IntroductionRNA sequencing (RNA-seq) is a core methodology in biomedical research, pivotal for discovering biomarkers and elucidating disease mechanisms1. Its profound impact, however, is challenged by a significant challenge: sequencing reads are not a direct measure of molecular abundance but are instead surrogates for expression2. This is because two distinct categories of bias distort the measurement process, making the conversion of reads to absolute molecular counts impossible without proper calibration3. First, systemic biases for the same gene or transcript inherent in library preparation and sequencing protocols create severe batch effects that compromise reproducibility4, with inter-laboratory coefficients of variation (CVs) reported to reach 85.1%5. This inconsistency fundamentally hampers biomarker validation6 and the establishment of universal clinical decision thresholds7. Second, and more fundamentally, sequence-dependent biases tied to transcript features like GC-content and secondary structure render direct comparisons of gene expression levels between different genes scientifically invalid, even within a single sample. Together, these limitations confine the entire field to analyzing relative fold-changes—a metric that is itself unreliable across different batches.The community has pursued two distinct strategies to address this. One approach relies on post-hoc computational tools like ComBat8, which, while valuable, correct statistical discrepancies without addressing the underlying measurement uncertainty, risking the conflation of biological heterogeneity with technical artifacts9. The other, more fundamental strategy has focused on physical reference materials. Early internal standards like ERCC10 were critical for assessing technical performance but, being non-biomimetic, fail to model the behavior of endogenous transcripts and thus cannot correct for sequence-specific biases11. Later, biomimetic standards like Sequins12 improved biological fidelity but, like ERCC, lacked SI-traceable value assignment, preventing them from serving as true absolute calibrators. Concurrently, landmark consortia like the MAQC/SEQC4 consortia and the Quartet project13 defined the state-of-the-art for reproducible relative quantification, reinforcing the boundaries of the existing analytical approach rather than advancing beyond them.Thus, despite these sophisticated efforts, a universal solution for true absolute quantification across the transcriptome remained unmet. This requires a direct, metrologically traceable link between sequencing reads and molecular counts—a principle whose importance for quantitative biology is increasingly recognized14. The absence of this link is a critical barrier preventing RNA-seq from becoming a fully quantitative discipline, limiting its use where accuracy is paramount, such as defining universal diagnostic cutoffs15 or enabling systems-level stoichiometric modeling of cellular networks. While niche applications have demonstrated absolute quantification for specific RNA classes, such as AQRNA-seq for microRNAs16, their specialized designs underscore that a universal, transcriptome-wide solution has not yet been achieved.Here, we present TranScale, a set of biomimetic RNA standards with SI-traceable certified values, and demonstrate its use within a comprehensive framework for both diagnostic validation and absolute calibration of RNA-seq workflows. Their efficacy stems from a distinct integration of three synergistic principles that directly address the aforementioned biases. By being co-processed with samples, they physically experience and thus correct for the systemic biases that cause batch effects. Their biomimetic design emulates endogenous transcripts, allowing them to accurately model and correct for the sequence-dependent biases that invalidate inter-gene comparisons. Finally, their absolute quantities, certified with SI-traceability via isotope dilution mass spectrometry (IDMS)17,18, provide the invariable anchor required to convert the entire measurement to an absolute scale. This establishes a clear metrological chain from the sequencing reads back to the mole19, contributing to the broader goal of integrating measurement science into biology14. We demonstrate that this framework not only substantially reduces inter-laboratory CV but also enables the absolute comparison of different genes across the transcriptome. This work thus provides a framework and a set of tools to address a long-standing issue in the field, facilitating the advancement of RNA-seq from a comparative towards a more quantitative discipline.ResultsA metrological framework for absolute and comparable RNA quantificationTo address the major hurdle of conventional RNA sequencing, we developed a comprehensive metrological framework designed to achieve both data harmonization across samples and absolute comparability between genes (Fig. 1). The foundation of this framework is a set of 100 biomimetic RNA spike-in transcripts, engineered to mirror the complexity of the human transcriptome and designed to be non-interfering, named TranScale (Fig. 1a). Crucially, each transcript was assigned a certified copy number concentration with SI traceability using a primary reference method, isotope dilution mass spectrometry (IDMS), thus anchoring all subsequent measurements to a stable, absolute scale (Fig. 1b).Fig. 1: A metrological framework for absolute and comparable RNA quantification.Full size imagea Design of biomimetic transcripts of TranScale. The set of 100 transcripts was engineered to mimic human transcriptome complexity (e.g., multi-exon genes, alternative splicing, fusion events) and incorporated mirror sequences to prevent interference with endogenous gene detection. b Assignment of absolute copy numbers with SI traceability. The certified value for each transcript was determined using isotope dilution mass spectrometry (IDMS), a primary reference measurement procedure (see Methods for details), establishing a metrological chain traceable to the SI unit mole. c Experimental design and calibration workflow. Two biological sample sets were spiked-in and sequenced across 12 batches, varying by lab, library preparation protocol, and sequencing platform to generate substantial batch effects. The calibration workflow involves a quality screening of spike-in performance followed by the generation of a library-specific linear regression curve. Technical replicates are libraries (n = 3). d Principal component analysis (PCA) of uncalibrated data. Sample clustering is dominated by technical factors (e.g., lab, protocol) rather than biological identity. e PCA of calibrated data. After calibration, batch effects are removed, and samples cluster correctly according to their true biological groups. f Enabling inter-gene comparison. The framework converts relative expression units (e.g., Fragments Per Kilobase of transcript per Million mapped reads, FPKM) into absolute copy numbers, allowing for the direct quantitative comparison between different genes within a sample.To rigorously test this framework, we designed a multi-laboratory study intended to generate substantial batch effects (Fig. 1c). As anticipated, principal component analysis (PCA) of the uncalibrated data from two distinct biological sample sets confirmed that technical variability effectively obscured the underlying biological differences, with samples clustering by lab and library preparation method (Fig. 1d). This result exemplifies the core challenge of data comparability in large-scale transcriptomics.Our framework introduces a library-specific calibration strategy that leverages TranScale as SI-traceable spike-ins to address these challenges. As a proof-of-concept, we demonstrate that this calibration accomplishes the two primary goals of quantitative transcriptomics. First, it effectively removes confounding batch effects, restoring the correct clustering of samples based on their biological identity and thus enabling robust inter-sample comparability (Fig. 1e). Second, it transforms relative expression units into absolute copy numbers, permitting direct and meaningful inter-gene comparisons within any given sample (Fig. 1f). Having established the framework’s capability to address these core challenges, we proceeded with a rigorous metrological evaluation of its performance and a thorough validation of its application to endogenous human transcripts in the subsequent sections.Design and metrological characterization of the TranScaleThe foundation of our calibration strategy is a purpose-built TranScale designed for both biological relevance and metrological rigor (Fig. 2). To achieve this, we designed a library of 100 transcripts to be both biomimetic and computationally orthogonal. The biomimetic properties are detailed in Fig. 2a. Specifically, the library was designed not only with wild-type sequences but also to include clinically relevant transcript variants such as alternative splicing isoforms, single-nucleotide variants, and fusion genes, thus reflecting the complexity of biological samples. Furthermore, the 100 transcripts cover a wide length distribution (500–3800 nt) (Supplementary Table 1), and their proportional representation across different length bins was designed to be comparable to that of the human transcriptome. Similarly, the GC content of the spike-ins (40–50%) spans the range typically observed in human genes. All transcripts were confirmed to have high purity (Supplementary Table 2, Supplementary Fig. 1). Critically, we utilized inverted mirror sequences of human genes. This design ensures that the spike-ins behave similarly to endogenous transcripts during the physical steps of library preparation and sequencing, while remaining computationally distinct, thus preventing analytical ambiguity.Fig. 2: Design and metrological characterization of the TranScale.Full size imagea Physicochemical properties of the 100 RNA spike-in transcripts of TranScale. Plots show the composition of different transcript types (pie chart, left), the distribution of transcript lengths (line plot, middle left), the proportion of transcripts in different length bins compared to human transcripts (bar chart, middle right), and the GC content compared to human transcripts (dot plot, right). b Certified values for TranScale. The heatmap shows the log2 (Mix1/Mix2) concentration ratios across different concentration tiers (left). Line plots show the certified log2 concentrations with expanded uncertainties for each of the 100 transcripts in Mix1 (top right), Mix2 (middle right), and their log2 ratio (bottom right). Technical replicates are defined as the independent measurements performed to determine the absolute copy number concentration of each gene in the TranScale (n = 12). Data are presented as reference values ± expanded uncertainties (k = 2) of TranScale. c Orthogonal validation of certified values. Scatter plots compare the copy numbers assigned by IDMS (x-axis) with those determined by RT-dPCR (y-axis) for Mix1, Mix2, and their ratio. The identity line (y = x) and Pearson’s correlation coefficient (r) are shown. d Assessment of sequencing orthogonality. The bar chart shows the percentage of total reads aligning to the TranScale sequences versus the human genome (GRCh38) alone across different library preparation (L1, L2) and sequencing platforms (ILL, DNB). Technical replicates are libraries (n = 3). Data are presented as mean values ± SD. Source data for this figure is available in the Source Data file.Next, to establish a metrological anchor for absolute quantification, we first assigned certified values with SI traceability to the individual stock solutions of each of the 100 transcripts using isotope dilution mass spectrometry (IDMS), a gold-standard reference measurement procedure (Supplementary Fig. 2, Supplementary Table 3). To ensure this traceability, the absolute concentration of each transcript was determined against primary standards from the National Institute of Metrology (NIM) (Supplementary Table 4), and the procedure demonstrated high consistency across all measurands (Supplementary Table 5). The copy number concentration of each of the 100 transcript stocks was determined by averaging 12 measurement results derived from four repeated digestions across two independent vials. This rigorous characterization process yielded high precision, with relative standard deviations (RSDs) for all transcripts ranging from 1.06% to 9.04% (all well below 10%) (Supplementary Data 1). Following the theoretical design matrix (Fig. 2b, heatmap), these certified stocks were then used to prepare two master mixes (Mix 1 and Mix 2) via precision gravimetry.The final certified concentrations in these mixes were derived from the certified stock values and high-precision gravimetric data. The certified absolute concentrations for all 100 transcripts in both Mix 1 and Mix 2 of TranScale are presented (Fig. 2b, top and middle right plots; Supplementary Data 2). These values span the intended 10⁵-fold dynamic range, making them suitable for calibrating a wide spectrum of transcript abundances. Crucially, a comprehensive uncertainty budget was established for each transcript to define the final certified values and their associated expanded uncertainties (k = 2). This evaluation systematically combined the uncertainty contributions from the initial characterization (uchar, incorporating uncertainties from both IDMS and gravimetric preparation), between-vial homogeneity (ubb), and long-term stability (ults), as detailed in Supplementary Data 3. The final expanded relative uncertainties for the absolute copy number concentrations were confirmed to be ≤16% for Mix 1 and ≤18% for Mix 2. The resulting uncertainty for the critical Mix 1-to-Mix 2 expression ratios was confirmed to be ≤23%. Furthermore, the log₂ expression ratios between the two mixes were certified to establish the definitive metrological anchor for relative quantification. While the design targeted a nominal 256-fold range (Log₂ from −4 to +4), the certified values defined an actual 378-fold dynamic range (Log₂ from −3.84 to +4.73) (Fig. 2b, bottom right plot).To independently confirm the accuracy of the IDMS-certified values of TranScale, we performed orthogonal validation using one-step reverse transcription digital PCR (RT-dPCR) (Supplementary Table 6, Supplementary Data 4, Supplementary Fig. 3-7), and their reverse transcription efficiencies were validated (Supplementary Table 7). For a representative subset of 20 transcripts spanning the full concentration range, the copy numbers determined by RT-dPCR showed a high degree of concordance with the IDMS-assigned values (Pearson’s r > 0.99), verifying the ground-truth accuracy of our RM (Fig. 2c, Supplementary Table 8). Furthermore, TranScale successfully met the stringent ISO criteria for homogeneity and stability (Supplementary Table 9, Supplementary Data 5–8), ensuring its reliability for widespread distribution and use.Finally, we confirmed the sequencing orthogonality of the spike-in RM. We sequenced TranScale alone using multiple library preparations (poly(A) selection, rRNA depletion) and sequencing platforms (Illumina NovaSeq, MGI DNBSEQ-T7). In all cases, reads aligned with high efficiency (>99.1%) to the TranScale reference sequences, while a negligible fraction ( 0.1 in all three technical replicates within each qualifying batch. High-quality batches were screened based on absolute quantitative performance evaluation and inter-batch statistical consistency. This process generated a new, dual-function reference dataset containing absolute copy numbers for 13,483 genes in RM D5 and 13,862 genes in RM D6 (Fig. 6a, Supplementary Fig. 12, Supplementary Data 12). The metrological quality of this dataset was high, with over 84% of the assigned absolute values having a relative standard uncertainty (uchar) below 20%.Fig. 6: Calibrated absolute quantification enables inter-gene comparability.Full size imagea Density plots of log₂ absolute copy number distributions for quantified genes in samples D5 (blue, 11,398 genes) and D6 (green, 12,164 genes). Orthogonal validation of calibrated RNA-seq measurements. Scatter plots compare log2 absolute copy numbers from RNA-seq with those from RT-dPCR for (b) sample D5 and (c) sample D6. d Comparison of the log2 expression ratio (D5/D6) measured by RNA-seq against the reference ratio from RT-dPCR. Measurement error (ME) of the absolute quantities (with mean-normalization) with absolute datasets of D5 (e) and D6 (f) constructed by TranScale across eight sequencing batches. Data are presented for genes that were detected in every batch and covered by the established absolute quantification reference datasets (N = 10,000 for D5 and D6). Calibration performance evaluation. Plots show D5/D6 ratio MEs before (g) and after (h) library-specific calibration by TranScale. MEs are calculated by the D5/D6 ratio from the reference datasets reported previously. Data are presented for genes that were detected in every batch and covered by the external ratio-based quantification reference datasets (N = 6740 for D5 and D6). For violin plots, the white box represents the interquartile range (IQR), and the whiskers extend to 1.5 × IQR. The center line is the median, and the shaded area shows the probability density. Technical replicates are libraries (n = 3). i Comparison of quantification profiles for a set of housekeeping genes using a relative metric (log2 FPKM, left) versus calibrated absolute copy numbers (log2 copy number, right). j Direct comparison of calibrated absolute expression levels (log2 copy number) for a set of housekeeping genes (top) and tumor-associated genes (bottom). Source data for this figure is available in the Source Data file.To rigorously validate this new absolute reference dataset, we performed extensive orthogonal and external benchmarking. First, absolute copy numbers measured by our calibrated RNA-seq showed strong concordance with quantities measured by a reference method, RT-dPCR, for 46 target genes in both D5 (Fig. 6b, Pearson’s r = 0.89) and D6 (Fig. 6c, Pearson’s r = 0.89). The expression ratios (D5/D6) derived from these absolute values were also highly concordant with those from RT-dPCR (Fig. 6d, Pearson’s r = 0.88).With this internally validated absolute dataset of endogenous genes in hand, we could test the central premise of our framework: whether the diagnostic paradigm developed using spike-ins holds true for the endogenous transcriptome. We applied the same dual-metric analysis to our dPCR-validated dataset, calculating the Absolute and Ratio Measurement Error (ME) distributions for thousands of endogenous genes across the eight batches (Fig. 6e–g). The results provided a powerful, transcriptome-wide validation of the two distinct and non-intuitive failure modes first identified using our calibrators. The L2_DNB_P batch, for instance, unequivocally confirmed the “subtle paradox.” Mirroring its behavior with the spike-ins, this workflow exhibited highly dispersed Absolute MEs for endogenous genes, indicating a chaotic internal measurement process. Critically, this severe flaw was again masked by a deceptively precise Ratio ME distribution (Fig. 6g). This provides definitive proof that the danger of high relative precision masking poor absolute accuracy is a real-world risk for endogenous gene quantification. Likewise, the L2_DNB_R batch confirmed the second, equally critical paradox. Just as with the calibrators, this workflow displayed a remarkably consistent internal process for endogenous genes, evidenced by its tightly compacted Absolute ME distribution (Fig. 6e, f). Yet, it again failed to preserve relative quantification, revealed by its dispersed Ratio ME distribution (Fig. 6g).By demonstrating that the exact same paradoxical behaviors discovered in our controlled system are replicated at the transcriptome scale, we confirm that these vulnerabilities are not theoretical but are inherent risks in the measurement of endogenous genes. This establishes that a dual-metric system is not merely an improvement but a fundamental necessity for robust quality control in any RNA-seq experiment. As a final validation, we benchmarked our data against the original Quartet “ground truth” ratios13. The calibration significantly reduced the ME and tightened the distribution of measured ratios compared to the uncorrected data (Fig. 6g, h). This demonstrates that our absolute correction also improves the accuracy of relative quantification against an established external standard.Having validated the framework, we applied it to reveal the true quantitative landscape of endogenous genes. We observed that conventional FPKM values present a compressed and distorted view of housekeeping gene expression. In contrast, our calibration transforms the data to absolute molecular counts, restoring the full dynamic range and revealing a well-defined quantitative structure (Fig. 6i). This fundamental restoration of the data landscape enables direct and meaningful comparisons between the absolute expression levels of different genes (Fig. 6j).This new capability for absolute quantification provides novel biological insights by enabling the quantitative dissection of distinct biological relationships. First, we examined the link between a master regulator and its direct target. In the constructed absolute quantification dataset D5, our data reveal that the absolute transcript level of the MET proto-oncogene is about 150-fold lower than that of its key downstream adaptor, the GRB2, and that is highly consistent with the results verified by RT-dPCR (log₂ copy number ratio of GRB2/MET ≈ 7.3 vs. ≈ 7.7, Supplementary Fig 13). Our study enabled the expression profile quantification within any given sample, and also validated the absolute quantification of this pair of genes in dataset D6. Likewise, the copy number ratio of these two genes across both samples was consistent with the RT-dPCR results (Fig. 6b–d, Supplementary Fig 13). This observation moves beyond simple correlation to provide, using a generalized method, an estimate of the transcript abundance ratio within a critical therapeutic pathway. In a second example, our framework provides a quantitative link between genomic structure and transcriptional output for tumor samples. We observed that the absolute transcript levels of ERBB2 and GRB7, two genes known to be co-amplified on chromosome 17q12 in tumor tissue, such as breast cancer20. We verified that the basal expression level of GRB7 is low in normal sample (11.78 log2 copies per μg of total RNA for D5), but is elevated by over 400-fold (20.42 log2 copies per μg of total RNA) in the tumor sample LCA (Supplementary Fig. 13a). Furthermore, the expression levels of GRB7 and ERBB2 are approximately equal, as an excellent concordance (20.1 vs. 20.4 log2 copies per μg of total RNA) between TranScale-calibrated RNA-seq and dPCR absolute measurements (Supplementary Fig. 13a), this co-overexpression provides biological insight into the enhancement of oncogenic signaling, and quantitative evidence that a genomic co-amplification event is translated into a near 1:1 ratio of transcript output—a precise mapping from genome to transcriptome that was previously difficult to ascertain with confidence (Fig. 6i). Collectively, these results, validated by both an orthogonal method and an external reference standard, demonstrate that our application framework successfully mitigates sequence-dependent biases for endogenous transcripts. This enables robust, accurate, and direct comparisons of absolute expression levels between different genes, resolving a fundamental limitation in transcriptomics.Demonstration of robust diagnostic classificationFinally, to illustrate the potential of TranScale in identifying biological insights and supporting clinical decision-making, we conducted a diagnostic simulation targeting the ERBB2 (HER2) oncogene across 12 independent sequencing libraries. We established a “ground truth” diagnostic cutoff (18.5 of log2 copies/µg total RNA) based on reference dPCR data, which clearly distinguished the tumor model (LCA) from normal controls (Quartet D5/D6). We observed that technical batch effects in uncalibrated relative quantification (FPKM) frequently masked the overexpression phenotype of the tumor samples, leading to inconsistent classifications. In contrast, TranScale calibration effectively neutralized these variations, recovering the true absolute abundance of ERBB2 and ensuring 100% concordance with the dPCR-defined diagnostic classification across all platforms and laboratories (Supplementary Fig. 14). These results highlight the capability of SI-anchored absolute quantification to reveal biological signals that may be obscured by technical noise in multi-center studies.DiscussionRNA-seq’s quantitative power is constrained by equating read counts with molecular quantity1,21, a practice causing batch effects that hinder data integration22 and clinical benchmarks23,24. The field has relied on retrospective computational normalization25,26, post-hoc methods that treat statistical symptoms, not the underlying measurement problem, and risk conflating artifacts with biological signals27. We address this by establishing a metrological framework that redefines RNA-seq as an absolute quantitative method. By introducing SI-traceable, biomimetic TranScale calibrators, we create a chain of traceability from sequencing reads to the SI unit (the mole). This physical calibration reduces inter-laboratory CV from >85% to 3800 nt) yielded impure products due to premature termination. Prioritizing analytical validity, we systematically shortened these sequences to a maximal length that permitted consistent, high-purity synthesis.(iii)Purification and verification: The resulting full-length RNA was purified using the MEGAclear Kit (Thermo Fisher), and its concentration was estimated with a Nanodrop. The integrity and purity of each of the 100 final transcripts were rigorously verified as a single, sharp peak on an Agilent 2100 Bioanalyzer (RNA 6000 Nano kit).Final panel composition and sufficiencyThe final TranScale panel consists of 100 well-defined, high-purity RNA transcripts. Despite the necessary length adjustments for quality control, the panel spans a significant range of lengths (500–3800 nt). Importantly, the utility of the panel is not contingent on the detection of all 100 calibrators. Our operational analyses show that robust calibration models (R² > 0.95) can be reliably constructed using as few as 20-30 well-distributed calibrators, confirming the sufficiency of the final panel for its intended application (Supplementary Fig. 10).Isotope Dilution Mass Spectrometry (IDMS) for TranScaleThe absolute concentration (copy number) of each of the 100 purified RNA transcripts of TranScale RM was certified using Isotope Dilution Mass Spectrometry (IDMS), a primary reference measurement procedure recognized by the Joint Committee for Traceability in Laboratory Medicine (JCTLM). The certification workflow was as follows:NMP standards preparationAs standards, the certified RMs of adenosine 5′-monophosphate (AMP), guanosine 5′-monophosphate disodium salt (GMP), cytidine 5′-monophosphate (CMP), and uridine 5′-monophosphate disodium salt (UMP) were obtained from the National Institute of Metrology, China (NIM) (NMPs). The corresponding isotope-labeled (13 C, 15 N) nucleotide monophosphates (LNMPs) (Silantes, Germany) were used as internal controls.Quantification of NMPs standardsFor quantification of transcripts, NMPs in the digested samples were separated completely using an SB-AQ C18 column (Agilent, USA), 0.1% formic acid (v/v) was used as mobile phase in a flow rate of 0.2 mL/min maintained at 30 °C. Signals of the well-separated NMPs were detected by SCIEX QTRAP® 6500 + LC-MS/MS in positive ion and multiple reaction monitoring (MRM) mode. Other instrumental and mass spectrometer data acquisition parameters, including Q1, Q3, DP, EP, CE and CXP, were optimized for the NMPs and LNMPs detection (Supplementary Table 3). A known amount of a corresponding stable isotope-labeled internal standard (LNMPs) for each ribonucleoside (NMPs). Mixtures with mass ratios of NMPs to LNMPs of 0.4, 0.8, 1.2, 1.6, and 2.0 were performed, respectively. NMP concentrations (μg/g) in transcripts were determined according to the standard curves of mass ratio and IDMS peak area ratio of NMPs to LNMPs (Supplementary Fig. 2). The raw mass spectrometry data were processed for absolute quantification using Analyst software (version 1.6, AB Sciex).Hydrolysis of RNA transcriptsAn aliquot of the purified RNA transcript was subjected to complete enzymatic hydrolysis to break it down into its constituent ribonucleosides. According to the protocol of our previous studies, briefly, each of the generated transcripts was diluted to a concentration of 1 ng/μL, and then 50 μL RNA sample was mixed with Phosphodiesterase I from Crotalus adamanteus venom (SVP, Sigma-Aldrich, USA) at the final concentration of 0.002 U/μL, 5 μL of LNMPs mixture was added. All reactions were formulated by the gravimetric method. After 25 min of incubation at 25 °C, the reaction was terminated by incubation at 80 °C for 15 min48.Quantification of target transcriptsThe molar concentration of each of the four NMPs was determined with high precision and accuracy by measuring the abundance ratio of the natural analyte to its corresponding LNMPs. According to the quantification method of NMPs, each RNA sample was measured twice independently, and each digestion was analyzed in triplicate. All the above measurements were independently conducted by two operators. The final mass fraction of each NMP in the digested RNA sample was expressed as Eq. (1):$${W}_{{RNA}}=\frac{{W}_{X}\times {M}_{{RNA}}}{{M}_{{NMP}}\times N}$$(1)where WRNA is the mass fraction of the RNA sample in micrograms per gram, Wx is the mass fraction of the selected NMP in the RNA sample, MRNA is the molecular mass of the RNA molecule, MNMP is the molecular mass of the selected NMP, and N is the number of the selected NMP in the RNA sample. The copy number of RNA (n, in copies per microgram) was expressed as Eq. (2) according to the Avogadro constant and RNA molecular weight48. Where NA is Avogadro’s constant. A density of 1.00 g/mL was taken into account in the calculation. The copy numbers (copies/μL) of transcripts were calculated.$$n=\frac{{W}_{{RNA}}\times {NA}}{{M}_{{RNA}}\times {10}^{9}}$$(2)Consistency check of IDMSConcentration of a specific RNA transcript was calculated independently from the concentration of each of the four nucleosides, based on the known sequence of that transcript (e.g., [RNA Transcript] = [Adenosine] / number of ‘A’s in sequence). The final certified value for each transcript was assigned as the average of the four independent calculations, provided they showed high consistency (e.g., relative standard deviation 0.1 threshold was established to exclude these unreliable data points, ensuring that the calibration model is constructed only from robust and accurately quantified calibrators.Dynamic Range threshold (>2¹⁰-fold)This criterion ensures that the calibration model is not built on a narrow concentration range. A wide dynamic range is essential for the model to have sufficient leverage to accurately determine the slope, ensuring its validity when applied to endogenous transcripts across the full expression spectrum.Coefficient of determination (R²) threshold (>0.95)An R² > 0.95 is a widely accepted standard for demonstrating a strong linear relationship. Setting this as the final gate serves as the ultimate confirmation that the foundational assumption of linearity holds true for the specific library being processed, providing high confidence in the subsequent calibration.Minimum number of calibrators for robust model fittingA critical, implicit QC gate in our framework is the number of calibrators that successfully pass the ME, FPKM, and dynamic range criteria. While a linear model can be generated from a few points, its reliability and predictive power for the entire transcriptome depend on a statistically robust foundation. Based on our multi-batch analysis, even lower-quality libraries that were ultimately salvageable for calibration consistently yielded a substantial number of calibrators (typically >30, as shown in Fig. 4a). Therefore, we establish a practical guideline: a library should yield a minimum of 20–30 high-confidence calibrators to proceed with high-confidence absolute quantification.This recommendation is based on the following rationale: (1) Statistical Stability: A sufficient number of calibrators ensures that the regression parameters (slope and intercept) are stable and accurately reflect the library’s systemic bias, rather than being skewed by random noise or a few outliers. (2) Diagnostic Power: The number of passing calibrators serves as a powerful meta-indicator of overall library quality. A failure to meet this minimum threshold strongly suggests underlying issues (e.g., significant non-linear biases, poor library preparation) that make the data unsuitable for absolute quantification, even if a model with a high R² can be formally generated. Libraries falling below this guideline should be flagged, and their use should be limited to relative analyses.Construction and metrological characterization of endogenous gene reference datasetsTo establish robust reference datasets for endogenous gene expression (designated D5 and D6), we processed data from eight independent measurement batches through a stringent filtering and characterization workflow. Beginning with the complete Ensembl annotation (n = 58,735), genes were retained only if detected in at least six of eight batches, present in all three technical replicates per batch, and exhibited an FPKM > 0.1 across all replicates. For this filtered set, FPKM values were converted to absolute copy numbers using the library-specific linear models derived from the co-processed TranScale calibrators. The final certified reference value for each gene was assigned as the arithmetic mean of its corrected absolute copy numbers from all qualifying batches. For the purpose of this study, the measurement uncertainty was evaluated based on the dominant experimental component. This characterization uncertainty (u_char), a Type A evaluation, was calculated as the relative standard deviation (RSD) of the final calibrated copy numbers for each gene across all contributing measurement batches. This approach is based on the principle that for a complex, multi-stage workflow like RNA-seq, the experimentally observed reproducibility (u_char) is the largest and most practically relevant contributor to the combined uncertainty, significantly outweighing the Type B uncertainties propagated from the calibration standards. Therefore, u_char provides a direct and robust measure of the quality and consistency of the final reference values. Accordingly, the final reference gene sets were filtered based on this primary uncertainty metric, retaining only those with a relative u_char of ≤ 20%. For the D5/D6 ratio dataset, values and their combined uncertainties were calculated by propagating the respective u_char values using the standard formula for division according to Eq. (12).$${u}_{D5/D6}=\sqrt{{u}_{D5\,}^{2}+{u}_{D6\,}^{2}}$$(12)Validation of reference datasets using RT-dPCRThe absolute gene expressions in the reference datasets were further validated by dPCR; primers of target genes in D5 and D6 were listed in Supplementary Data 12. First, initial cDNA synthesis was performed by incubating 2 μL RNA with 4 μL of 5× PrimeScript IV cDNA Synthesis Mix (Takara #6215 A) containing PrimeScript IV RTase, RNase Inhibitor, Oligo dT Primer, and dNTPs supplemented with 1 μL random hexamers. Nuclease-free water was added to achieve a 20-μL reaction volume. This reaction mixture was incubated at 30 °C for 10 min and then for 15 min at 42 °C and finally for 5 min at 95 °C for termination. Second, dPCR reactions were employed by the Bio-Rad QX200 Droplet Digital PCR System with 20-μL mixtures containing: 10 μL EvaGreen Supermix, 2 μL primer pair, 2 μL cDNA template, and 6 μL RNase-free ddH₂O. Following droplet generation with 70 μL oil, 40-μL droplets were transferred to a 96-well plate. Amplification conditions comprised: 5 min at 95 °C; 40 cycles of 95 °C for 30 s and 60 °C for 1 min; followed by signal stabilization (4 °C for 5 min and 90 °C for 5 min). Signals were acquired using the Droplet Reader, with all reactions performed in duplicate. Absolute quantities were verified by comparing measured absolute copy numbers of selected genes in D5 and D6 against absolute copy numbers in the reference dataset, and also assessing D5/D6 copy number ratios versus established reference ratios for Ratio consistency.Diagnostic robustness simulationTo simulate a clinical diagnostic scenario, we utilized ERBB2 expression data from 12 sequencing libraries. A diagnostic cutoff was empirically defined as the midpoint separating the dPCR-quantified absolute copy numbers of the tumor model (LCA) and normal controls (Quartet D5 and D6). For the uncalibrated analysis, raw FPKM values were log2-transformed. For the calibrated analysis, TranScale-derived absolute copy numbers were used. We calculated the consistency of diagnostic classification (Tumor vs. Normal) across all batches relative to the dPCR-defined ground truth.Statistics & ReproducibilityStudy design and Sample sizeNo statistical method was used to predetermine sample size. The sample sizes were 12 technical replicates for initial measurement of each transcript of TranScale, and 3 independent technical replicates for values validation to ensure sufficient precision for measuring absolute copy number concentrations and to rigorously evaluate batch-effect correction as per metrological standards. We sequenced 3 replicates of each of the four RNA samples (D5, D6, LCA, and LCN) using 2 commercially available short-read sequencing protocols: PolyA and RiboZero, generating 12 data batches. Each batch included 6 libraries, resulting in a total of 72 libraries. These sample sizes are sufficient to provide within-batch technical replication, cross-protocol comparisons, and cross-batch/laboratory reproducibility assessment. Details are illustrated explicitly in Fig. 1.Data exclusionAll data from planned experiments have been included. All attempts at replication were successful.RandomizationThe experiments were not randomized, as the study did not involve group comparisons requiring random allocation. Samples were allocated by a pre-specified, balanced design. Each batch contained 6 libraries; there were 3 technical replicates for 2 paired samples (samples D5 and D6, or samples LCA and LCN).BlindingThe investigators were not blinded to allocation during experiments and outcome assessment, as the study relied on objective bioinformatic pipelines and predefined calibration standards.Data analysisAll attempts at replication were successful. Data processing and statistical analyses were performed using R (version 4.5.0) and associated packages, including ggplot2 (version 3.5.2) for visualization. Graphs were generated using GraphPad Prism (version 10.1.2), Origin (version 2021), and Adobe Illustrator (version 2025). Specific statistical tests used for homogeneity and stability assessments are detailed in the relevant results and methods sections above.Ethics StatementThis research complies with all relevant ethical regulations. The use of RNA materials derived from established clinical cell lines in this study was reviewed and approved by the Ethics Committee of the National Cancer Center/ Cancer Hospital, Chinese Academy of Medical Sciences (Approval no. 24/427-4707).Reporting summaryFurther information on research design is available in the Nature Portfolio Reporting Summary linked to this article.Data availabilityThe raw sequence data in this paper have been deposited in the Genome Sequence Archive (GSA) under the accession code GSA-Human: HRA01324452. Source data are provided with this paper.Code availabilitySource code for gene expression analysis and the TranScale calibration pipeline have been deposited on GitHub, available at https://github.com/zhyu0807/TranScale/tree/main53, and have been archived in Zenodo under: https://doi.org/10.5281/zenodo.1844619054.ReferencesStark, R., Grzelak, M. & Hadfield, J. RNA sequencing: the teenage years. Nat. Rev. Genet. 20, 631–656 (2019).Article CAS PubMed Google Scholar Zheng, Y. et al. Multi-omics data integration using ratio-based quantitative profiling with Quartet reference materials. Nat. Biotechnol. 42, 1133–1149 (2024).Article CAS PubMed Google Scholar Mortazavi, A., Williams, B. A., Mccue, K., Schaeffer, L. & Wold, B. Mapping and quantifying mammalian transcriptomes by RNA-Seq. Nat. Methods (2008).SEQC Consortium A comprehensive assessment of RNA-seq accuracy, reproducibility and information content by the Sequencing Quality Control Consortium. Nat. Biotechnol. 32, 903–914 (2014).Article Google Scholar Wang, D. et al. A real-world multi-center RNA-seq benchmarking study using the Quartet and MAQC reference materials. Nat. Commun. 15, 6167 (2024).Article ADS CAS PubMed PubMed Central Google Scholar Begley, C. G. & Ellis, L. M. Raise standards for preclinical cancer research. Nature 483, 531–533 (2012).Article ADS CAS PubMed Google Scholar Byron, S. A. et al. Translating RNA sequencing into clinical diagnostics: opportunities and challenges. Nat. Rev. Genet. 17, 257–271 (2016).Article CAS PubMed PubMed Central Google Scholar Johnson, W. E., Li, C. & Rabinovic, A. Adjusting batch effects in microarray expression data using empirical Bayes methods. Biostatistics 8, 118–127 (2007).Article PubMed Google Scholar Nygaard, V., Rødland, E. A. & Hovig, E. Methods that remove batch effects while retaining group differences may lead to exaggerated confidence in downstream analyses. Biostatistics 17, 29–39 (2016).Article MathSciNet PubMed Google Scholar External RNA Controls Consortium The External RNA Controls Consortium: a progress report. Nat. Methods 2, 731–734 (2005).Article Google Scholar Risso, D., Ngai, J., Speed, T. P. & Dudoit, S. Normalization of RNA-seq data using factor analysis of control genes or samples. Nat. Biotechnol. 32, 896–902 (2014).Article CAS PubMed PubMed Central Google Scholar Hardwick, S. A. et al. Spliced synthetic genes as internal controls in RNA sequencing experiments. Nat. Methods 13, 792–798 (2016).Article CAS PubMed Google Scholar Yu, Y. et al. Quartet RNA reference materials improve the quality of transcriptomic data through ratio-based profiling. Nat. Biotechnol. 42, 1118–1132 (2023).Article PubMed PubMed Central Google Scholar Better research through metrology. Nat. Methods 15, 395 https://doi.org/10.1038/s41592-018-0035-x (2018).Milavec, M. et al. Metrological framework to support accurate, reliable, and reproducible nucleic acid measurements. Anal. Bioanal. Chem. 414, 791–806 (2022).Article CAS PubMed Google Scholar Hu, J. F. et al. Quantitative mapping of the cellular small RNA landscape with AQRNA-seq. Nat. Biotechnol. 39, 978–988 (2021).Article CAS PubMed PubMed Central Google Scholar Fu, L. et al. Simultaneous quantification of methylated cytidine and adenosine in cellular and tissue RNA by nano-flow liquid chromatography–tandem mass spectrometry coupled with the stable isotope-dilution method. Anal. Chem. 87, 7653–7659 (2015).Article ADS CAS PubMed PubMed Central Google Scholar Fassett, J. D. & Paulsen, P. J. Isotope dilution mass spectrometry for accurate elemental analysis. Anal. Chem. 61, 643A–649A (1989).Article ADS CAS Google Scholar De Bièvre, P., Dybkaer, R., Fajgelj, A. & Hibbert, D. B. Metrological traceability of measurement results in chemistry. Pure Appl. Chem. 79, 1755–1761 (2007).Google Scholar Sircoulomb, F. et al. Genome profiling of ERBB2-amplified breast cancers. BMC Cancer 10, 539–539 (2010).Article PubMed PubMed Central Google Scholar Shi, L. et al. Microarray scanner calibration curves: characteristics and implications. BMC Bioinforma. 6, S11 (2005).Article Google Scholar Luecken, M. D. & Theis, F. J. Current best practices in single-cell RNA-seq analysis: a tutorial. Mol. Syst. Biol. 15, e8746 (2019).Article PubMed PubMed Central Google Scholar McNutt, M. Reproducibility. Science 343, 229–229 (2014).Article ADS CAS PubMed Google Scholar Baker, M. Reproducibility crisis. Nature 533, 353–366 (2016).ADS Google Scholar Zhang, Y., Parmigiani, G. & Johnson, W. E. ComBat-seq: batch effect adjustment for RNA-seq count data. NAR Genom. Bioinform. 2, lqaa078 (2020).Article PubMed PubMed Central Google Scholar Tran, H. T. N. et al. A benchmark of batch-effect correction methods for single-cell RNA sequencing data. Genome Biol. 21, 12 (2020).Article CAS PubMed PubMed Central Google Scholar Hicks, S. C., Townes, F. W., Teng, M. & Irizarry, R. A. Missing data and technical variability in single-cell RNA-sequencing experiments. Biostatistics 19, 562–578 (2018).Article MathSciNet PubMed PubMed Central Google Scholar Cole, M. B. et al. Performance assessment and selection of normalization procedures for single-cell RNA-seq. Cell syst. 8, 315–328. e318 (2019).Article CAS PubMed PubMed Central Google Scholar Robinson, M. D. & Oshlack, A. A scaling normalization method for differential expression analysis of RNA-seq data. Genome Biol. 11, R25 (2010).Article PubMed PubMed Central Google Scholar Svensson, V., Vento-Tormo, R. & Teichmann, S. A. Exponential scaling of single-cell RNA-seq in the past decade. Nat. Protoc. 13, 599–604 (2018).Article CAS PubMed Google Scholar Lun, A. T., Bach, K. & Marioni, J. C. Pooling across cells to normalize single-cell RNA sequencing data with many zero counts. Genome Biol. 17, 75 (2016).Article PubMed PubMed Central Google Scholar Fu, Y., Wu, P.-H., Beane, T., Zamore, P. D. & Weng, Z. Elimination of PCR duplicates in RNA-seq and small RNA-seq using unique molecular identifiers. BMC Genomics 19, 531 (2018).Article PubMed PubMed Central Google Scholar Kivioja, T. et al. Counting absolute numbers of molecules using unique molecular identifiers. Nat. Methods 9, 72–74 (2012).Article CAS Google Scholar Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 15, 550 (2014).Article PubMed PubMed Central Google Scholar Begley, C. G. & Ioannidis, J. P. Reproducibility in science: improving the standard for basic and preclinical research. Circ. Res. 116, 116–126 (2015).Article CAS PubMed Google Scholar Hoadley, K. A. et al. Cell-of-origin patterns dominate the molecular classification of 10,000 tumors from 33 types of cancer. Cell 173, 291–304. e296 (2018).Article ADS CAS PubMed PubMed Central Google Scholar Travaglini, K. J. et al. A molecular cell atlas of the human lung from single-cell RNA sequencing. Nature 587, 619–625 (2020).Article ADS CAS PubMed PubMed Central Google Scholar Caldas, J. & Vinga, S. Global meta-analysis of transcriptomics studies. PLoS One 9, e89318 (2014).Article ADS PubMed PubMed Central Google Scholar National Center for Research Platform. NCRM [Internet]. [cited 2025-08-19]. Available from: https://www.ncrm.org.cn/.Patro, R., Duggal, G., Love, M. I., Irizarry, R. A. & Kingsford, C. Salmon provides fast and bias-aware quantification of transcript expression. Nat. Methods 14, 417–419 (2017).Article CAS PubMed PubMed Central Google Scholar Robert, C. & Watson, M. Errors in RNA-Seq quantification affect genes of relevance to human disease. Genome Biol. 16, 177 (2015).Article PubMed PubMed Central Google Scholar Lähnemann, D. et al. Eleven grand challenges in single-cell data science. Genome Biol. 21, 31 (2020).Article PubMed PubMed Central Google Scholar Moses, L. & Pachter, L. Museum of spatial transcriptomics. Nat. Methods 19, 534–546 (2022).Article CAS PubMed Google Scholar Townes, F. W., Hicks, S. C., Aryee, M. J. & Irizarry, R. A. Feature selection and dimension reduction for single-cell RNA-Seq based on a multinomial model. Genome Biol. 20, 295 (2019).Article CAS PubMed PubMed Central Google Scholar Eraslan, G. et al. Single-nucleus cross-tissue molecular reference maps toward understanding disease gene function. Science 376, eabl4290 (2022).Article CAS PubMed PubMed Central Google Scholar Vesper, H. W., Myers, G. L. & Miller, W. G. Current practices and challenges in the standardization and harmonization of clinical laboratory tests. Am. J. Clin. Nutr. 104, 907S–912S (2016).Article CAS PubMed PubMed Central Google Scholar Saadatpour, A., Lai, S., Guo, G. & Yuan, G. C. Single-cell analysis in cancer genomics. Trends Mol. Med. 31, 576–586 (2015).CAS Google Scholar Niu, C. et al. Accurate quantification of SARS-CoV-2 RNA by isotope dilution mass spectrometry and providing a correction of reverse transcription efficiency in droplet digital PCR. Anal. Bioanal. Chem. 414, 6771–6777 (2022).Article CAS PubMed PubMed Central Google Scholar State Administration for Market Regulation. General and statistical principles for characterization of reference materials. JJF 1343-2022. (Standardization Administration of China, 2022).Chai, T. & Draxler, R. R. Root mean square error (RMSE) or mean absolute error (MAE)?–Arguments against avoiding RMSE in the literature. Geosci. Model. Dev. 7, 1247–1250 (2014).Article ADS Google Scholar Ali Abd Al-Hameed, K. Spearman’s correlation coefficient in statistical analysis. Int. J. Nonlinear Anal. Appl. 13, 3249–3255 (2022).Google Scholar Zhang, Y. et al. A metrological foundation for absolute transcriptomics using International System of Units-anchored calibrators, Genome Sequence Archive (GSA), https://ngdc.cncb.ac.cn/gsa-human/browse/HRA013244 (2026).Zhang, Y. et al. A metrological foundation for absolute transcriptomics using International System of Units-anchored calibrators, https://github.com/zhyu0807/TranScale/tree/main (2026).Zhang, Y. et al. A metrological foundation for absolute transcriptomics using International System of Units-anchored calibrators. Zenodo, https://doi.org/10.5281/zenodo.18446190 (2026).Download referencesAcknowledgementsThis work was supported in part by the National Key Research and Development Program of China (grant 2023YFF0613300 to L.D.), the Basic Research Fund Project of the National Institute of Metrology, China (grant AKYZD2202 to L.D.), and the National High-level Metrology Talent Cultivation Program (grant JLTD2601 to L.D.).Author informationAuthor notesThese authors contributed equally: Yu Zhang, Bingwen Yang, Ying Yu.Authors and AffiliationsCenter for Advanced Measurement of Science, National Institute of Metrology, Beijing, ChinaYu Zhang, Bingwen Yang, Xia Wang, Chunyan Niu, Yongzhuo Zhang, Yang Liu, Jingshu Li, Caihang Zhang, Jiayi Yang, Zheng Liu, Zhiyu Tang, Yunhua Gao, Xiang Fang & Lianhua DongState Key Laboratory of Genetics and Development of Complex Phenotypes, Human Phenome Institute and School of Life Sciences, Fudan University, Shanghai, ChinaYing Yu, Yuanting Zheng & Leming ShiShenzhen Institute for Technology Innovation, National Institute of Metrology, Shenzhen, ChinaYang LiuCollege of Chemical and Life Sciences, Beijing University of Technology, Beijing, ChinaJiayu TianDepartment of Pathology, Institute of Basic Medical Sciences, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, ChinaYuqin LiuState Key Laboratory of Molecular Oncology, Department of Etiology and Carcinogenesis, National Cancer Center/ National Clinical Research Center for Cancer/ Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, ChinaTing XiaoNational Center for Clinical Laboratories, Institute of Geriatric Medicine, Chinese Academy of Medical Sciences, Beijing Hospital/ National Center of Gerontology, Beijing, ChinaRui ZhangInternational Human Phenome Institutes (Shanghai), Shanghai, ChinaLeming ShiAuthorsYu ZhangView author publicationsSearch author on:PubMed Google ScholarBingwen YangView author publicationsSearch author on:PubMed Google ScholarYing YuView author publicationsSearch author on:PubMed Google ScholarXia WangView author publicationsSearch author on:PubMed Google ScholarChunyan NiuView author publicationsSearch author on:PubMed Google ScholarYongzhuo ZhangView author publicationsSearch author on:PubMed Google ScholarYang LiuView author publicationsSearch author on:PubMed Google ScholarJingshu LiView author publicationsSearch author on:PubMed Google ScholarCaihang ZhangView author publicationsSearch author on:PubMed Google ScholarJiayi YangView author publicationsSearch author on:PubMed Google ScholarJiayu TianView author publicationsSearch author on:PubMed Google ScholarZheng LiuView author publicationsSearch author on:PubMed Google ScholarZhiyu TangView author publicationsSearch author on:PubMed Google ScholarYunhua GaoView author publicationsSearch author on:PubMed Google ScholarYuanting ZhengView author publicationsSearch author on:PubMed Google ScholarYuqin LiuView author publicationsSearch author on:PubMed Google ScholarTing XiaoView author publicationsSearch author on:PubMed Google ScholarRui ZhangView author publicationsSearch author on:PubMed Google ScholarXiang FangView author publicationsSearch author on:PubMed Google ScholarLeming ShiView author publicationsSearch author on:PubMed Google ScholarLianhua DongView author publicationsSearch author on:PubMed Google ScholarContributionsL.D., L.S. and X.F. conceived and supervised the study. Y.Z., B.Y., Y.Y., X.W., C.N., Y.Z.Z., Y.L., J.L., C.Z., J.Y., J.T., Z.L., Z.T., Y.G., Y.T.Z., Y.Q.L., T.X. and R.Z. performed data analysis and/or interpretation. Y.Q.L. and T.X. provided paired cell line samples. Y.L. and J.L. cultured the cell lines and prepared RNA reference materials. Y.Z. and C.Z. characterized the reference materials TranScale. B.Y. and Y.L. performed RNA library preparation. X.W., Y.Y., Y.Z. and J.T. performed RT-dPCR validation. L.D. and Y.Z. managed the TranScale and reference datasets. Y.Z. generated most figures. Y.Z. and L.D. wrote the initial draft. L.D., L.S. and X.F. critically revised the manuscript. All authors reviewed and approved the final manuscript.Corresponding authorsCorrespondence to Xiang Fang, Leming Shi or Lianhua Dong.Ethics declarationsCompeting interestsThe authors declare no competing interests.Peer reviewPeer review informationNature Communications thanks Wei Xu and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.Additional informationPublisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.Supplementary informationSupplementary Information (download PDF )Peer Review File (download PDF )Description of Additional Supplementary Files (download PDF )Supplementary Data 1-12 (download ZIP )Reporting Summary (download PDF )Source dataSource Data (download XLSX )Rights and permissionsOpen Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.Reprints and permissionsAbout this article