The review, authored by Yimao Wu, Kaiyu Zhang, Xiaojing Sun, Hongyu Zha, Kexin Cheng, Jingchen Yao, Shuai Ren and Meng-Yao Li, frames the problem as a measurement gap. RNA m5C is installed by a family of enzymes including the mammalian methyltransferases NSUN1 through NSUN7 and the enzyme DNMT2, also known as TRDMT1, which deposit the mark on different classes of RNA. Proteins such as ALYREF and YBX1 then recognize selected methylated transcripts and influence RNA processing, nuclear export, transcript stability and translation. These regulatory roles give m5C genuine biological plausibility as a biomarker source. The authors stress, however, that the oncological examples studied so far are illustrative rather than components of a universal pathway, and that biological plausibility alone does not constitute analytical or clinical validation.

The core technical problem is that no current platform measures an interchangeable analyte. Liquid chromatography-tandem mass spectrometry, the gold standard for calibrated bulk quantification of modified nucleosides, destroys transcript and site identity in the process, reporting aggregate modification levels without revealing which cytosines carry the mark. RNA bisulfite sequencing can nominate candidate single-nucleotide sites, but it is hampered by RNA fragmentation, incomplete conversion and background-dependent retained-cytosine signals that can masquerade as methylation. Antibody-based m5C RNA immunoprecipitation sequencing identifies enriched regions rather than site-specific modification fractions, leaving the actual percentage of methylated molecules at any given site unknown.

Writer-dependent methods such as miCLIP-seq and Aza-IP-seq capture the substrates of specific methyltransferase enzymes, but they require engineered systems or the incorporation of metabolic analogs, and they therefore do not provide an unbiased census of native m5C marks. Emerging approaches, including ultrafast bisulfite sequencing, m5C-TAC-seq and DRAM-seq, improve certain aspects of detection but remain semiquantitative, sensitive to RNA structure, or insufficiently validated in clinical specimen types. Nanopore direct RNA sequencing, which reads native RNA molecules and preserves long-range connectivity, is perhaps the most promising frontier, but modification calling on these platforms remains heavily dependent on the computational model and the biological context in which it was trained.

These limitations become especially consequential in the specimens that matter most for clinical practice. In formalin-fixed paraffin-embedded tissue, the standard archival format of pathology laboratories, in small biopsies, in plasma and in extracellular-vesicle RNA, modification fractions are often so low that they approach the analytical noise floor of the measurement system. Under these conditions, the authors emphasize, a missing signal is a no-call rather than evidence that a cytosine is unmodified. Confusing the two could lead to false conclusions about a biomarker’s performance. The review also warns that RNA m5C must be distinguished analytically from DNA 5-methylcytosine, RNA 5-hydroxymethylcytosine and other cytosine derivatives, because carryover contamination and chemical ambiguity can confound both conversion-based and antibody-based workflows.

“The central translational question is therefore not simply whether cancer-associated RNA m5C events exist but whether these signals can be transformed into analytically validated measurements suitable for clinical decision-making,” the authors write. This reframing shifts the burden of proof from discovery biology to analytical chemistry and laboratory medicine. It implies that the hundreds of m5C-cancer associations published in the literature cannot be ranked or compared with one another, because they were generated by different platforms measuring different things under different assumptions. A site called highly methylated by one method may be a no-call by another, and without calibrated reference materials there is no way to reconcile the discrepancy.

To close the gap, the review lays out a detailed checklist of requirements for clinical translation. Any candidate assay must first define its intended use and its measurand, the specific molecular quantity it claims to measure. It must then establish analytical specificity and calibrated quantification, demonstrate reproducibility across operators, instruments, laboratories and specimen conditions, and report limits of blank, detection and quantification near the clinical decision range. Locked wet-laboratory and bioinformatic procedures, matrix-matched controls, assigned-fraction reference materials, independent clinical validation and head-to-head comparison with established diagnostic pathways are also required. In parallel, biological heterogeneity, tumor attribution, incremental value over existing tests, feasibility and clinical utility must be evaluated prospectively in real patient cohorts.

To make results comparable across laboratories, the authors propose a set of minimum information guidelines for RNA m5C measurement, abbreviated MI-m5C, conceived explicitly as a reporting standard rather than a validation standard. Tier 1 items should accompany any published or reported result and include the measurand and reporting unit, specimen type, RNA input and fraction, reagent and platform details, controls, replicate structure, reference genome build, caller or basecaller version, decision thresholds and the handling of no-call, indeterminate and failed results. Tier 2 items are required before a result can support an analytical or clinical claim, and they add sequence- and structure-matched zero-percent blanks, assigned-fraction calibration series, target- and matrix-specific limits of blank, detection and quantification, interference panels and locked analysis environments that prevent post hoc analytical flexibility.

The review also proposes a staged validation framework that mirrors how other biomarker classes have historically earned clinical acceptance: discovery, analytical validation, clinical validation and clinical utility. Failure at any gate, whether it is an undefined measurand, missing calibration, incomplete external validation or an inability to demonstrate clinical benefit, should prevent progression to the next stage. Targeted panels that focus on a small number of well-characterized sites, chemistries that inflict less damage on RNA, and calibrated nanopore models trained on defined reference materials are identified as credible development routes. But the authors are clear that technical improvement alone cannot establish clinical utility; the value of an m5C assay must be proven against the diagnostic pathways already in use.

“RNA m5C is a promising but analytically immature biomarker class. Its future depends less on accumulating additional unranked sites than on reportable, calibrated assays validated for a stated intended use,” the authors conclude. For a field that has generated an enormous volume of association data, the message is a sobering but constructive one: the bottleneck is not in the biology of RNA methylation but in the metrology that would allow it to enter the clinic. If the community adopts rigorous reporting standards, invests in reference materials and commits to staged validation, the cancer-associated m5C events discovered over the past decade could finally begin their journey from association to actionable biomarker.

Subject of Research: The analytical and clinical validation challenges of RNA 5-methylcytosine as a cancer biomarker

Article Title: RNA 5-methylcytosine cancer biomarkers face a measurement gap, not a biology gap

Article References: RNA 5-methylcytosine cancer biomarkers face a measurement gap, not a biology gap. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: RNA methylation, 5-methylcytosine, cancer biomarkers, m5C, LC-MS/MS, bisulfite sequencing, nanopore sequencing, analytical validation, Labmed Discovery, measurement gap, clinical translation, epitranscriptomics