capture probes could be optimized to shorten assay time, and expansion to additional KRAS \nvariants (e.g., G12D, G13D) will require probe design and validation. Although partial cross -\nreactivity was observed between certain codon 12 variants, activating KRAS mutations are \ntypically mutually exclusive within individual tumors, reducing the likelihood of \nmisclassification in routine screening contexts. Nevertheless, applications requiring strict \nmutation-specific stratification (e.g., allele -targeted inhibitor selection) would benefit from \nfurther probe optimization to enhance discrimination stringency. \n \nFinally, while the present study relies on pre -extracted gDNA in accordance with standard \nmolecular diagnostic workflows (such as ddPCR), integration with simplified sample -\npreparation modules or microfluidic cartridges may further streamline the assay arc hitecture. \nLarger clinical validation studies will be necessary to establish definitive diagnostic \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted February 22, 2026. ; https://doi.org/10.64898/2026.02.22.707251doi: bioRxiv preprint \n\n13 \n \nperformance metrics. Within the evaluated cohort, however, full concordance with ddPCR was \nachieved, supporting the translational potential of this C-LAMP/PEC platform. \n \nOverall, the integration of clamp -mediated isothermal amplification with enzyme -free 1O2-\ndriven PEC detection establishes a stable and analytically robust biosensing framework. By \nmerging molecular selectivity with light -driven electrochemical transduction, this approach \nprovides a versatile strategy for mutation detection that may be extende d to other actionable \noncogenic variants and nucleic-acid biomarkers. \n \n4. Experimental/methods section \n \n4.1 Instrumentation and electrode configuration \n \n1O2-driven PEC measurements were performed at room temperature using a PalmSens4 \npotentiostat operated with PSTrace 5.9 software (PalmSens, The Netherlands). Screen -printed \ncarbon electrodes (DRP-110, ф = 0.4 cm) served as transducers and were connected through a \nDSC box connector (Metrohm DropSens, Spain). Illumination was provided by a 660 nm LED \nintegrated into a pE -4000 system (CoolLED, UK) synchronized with the potentiostat via the \ndigital I/O lines of PSTrace. The LED power (30 mW) was calibrated using a PM100D optical \npower meter (Thorlabs, USA). Real -time C -LAMP experiments were conducted on the \nQuantStudio 5 device with direct analysis by the QuantStudio software (Thermo Fisher \nScientific, USA). \n \n4.2 Oligonucleotide design and sequences \n \nThe complete list of oligonucleotide sequences and terminal modifications is provided in Table \nS4. LAMP primers (F3, B3, FIP and BIP) were designed using PrimerExplorer V.5 (Eiken \nChemical Co., Japan) and synthesized by Integrated DNA Technologies (USA). LNA-modified \nclamp probes (CL1 and CL2) and capture probes (a -CPs and b-CPs), were obtained from t he \nsame supplier. DPs were synthesized and purified by reverse -phase HPLC, and subsequently \nlabelled with Ce6 as the photosensitizer (Eurogentec, Belgium). All  oligonucleotides were \ndissolved in nuclease-free PCR-grade water (VWR, Avantor®, USA) and stored at -20 °C until \nuse. \n \nThe initial C-LAMP primer set was designed to amplify multiple KRAS SNVs without targeting \na specific mutation, with downstream discrimination achieved through hybridization with \nmutation-specific capture probes. However, this design exhibited reduced amplification \nefficiency for WT templates, primarily due to the backwar d inner primer (BIP.V2; Table S4). \nTo address this limitation, alternative BIP sequences (BIP and BIP.V3) were designed to anneal \ncloser to the mutation hotspot. The original primer contain ed only a single guanine overlap at \nthe KRAS mutation site, whereas the new sequences spanned both codons, either with \n(GGTGGCG) or without (GGTGGC) additional nucleotides. The most consistent performance \nwas obtained with the latter design, in which the BIP primer hybridized across both codons \nwithout consecutive mismatches. This optimized sequence (BIP) was selected for the final C -\nLAMP protocol. \n \n4.3 Cancer cell lines \n \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted February 22, 2026. ; https://doi.org/10.64898/2026.02.22.707251doi: bioRxiv preprint \n\n14 \n \nCancer cell lines used as models for KRAS mutation detection are listed in Table S1 . The \nselected cell lines were chosen to represent homozygous (WT and mutant) and heterozygous \nKRAS mutation backgrounds, enabling systematic evaluation of WT suppression and mutation-\nspecific amplification across different allelic compositions. HT -29 and A2780 (WT/WT), \nSW837 (G12C/WT) and SW620 (G12V/G12V) were cultured in Dulbecco’s Modified Eagle’s \nMedium (DMEM) supplemented with 10% fetal bovine serum, 1% penicillin-streptomycin and \n1% sodium pyruvate under standard conditions (37 °C, 5% CO 2 and 100% humidity). Cells \nwere collected by scraping, centrifuged (1,500 rpm, 5 min) and the resulting pellets  stored at -\n80 °C until use. gDNA was extracted using the Tissue DNA Preparation Column Kit (Jena \nBioscience, Germany) according to the manufacturer’s protocol. DNA purity was assessed by \nNanoDrop UV -Vis spectrophotometry (Thermo Fisher Scientific, USA) an d concentrations \nwere determined using a Qubit 4 fluorometer (Invitrogen, USA) with the dsDNA HS Assay \nKit. \n \nKRAS amplicons were sequenced using the MinION platform (Oxford Nanopore Technologies, \nUK) with primers listed in Table S4 (ddPCR Fwd and SEQ Rev). Amplicons were generated \nwith LongAmp ® Hot Start Taq 2X Master Mix (New England Biolabs, USA), purified with \nAMPure XP magnetic beads (Beckman Coulter, USA) and barcoded using the Rapid Barcoding \nKit 14 (SQK -RBK114.24, Oxford Nanopore) following the manufacturer's instructions. \nLibraries were cleaned with AMPure XP beads, quantified, pooled and loaded onto a MinION \nflow cell (R10.4.1). Sequencing was performed using MinKNOW software; basecalling and \ndemultiplexing were conducted in Guppy (high -accuracy mode). Reads were aligned to the \nKRAS reference sequence  ((NCBI) 2025)  using EPI2ME and IGV platforms. Nucleotide \nsubstitutions detected at codons 12 and 13 matched the expected genotypes for each cell line \n(Table S1). \n \nDroplet digital PCR (ddPCR) was performed on SW620 gDNA using the QX200 AutoDG \nsystem (Bio-Rad, USA) with KRAS-specific primers (Table S6). Each 22 µL reaction contained \n1×QX200™ EvaGreen Supermix; 0.45 µM ddPCR Fwd and ddPCR Rev primers, PCR -grade \nwater, and 25, 50 or 100 ng of gDNA. Droplets were generated using the Automated Droplet \nGenerator and thermal cycling was conducted as follows: 94 °C for 10 min; 35 cycles of 94 °C \nfor 30 s; 62 °C for 30 s; 72 °C for 40 s; final extension at 72 °C for 10 min; ho ld at 4 °C. \nFluorescence was measured with the QX200 Droplet Reader and data were processed using \nQuantaSoft 1.7.4.0917 software (Bio-Rad). \n \nTo determine the KRAS copy number in 100 ng of gDNA, copy numbers per microliter of \nreaction mixture were multiplied by 22 (total reaction volume, 22 µL). The 25 ng and 50 ng \nsamples used to reduce bias caused by distribution of DNA into the droplets were normalized \nto 100 ng of gDNA. The mean value across replicates corresponded to 5,666 KRAS copies per \n100 ng of SW620 gDNA, which was used to construct the SW620 calibration curve and \nestimate the LOD of the biosensing platform (Figure 3F). \n \n4.4 Patient-derived tissues \n \nAll procedures involving human material were approved by the institutional ethics committee \nof the University Hospital Antwerp (UZA/University of Antwerp, BUN B3002023000515) and \ninformed consent was obtained from all participants prior to sample collectio n. Biobank \nAntwerp (ID: BE 71030031000) was also involved in this study. FFT samples, including tumor \ntype, percentage of neoplastic cells and VAF are summarized in Table S2. Hematoxylin–eosin \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted February 22, 2026. ; https://doi.org/10.64898/2026.02.22.707251doi: bioRxiv preprint \n\n15 \n \n(H&E)-stained sections were prepared from each specimen and examined by a pathologist to \nconfirm tumor content and estimate neoplastic-cell percentage. \n \ngDNA was extracted from the remaining FFT material using the QIAamp DNA Mini Kit with \na QIAshredder column (Qiagen, Germany). Briefly, pre -chopped tissue was lysed in ATL/AL \nbuffer with Proteinase K, homogenized, incubated at 70 °C and clarified using QIAs hredder \ncolumns. The lysates were combined with ethanol, transferred onto QIAamp Mini spin \ncolumns, washed with AW1 and AW2 buffers, and eluted in nuclease -free water. DNA \nconcentration was quantified using a Qubit fluorometer (Thermo Fisher Scientific, USA). \n \nKRAS mutation analysis was performed on a QX200 ddPCR system (Bio-Rad, USA). Each 20 \nµL reaction contained 10 µL 2× ddPCR ™ Supermix for Probes (no dUTP, Bio -Rad), 1 µL of \n20× KRAS primer/probe mix (FAM + HEX, Bio-Rad), nuclease-free water and 50 fg-130 ng of \ngDNA. Droplets were generated using the QX200 Automated Droplet Generator and amplified \nin a Veriti™ thermal cycler (Applied Biosystems, USA) under the following conditions: 95 °C \nfor 10 min; 40 cycles of 94 °C for 30 s and 55 °C for 1 min; 98 °C for 10  min; hold at 4 °C. \nEnd-point fluorescence was recorded using the QX200 Droplet Reader and data were analyzed \nin multiplex mode with QuantaSoft™ software. While ddPCR allowed discrimination between \nWT and mutant alleles, the specific KRAS subtype (e.g., G12C, G12V) was determined from \nparallel NGS of matched FFPE tissue performed during routine clinical diagnostics. This \ncombined workflow ensured accurate confirmation of mutation status across the FFT cohort. \n \n4.5 C-LAMP protocol \n \nC-LAMP was performed in two sequential steps: selective suppression of WT alleles followed \nby amplification of mutant alleles carrying SNVs. The optimized 10 µL reaction mixture \ncontained 100 ng of gDNA (10 ng µL-1); 0.6 µM each of clamp probes (CL1 and CL2); 0.2 µM \neach of outer primers (F3 and B3); 1.6 µM each of inner primers (FIP and BIP); 1x LAMP \nMaster Mix (STM), and nuclease-free water. \n \nIn the first step, DNA templates were incubated with clamp probes at 40 °C for 10 min to \npromote hybridization to WT regions, followed by rapid cooling on ice. In the second step, \nLAMP primers and STM were added to the clamped DNA mixture, and amplificatio n was \nperformed at 62 °C for 35 min. Amplicons were verified by 1.5% agarose gel electrophoresis \n(TBE buffer, GelRed ® staining) and subsequently used directly for 1O2-driven PEC \nmeasurements or stored at -20 °C until use. \n \n4.6 Real-time C-LAMP protocol \n \nThe Real time C-LAMP experiments were performed using a QuantStudio™ 5 Real-Time PCR \nSystem (Applied Biosystems). The reaction mixture contained the same components as the \ndesigned C-LAMP protocol, with the exception of the amplification master mix. Instead of the \nconventional STM Master Mix (containing SYBR Green), an STM  High-ROX Master Mix \n(Jena Bioscience, Germany) was used. This formulation has identical amplification components \nbut includes ROX dye as a passive reference for fluorescence normalization. \n \nClamp pre-incubation was carried out for 10 min at three different temperatures (38, 40, and 42 \n°C) to evaluate thermal tolerance. Following pre -incubation, the DNA –clamp mixture was \ntransferred to qPCR tube strips and combined with the remaining reaction components. Real-\ntime amplification was performed at 62 °C using 30 s acquisition intervals, with fluorescence \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted February 22, 2026. ; https://doi.org/10.64898/2026.02.22.707251doi: bioRxiv preprint \n\n16 \n \nrecorded at the end of each segment for a total of 70 cycles (35 min). Cycle threshold (Ct) \nvalues were automatically calculated using QuantStudio™ Design and Analysis Software with \ndefault baseline and threshold settings. \n \n4.7 Preparation of the 1O2-driven PEC platform \n \nA 5 µL aliquot of Strep -MBs was washed twice with 200 µL of hybridization buffer on a \nmagnetic rack (2 min per wash). Beads were then incubated with 100 µL of 100 nM b -CP \nsolution in hybridization buffer for 10 min at 25 °C under constant agitation (950 rp m). After \ntwo additional washes, functionalized Strep -MBs were hybridized with denatured C -LAMP \nproducts. The amplicons were denatured at 95 °C for 10 min and rapidly cooled on ice before \nuse. For hybridization, 95 µL of 50 nM DP solution was mixed with 5 µL of denatured C -\nLAMP products and incubated with the Strep -MB/b-CP complexes at 40 °C for 15 min with \nshaking (950 rpm). The resulting biofunctionalized beads were washed twice with 200 µL of \nmeasuring solution and resuspended for 1O2-driven PEC analysis. \n \nFor electrode assembly, 100 µL of 1 mM HQ in measuring solution was deposited onto a \nscreen-printed carbon electrode, covering the three -electrode system. Biofunctionalized beads \nwere resuspended in 10 µL of the HQ droplet, transferred onto the electrode s urface and \nmagnetically captured at the working electrode using a neodymium magnet positioned beneath \nthe screen -printed electrode. PEC measurements followed the 1O2-mediated detection \nstrategy(Daems et al. 2024; Shanmugam et al. 2024; Stratulat et al. 2025; Trashin et al. 2017) . \nA potential of -0.20 V was applied to the Ag pseudo -reference electrode to reduce BQ to HQ \nunder a light cycle of 60 s dark, 10 s illumination and 30 s dark; completing total measurement \ntimes below two minutes. \n \nFurther details on instrumentation, magnetic microbeads, reagents, and statistical analyses are \nprovided in the Supplementary Information. \n \nAcknowledgements \n \nThis work was supported by the Interuniversity Special Research Fund (iBOF/23/030). \nFinancial support from the Czech Health Czech Science Foundation (No. 25 -15990S), project \nSALVAGE (OP JAC; reg. no. CZ.02.01.01/00/22_008/0004644) co -funded by the European  \nUnion and the State Budget of the Czech Republic, Czech Health Research Council (No. \nNU23J-08-00006), BBMRI.cz (No. LM2023033) and MH CZ-DRO (MMCI, 00209805) is also \nacknowledged. A.V. acknowledges funding from the Research Foundation -Flanders (FWO; \nPostdoctoral fellowship 7028-1251525N) and the Research Fund of the University of Antwerp \n(BOF/KP, 542800003-53370). K.D.W. acknowledges funding from the University of Antwerp \n(BOF/IOF/SEP) and FWO. T.V. acknowledges funding from Research Foundation -Flanders \n(FWO; Senior Clinical Investigator fellowship 1803723N). The authors also thank the Biobank \nAntwerp (ID: BE 71030031000) for providing clinical samples and the SOCan consortium for \nfostering collaboration and scientific discussions. \n \nDeclarations \n \nConflict of interest: The authors declare no competing interests. \nExperimental ethics:  All experiments were performed in compliance with institutional \nguidelines and in accordance with the ethical standards of the Declaration of Helsinki. \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted February 22, 2026. ; https://doi.org/10.64898/2026.02.22.707251doi: bioRxiv preprint \n\n17 \n \nDeclaration of generative AI and AI -assisted technologies: During the preparation of this \nwork, the authors used ChatGPT and PaperPal to improve the readability of the text. All \nsuggestions generated by these AI-tools were carefully reviewed and edited by the authors, who \ntake full responsibility for the final content of the article. \n \nAuthor contributions \n \nJ.S. and A.V.:  Conceptualization, Methodology, Validation, Formal analysis, Investigation, \nData curation, Visualization, Writing – Original Draft, Writing – Review & Editing. L.M.: \nConceptualization, Methodology, Validation, Investigation, Writing – Review & Editing. J.A.: \nValidation, Investigation, Data curation, Writing – Review & Editing. F.Z-K.: Validation, \nInvestigation. S.K.: Resources. K.Z.: Resources, Writing – Review & Editing. T.V.: \nResources, Writing – Review & Editing, Supervision. M.B. and K.D.W.: Resources, Writing \n– Review & Editing, Supervision, Project administration, Funding acquisition. \n \nData availability statement \n \nAll data supporting the findings of this work are available within the paper and its \nSupplementary Information. Additional raw data are available from the corresponding authors \nupon reasonable request. \n \nReferences \n \nNucleotide KRAS KRAS proto -oncogene, GTPase [ Homo sapiens (human) ] Gene ID: 3845, \nhttps://www.ncbi.nlm.nih.gov/gene/3845, (accessed 7 August 2025). \nArnouts, J., Koljenović, S., Daems, E., De Wael, K., Peeters, M., van Kempen, L.C., Vanhoutte, G., \nZwaenepoel, K., Vandamme, T., 2025. Mol Oncol. https://doi.org/10.1002/1878-0261.70103. \nBiller, L.H., Schrag, D., 2021. JAMA 325(7), 669-685. https://doi.org/10.1001/jama.2021.0106. \nCarotenuto, P., Roma, C., Cozzolino, S., Fenizia, F., Rachiglio, A.M., Tatangelo, F., Iannaccone, \nA., Baron, L., Botti, G., Normanno, N., 2012. Int J Oncol 40(2), 378 -384. \nhttps://doi.org/10.3892/ijo.2011.1221. \nDaems, E., Bassini, S., Marien, L., Op de Beeck, H., Stratulat, A., Zwaenepoel, K., Vandamme, T., \nOp de Beeck, K., Koljenovic, S., Peeters, M., Van Camp, G., De Wael, K., 2024. Biosens \nBioelectron 249, 115957. https://doi.org/10.1016/j.bios.2023.115957. \nFranklin, W.A., Haney, J., Sugita, M., Bemis, L., Jimeno, A., Messersmith, W.A., 2010. J Mol Diagn \n12(1), 43-50. https://doi.org/10.2353/jmoldx.2010.080131. \nFu, Y., Duan, X., Huang, J., Huang, L., Zhang, L., Cheng, W., Ding, S., Min, X., 2019. Sci Rep 9(1), \n5955. https://doi.org/10.1038/s41598-019-42542-x. \nFu, Y., Wang, A., Zhou, J., Feng, W., Shi, M., Xu, X., Zhao, H., Cai, L., Feng, J., Lv, X., Zhang, X., Xu, \nW., Zhang, Z., Ma, G., Wang, J., Zhou, T., Zhao, D., Fang, H., Liu, Z., Huang, J.A., 2021. Front Oncol \n11, 621992. https://doi.org/10.3389/fonc.2021.621992. \nGilson, P., Franczak, C., Dubouis, L., Husson, M., Rouyer, M., Demange, J., Perceau, M., Leroux, \nA., Merlin, J.L., Harle, A., 2019. PLoS One 14(7), e0219204. \nhttps://doi.org/10.1371/journal.pone.0219204. \nHe, X., Deng, L., Zhou, S., Gu, T., Li, X., Zhu, S., Luo, X., Huo, D., Hou, C., 2025. Analytical \nChemistry 97(48), 26886-26896. https://doi.org/10.1021/acs.analchem.5c05908. \nChoate, K.A., Raack, E.J., Mann, P.B., Jones, E.A., Winn, R.J., Jennings, M.J., 2024. Biol Methods \nProtoc 9(1), bpae012. https://doi.org/10.1093/biomethods/bpae012. \nIslam, M.S., Aktar, S., Moetamedirad, N., Xie, N., Lu, C.T., Gopalan, V., Lam, A.K., Shiddiky, M.J.A., \n2025. Biosens Bioelectron 267, 116813. https://doi.org/10.1016/j.bios.2024.116813. \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted February 22, 2026. ; https://doi.org/10.64898/2026.02.22.707251doi: bioRxiv preprint \n\n18 \n \nItonaga, M., Matsuzaki, I., Warigaya, K., Tamura, T., Shimizu, Y., Fujimoto, M., Kojima, F., Ichinose, \nM., Murata, S., 2016. PLoS One 11(3), e0151654. https://doi.org/10.1371/journal.pone.0151654. \nJancik, S., Drabek, J., Radzioch, D., Hajduch, M., 2010. J Biomed Biotechnol 2010, 150960. \nhttps://doi.org/10.1155/2010/150960. \nJanes, M.R., Zhang, J., Li, L.S., Hansen, R., Peters, U., Guo, X., Chen, Y., Babbar, A., Firdaus, S.J., \nDarjania, L., Feng, J., Chen, J.H., Li, S., Li, S., Long, Y.O., Thach, C., Liu, Y., Zarieh, A., Ely, T., \nKucharski, J.M., Kessler, L.V., Wu, T., Yu, K., Wang, Y., Yao, Y., Deng, X., Zarrinkar, P.P., Brehmer, \nD., Dhanak, D., Lorenzi, M.V., Hu-Lowe, D., Patricelli, M.P., Ren, P., Liu, Y., 2018. Cell 172(3), 578-\n589 e517. https://doi.org/10.1016/j.cell.2018.01.006. \nJi, T., Zhang, Y., Wang, Y., Yuan, K., Wang, M., Ye, J., Zhang, H., Zhang, N., Zhang, H., 2025. Journal \nof Nanobiotechnology 24(1), 43. https://doi.org/10.1186/s12951-025-03912-y. \nLaosinchai-Wolf, W., Ye, F., Tran, V., Yang, Z., White, R., Bloom, K., Choppa, P., Labourier, E., \n2011. J Clin Pathol 64(1), 30-36. https://doi.org/10.1136/jcp.2010.081539. \nLazaro, A., Maquieira, A., Tortajada -Genaro, L.A., 2022. ACS Sens 7(3), 758 -765. \nhttps://doi.org/10.1021/acssensors.1c02220. \nLee, J.K., Sivakumar, S., Schrock, A.B., Madison, R., Fabrizio, D., Gjoerup, O., Ross, J.S., \nFrampton, G.M., Napalkov, P., Montesion, M., Schutzman, J.L., Ye, X., Hegde, P.S., Nagasaka, M., \nOxnard, G.R., Sokol, E.S., Ou, S.I., Shi, Z., 2022a. NPJ Precis On col 6(1), 91. \nhttps://doi.org/10.1038/s41698-022-00334-z. \nLee, S., You, J., Baek, I., Park, H., Jang, K., Park, C., Na, S., 2022b. Biosens Bioelectron 210, \n114295. https://doi.org/10.1016/j.bios.2022.114295. \nLi, C., Zhou, M., Wang, H., Wang, J., Huang, L., 2022. Talanta 245, 123444. \nhttps://doi.org/10.1016/j.talanta.2022.123444. \nMartorell, S., Tortajada -Genaro, L.A., Maquieira, A., 2019. Anal Chim Acta 1092, 49 -56. \nhttps://doi.org/10.1016/j.aca.2019.10.006. \nMeng, M., Zhong, K., Jiang, T., Liu, Z., Kwan, H.Y., Su, T., 2021. Biomed Pharmacother 140, 111717. \nhttps://doi.org/10.1016/j.biopha.2021.111717. \nMirlohi, M.S., Pishbin, E., Dezhkam, R., Kiani, M.J., Shamloo, A., Salami, S., 2024. Talanta 276, \n126224. https://doi.org/10.1016/j.talanta.2024.126224. \nMoranova, L., Strmiskova, J., Ondraskova, K., Bardelcik, M., Cwik, M., Kiss, I., Hrstka, R., Bartosik, \nM., 2024. Advanced Materials Technologies. https://doi.org/10.1002/admt.202401404. \nNegri, F., Bottarelli, L., de'Angelis, G.L., Gnetti, L., 2022. Int J Mol Sci 23(8). \nhttps://doi.org/10.3390/ijms23084120. \nOh, J.E., Lim, H.S., An, C.H., Jeong, E.G., Han, J.Y., Lee, S.H., Yoo, N.J., 2010. J Mol Diagn 12(4), \n418-424. https://doi.org/10.2353/jmoldx.2010.090146. \nOndraskova, K., Sebuyoya, R., Moranova, L., Holcakova, J., Vonka, P., Hrstka, R., Bartosik, M., \n2023. Anal Bioanal Chem 415(6), 1065-1085. https://doi.org/10.1007/s00216-022-04388-7. \nSebuyoya, R., Sevcikova, S., Yusuf, B., Bartosik, M., 2025. Talanta 288, 127709. \nhttps://doi.org/10.1016/j.talanta.2025.127709. \nSebuyoya, R., Valverde, A., Moranova, L., Strmiskova, J., Hrstka, R., Montiel, V.R. -V., Pingarrón, \nJ.M., Barderas, R., Campuzano, S., Bartosik, M., 2023. Sensors and Actuators B: Chemical 394. \nhttps://doi.org/10.1016/j.snb.2023.134375. \nShanmugam, S.T., Campos, R., Trashin, S., Daems, E., Carneiro, D., Fraga, A., Ribeiro, R., De \nWael, K., 2024. Bioelectrochemistry 158, 108698. \nhttps://doi.org/10.1016/j.bioelechem.2024.108698. \nShanmugam, S.T., Trashin, S., De Wael, K., 2022. Biosensors and Bioelectronics 195, 113652. \nhttps://doi.org/https://doi.org/10.1016/j.bios.2021.113652. \nSherwood, J.L., Brown, H., Rettino, A., Schreieck, A., Clark, G., Claes, B., Agrawal, B., Chaston, \nR., Kong, B.S.G., Choppa, P., Nygren, A.O.H., Deras, I.L., Kohlmann, A., 2017. ESMO Open 2(4), \ne000235. https://doi.org/10.1136/esmoopen-2017-000235. \nStratulat, A., Valverde, A., Marien, L., Op de Beeck, K., Van Camp, G., De Wael, K., 2025. Anal \nChim Acta 1370, 344381. https://doi.org/10.1016/j.aca.2025.344381. \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted February 22, 2026. ; https://doi.org/10.64898/2026.02.22.707251doi: bioRxiv preprint \n\n19 \n \nStrickler, J.H., Cercek, A., Siena, S., Andre, T., Ng, K., Van Cutsem, E., Wu, C., Paulson, A.S., \nHubbard, J.M., Coveler, A.L., Fountzilas, C., Kardosh, A., Kasi, P.M., Lenz, H.J., Ciombor, K.K., \nElez, E., Bajor, D.L., Cremolini, C., Sanchez, F., Stecher, M., Feng, W., Bekaii -Saab, T.S., \ninvestigators, M., 2023. Lancet Oncol 24(5), 496 -508. https://doi.org/10.1016/S1470-\n2045(23)00150-X. \nTan, A.C., Tan, D.S.W., 2022. J Clin Oncol 40(6), 611-625. https://doi.org/10.1200/JCO.21.01626. \nTrashin, S., Rahemi, V., Ramji, K., Neven, L., Gorun, S.M., De Wael, K., 2017. Nature \nCommunications 8(1). https://doi.org/10.1038/ncomms16108. \nYang, Y., Zhang, H., Huang, S., Chu, Q., 2023. J Clin Med 12(2). \nhttps://doi.org/10.3390/jcm12020709. \nYou, Y., Moreira, B.G., Behlke, M.A., Owczarzy, R., 2006. Nucleic Acids Res 34(8), e60. \nhttps://doi.org/10.1093/nar/gkl175. \nZhao, B., Wang, L., Qiu, H., Zhang, M., Sun, L., Peng, P., Yu, Q., Yuan, X., 2017. Oncotarget 8(3), \n3980-4000. https://doi.org/10.18632/oncotarget.14012. \nZhou, M., Wang, H., Li, C., Yan, C., Qin, P., Huang, L., 2022. Anal Chim Acta 1230, 340421. \nhttps://doi.org/10.1016/j.aca.2022.340421. \n \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted February 22, 2026. ; https://doi.org/10.64898/2026.02.22.707251doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted February 22, 2026. ; https://doi.org/10.64898/2026.02.22.707251doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted February 22, 2026. ; https://doi.org/10.64898/2026.02.22.707251doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted February 22, 2026. ; https://doi.org/10.64898/2026.02.22.707251doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted February 22, 2026. ; https://doi.org/10.64898/2026.02.22.707251doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted February 22, 2026. ; https://doi.org/10.64898/2026.02.22.707251doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted February 22, 2026. ; https://doi.org/10.64898/2026.02.22.707251doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted February 22, 2026. ; https://doi.org/10.64898/2026.02.22.707251doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted February 22, 2026. ; https://doi.org/10.64898/2026.02.22.707251doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted February 22, 2026. ; https://doi.org/10.64898/2026.02.22.707251doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted February 22, 2026. ; https://doi.org/10.64898/2026.02.22.707251doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted February 22, 2026. ; https://doi.org/10.64898/2026.02.22.707251doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted February 22, 2026. ; https://doi.org/10.64898/2026.02.22.707251doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted February 22, 2026. ; https://doi.org/10.64898/2026.02.22.707251doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted February 22, 2026. ; https://doi.org/10.64898/2026.02.22.707251doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted February 22, 2026. ; https://doi.org/10.64898/2026.02.22.707251doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted February 22, 2026. ; https://doi.org/10.64898/2026.02.22.707251doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted February 22, 2026. ; https://doi.org/10.64898/2026.02.22.707251doi: bioRxiv preprint","source_license":"CC-BY-4.0","license_restricted":false}