A GLORI-seq experiment produces a dataset unlike any other in epitranscriptomics. Instead of enrichment peaks or editing rates, you receive a table of genomic positions, each with a number between 0 and 1 — the fraction of transcripts carrying m6A at that site. This is stoichiometric data, and it demands a different interpretive approach than the peak-based outputs most researchers are accustomed to.

This guide explains how to interpret the core outputs of a GLORI-seq experiment: the site-level modification table, QC metrics that validate data quality, the differential analysis that identifies condition-dependent m6A changes, and the annotation and prioritization steps that convert a list of sites into a validation-ready shortlist. It is written for researchers who have designed a GLORI-seq project — for guidance on experimental design, see Designing a GLORI-seq Project for Differential m6A Quantification — and are now preparing to interpret the data.

Figure 1: GLORI-seq data interpretation follows a structured workflow from raw reads through modification fractions to annotated differential candidates.

The Core Output: Site-Level Modification Table

The primary output of a GLORI-seq experiment is a site-level modification table. Each row represents one adenosine position in the transcriptome where m6A was detected, and each column provides information about that site.

A typical site-level table includes the following fields:

Genomic coordinate. The chromosome, position, and strand of the modified adenosine, reported in the reference genome build specified during project design.

Modification fraction. The proportion of reads at this position that retained an adenosine call after chemical deamination, representing the fraction of transcripts carrying m6A at this site. A value of 0.65 means approximately 65% of transcripts are modified at this position. Values are reported with associated confidence intervals that reflect read depth.

Read depth. The number of sequencing reads covering this position. Higher depth produces tighter confidence intervals around the modification fraction. Sites with low coverage may have wide confidence intervals and should be interpreted with caution.

Gene and transcript annotation. The gene symbol, transcript ID, and genomic region (5′ UTR, CDS, 3′ UTR, intronic, intergenic) associated with each site. This annotation is derived from the transcript reference specified in the project design and is essential for downstream interpretation.

Modification context. Whether the site falls within the canonical RRACH motif or a non-canonical sequence context. Most m6A sites conform to RRACH, but GLORI-seq detects sites regardless of motif, and the motif annotation helps distinguish high-confidence canonical sites from non-canonical ones that may warrant additional validation.

Sample-level values. If multiple samples were submitted, the modification fraction for each sample at each site is reported, enabling direct comparison across conditions and replicates.

This table is the foundation of all downstream analysis. Every interpretive step — differential testing, annotation, enrichment, prioritization — works from these site-level data. For researchers accustomed to peak files from MeRIP-seq, the shift is from interpreting regions to interpreting positions: each row is a specific adenosine with a specific modification level, not a broad region with an enrichment score.

Understanding Modification Fractions

The modification fraction is the central measurement in GLORI-seq, and interpreting it correctly is essential for sound biological conclusions.

What the fraction means. A modification fraction of 0.4 at an adenosine position means that, among the transcripts covering that position in the sequenced sample, approximately 40% carried m6A at that site and 60% were unmethylated. The fraction is a snapshot of the modification state across the transcript population at the moment of RNA extraction.

What the fraction does not mean. The modification fraction does not measure m6A abundance across all transcripts — it measures the proportion of the transcripts you sequenced that were modified at that position. If a gene is lowly expressed, the modification fraction at its m6A sites is estimated from fewer reads and has wider confidence intervals. The fraction does not directly report on m6A function; a site modified at 10% can be functionally important, and a site modified at 90% may be inconsequential, depending on the biological context.

Confidence intervals. Each modification fraction should be interpreted together with its confidence interval, which reflects the precision of the estimate given the read depth. A site with a modification fraction of 0.5 ± 0.05 (narrow interval) is well-estimated. A site with a modification fraction of 0.5 ± 0.25 (wide interval) is consistent with a broad range of true modification levels and should be treated as an uncertain estimate.

Distribution of modification levels. Published GLORI-seq data show that m6A modification fractions span the full range from near-zero to near-complete modification, with a broad distribution rather than a bimodal pattern. This means most sites are modified at intermediate levels — 20% to 60% is common — and the biologically interesting signal is often in the shift of these intermediate sites between conditions, not in sites that appear or disappear entirely.

Expression confounds. A change in modification fraction between conditions can reflect a change in m6A deposition or a change in transcript expression. If expression doubles but m6A deposition remains constant, the absolute number of modified reads doubles, but the modification fraction stays the same — GLORI-seq correctly reports no change. If you also need to distinguish expression changes from modification changes, matched RNA-seq data is required. For guidance on this decision, see Integrating RNA-seq and Epigenomic Data Analysis.

QC Metrics to Check Before Interpretation

Before diving into biological interpretation, several quality control metrics should be reviewed to confirm that the data supports the planned analyses.

Conversion efficiency. The fraction of unmethylated adenosines converted to inosines should exceed the laboratory threshold — typically 95% or higher. GLORI-seq reports approximately 99% conversion of unmethylated A under standard conditions. Samples with lower conversion efficiency produce inflated background and reduce the power to detect genuine m6A sites, particularly at low modification levels.

False-positive rate. The protocol includes an untreated RNA control or in vitro transcribed RNA control that carries no m6A. The fraction of sites called as modified in this negative control estimates the false-positive rate. Published data report approximately 1,856 false-positive sites with only 0.09% overlap with genuine m6A positions, indicating a low and well-characterized background.

Side-reaction rates. The chemical deamination produces minor side reactions — C-to-U conversion at less than 4% of cytidines and G-to-X conversion at approximately 3% of guanosines. These rates should be consistent across samples within a project. Outlier samples with elevated side-reaction rates may indicate processing issues and should be flagged.

Site count and saturation. The number of m6A sites detected per sample should be consistent across biological replicates within the same condition. At 50 Gb of sequencing per sample, approximately 80,000 sites are typically detected in mammalian transcriptomes; deeper sequencing increases this number. Large discrepancies in site count between replicates can indicate library preparation variability or sample quality differences.

Between-replicate concordance. Modification fractions at the same site across biological replicates should show reasonable agreement. While some biological variability is expected — and is precisely what replicates are designed to capture — systematic discordance between replicates warrants investigation. Pairwise scatter plots of modification fractions between replicates provide a visual check; correlation coefficients above 0.8–0.9 are typical for high-quality GLORI-seq data from matched conditions.

If any of these QC metrics fall outside expected ranges, consult with the service provider before proceeding with differential analysis. Running differential testing on data with unresolved QC issues risks identifying apparent differences that reflect technical artifacts rather than biology.

Figure 2: Five QC metrics — conversion efficiency, false-positive rate, side reactions, site counts, and replicate concordance — should be reviewed before proceeding to biological interpretation.

From Sites to Differential Modification Candidates

Differential analysis tests whether the modification fraction at each site differs between conditions. This is the step that connects GLORI-seq data to the biological question that motivated the experiment.

Statistical framework. Because the output at each site is a proportion (modified reads / total reads), differential testing typically uses models designed for proportion or count data. Beta-binomial models are common, as they account for both the modification fraction and the uncertainty associated with read depth. Sites with higher coverage and larger effect sizes are identified with greater confidence than sites with low coverage or small differences.

What differential analysis tests. For a two-group comparison — treatment versus control — the analysis tests the null hypothesis that the modification fraction is the same in both conditions at each site. Sites where the data are inconsistent with this null hypothesis are reported as differentially modified. The output includes an effect size — the difference in modification fraction between conditions — and a measure of statistical confidence, typically an adjusted p-value or false discovery rate.

Interpreting effect sizes. The effect size is the estimated difference in modification fraction between conditions. An effect size of +0.15 means the treatment condition has an estimated 15 percentage points higher modification at that site. Whether a given effect size is biologically meaningful depends on the system; a 10-percentage-point shift may be substantial for a site with baseline modification of 5%, but modest for a site with baseline modification of 60%. Report both the effect size and the baseline modification level when describing differentially modified sites.

Multiple testing correction. Differential testing is performed at tens of thousands of sites simultaneously, and the p-values must be corrected for multiple testing. The false discovery rate (FDR) — the expected fraction of reported positives that are false positives — is the standard metric. An FDR threshold of 0.05 is typical, meaning approximately 5% of sites called as differentially modified are expected to be false positives.

Filtering and prioritization. After differential testing, the list of significant sites is typically filtered by effect size to focus on changes that are likely to be biologically consequential. A common practice is to require both statistical significance (FDR < 0.05) and a minimum effect size (for example, |Δ modification fraction| > 0.1). This dual threshold removes sites where the statistical test detects a difference that is too small to be biologically interpretable.

For a deeper discussion of why single-base quantification enables this type of comparison, see Single-Base m6A Quantification: When Detection Is Not Enough.

Annotating and Contextualizing Differential Results

A list of differentially modified sites with coordinates and effect sizes is the starting point, not the endpoint. Annotation places these sites in biological context.

Gene feature annotation. Each differentially modified site is assigned to a gene feature category: 5′ UTR, coding sequence (CDS), 3′ UTR, or intronic. The distribution of differential sites across these categories can reveal broad patterns. For example, an enrichment of differential sites in 3′ UTRs may suggest condition-dependent regulation of mRNA stability, while enrichment in CDS may point to translation-related effects. The annotation should use the transcript reference specified during project design to ensure consistency.

Gene-level summarization. Since a single gene can harbor multiple m6A sites, gene-level summarization is often useful. A gene may contain several differentially modified sites that show consistent directionality — all increasing or all decreasing — or a mix of directions. Sites with consistent directionality within a gene strengthen the case that the gene's m6A landscape is condition-responsive. Sites with opposing directions within the same gene require closer scrutiny and may reflect isoform-specific regulation or annotation complexity.

Pathway and functional enrichment. Gene lists derived from differentially modified sites can be tested for enrichment in biological pathways, molecular functions, or curated gene sets. Standard tools such as Gene Ontology enrichment analysis or KEGG pathway analysis are applicable. The interpretive note is that m6A modification is not the same as gene expression — an enrichment in a pathway means the pathway's transcripts tend to carry condition-dependent m6A changes, not necessarily that the pathway is up- or down-regulated.

Sequence motif analysis. The sequences flanking differentially modified sites can be analyzed for enriched motifs. The canonical RRACH motif should be the most enriched motif in the full set of m6A sites. Differential sites that share additional sequence features beyond the canonical motif may point to condition-specific recognition by particular m6A writers or readers.

For projects combining GLORI-seq with other data types, Epigenomic Data Analysis services support integrated interpretation across modification, expression, and chromatin datasets.

Building a Validation Shortlist

Differential analysis on GLORI-seq data can produce hundreds to thousands of candidate sites. Validation bandwidth is always smaller. A disciplined shortlist is the bridge between discovery and confirmation.

Rank by confidence and effect. The initial rank should be driven by the data: sites with the strongest statistical support (lowest FDR) and largest effect sizes rise to the top. Within this set, prioritize sites where the direction of change is consistent with the biological expectation — for example, sites that lose modification in a writer knockout rather than gain it.

Favor well-covered sites. Sites with high read depth in all samples produce more reliable modification fraction estimates. A site with a large effect size but low coverage is a riskier validation candidate than a site with a moderate effect size but deep coverage across all replicates.

Annotate for validation tractability. For each shortlisted site, assess whether it can be validated with an orthogonal method. Sites within the RRACH motif are typically accessible to targeted validation methods such as SELECT-m6A sequencing. Sites in well-expressed genes with clear isoform structures are better candidates than sites in complex or poorly annotated loci. For targeted site validation, SELECT-m6A sequencing provides site-specific m6A quantification without requiring transcriptome-wide analysis.

Cross-reference with external evidence. Compare the shortlist against published datasets, RNA modification databases, and functional genomics resources. Sites that have been independently reported in related systems gain credibility, though absence from databases does not count against a site — it may reflect that the relevant condition or tissue has not yet been profiled.

Define go/no-go criteria. For each candidate, write a specific, measurable criterion for what result from the validation experiment would advance or retire the candidate. This prevents the shortlist from becoming an open-ended to-do list and ensures that validation experiments produce decisions, not just more data.

For a detailed framework on candidate prioritization across epitranscriptomics datasets, see Target Prioritization in Epitranscriptomics.

Figure 3: Building a validation shortlist from differential GLORI-seq results requires ranking, filtering, and defining decision criteria for each candidate.

What to Expect in a GLORI-seq Data Delivery

Understanding the structure of a typical GLORI-seq data delivery helps set expectations and plan the downstream analysis workflow.

Site-level modification table. The primary data file, provided in tabular format compatible with standard analysis tools (R, Python, Excel). Each row is a site, each column a sample or annotation field. This table is the starting point for custom analyses beyond the standard differential testing pipeline.

Differential modification results. A table of sites tested for differential modification, with effect sizes, confidence intervals, raw and adjusted p-values, and annotation fields. Sites meeting the significance and effect-size thresholds are flagged.

QC report. A summary of per-sample QC metrics, including conversion efficiency, false-positive rate estimates, side-reaction rates, site counts, and between-replicate concordance. This report confirms that the data meets quality thresholds for the planned analyses.

Annotation summary. A breakdown of site distribution across gene features, with separate summaries for all detected sites and for differentially modified sites. Enrichment of differential sites in specific gene features or pathways may be noted.

Visualization outputs. Standard visualizations typically include modification fraction distributions, volcano plots of differential results, chromosome-level views of modification patterns, and gene-level tracks showing modification fractions across transcript features.

For a complete overview of how these deliverables fit into a broader epitranscriptomics project, RNA modification services at CD Genomics provide integrated workflows from experimental design through data delivery and interpretation support.

Summary

Interpreting GLORI-seq results is a structured process that moves from site-level modification fractions through differential analysis to a validation-ready shortlist.

The modification fraction at each site — a number between 0 and 1 — is the fundamental data unit, and understanding its precision (confidence intervals), its distribution (intermediate levels are the norm), and its independence from expression is the foundation of sound interpretation. QC metrics, particularly conversion efficiency and between-replicate concordance, gate the transition from data review to biological interpretation. Differential analysis identifies sites where modification levels differ between conditions, with effect sizes and FDR-adjusted confidence measures attached. Annotation by gene feature, pathway, and motif places differential sites in biological context. And a disciplined shortlisting process — ranking by confidence and effect, filtering by coverage and tractability, and defining go/no-go criteria — converts the differential results into candidates that can be validated with orthogonal methods.

The interpretive workflow mirrors the structure of the GLORI-seq method itself: single-base, quantitative, and designed to support comparison. When the data is interpreted with the same rigor applied to its generation, GLORI-seq delivers modification-level insights that no detection-based method can provide.

FAQ

1. What is the most important number to look at in GLORI-seq results?

The modification fraction — the proportion of reads retaining an adenosine call at each site — is the central measurement. Each site receives a value between 0 and 1, representing the fraction of transcripts carrying m6A at that position. Always interpret this number together with its confidence interval, which reflects the precision of the estimate given the sequencing depth at that site.

2. How many differentially modified sites should I expect from a GLORI-seq experiment?

The number depends on the biological contrast, the effect-size threshold, and the sequencing depth. In published data, conditions with strong biological differences — such as writer knockout versus wild-type — can produce thousands of differentially modified sites. More subtle comparisons may yield hundreds. The number of differential sites is less important than the confidence attached to each one; a smaller set of high-confidence sites is more valuable than a larger set of borderline candidates.

3. What QC metrics indicate that GLORI-seq data is suitable for differential analysis?

Conversion efficiency should exceed 95%. The false-positive rate estimated from the untreated RNA control should be low (published data report approximately 0.09% overlap with genuine sites). Between-replicate concordance should be high, with modification fraction correlations above 0.8–0.9 for matched conditions. Side-reaction rates (C-to-U, G-to-X) should be consistent across samples and within expected ranges.

4. Can I interpret GLORI-seq differential results without matched RNA-seq?

Yes, for the interpretation of modification fraction changes themselves. GLORI-seq reports modification fractions that are independent of expression — if expression doubles and modification stays constant, the fraction stays the same. However, if the biological question requires distinguishing expression-driven changes from modification-driven changes in the absolute number of modified transcripts, matched RNA-seq is needed.

5. How should I prioritize differentially modified sites for validation?

Prioritize by confidence (lowest FDR), effect size (largest absolute difference in modification fraction), and coverage (highest read depth across all replicates). Then filter for validation tractability — sites accessible to orthogonal methods, in well-annotated genes with clear isoform structures. Cross-reference against published data for independent support, and define explicit go/no-go criteria for each candidate before beginning validation experiments.

References

1. Liu, Cong, Hanxiao Sun, Yunpeng Yi, Weiguo Shen, Kai Li, Ye Xiao, Fei Li, et al. "Absolute quantification of single-base m6A methylation in the mammalian transcriptome using GLORI." Nature Biotechnology, vol. 41, 2023, pp. 355–366. DOI: 10.1038/s41587-022-01487-9
2. Shen, Weiguo, Hanxiao Sun, Cong Liu, et al. "GLORI for absolute quantification of transcriptome-wide m6A at single-base resolution." Nature Protocols, vol. 19, 2024, pp. 1252–1287. DOI: 10.1038/s41596-023-00937-1
3. Garcia-Campos, Miguel Angel, Sarit Edelheit, Ursula Toth, Modi Safra, Ran Shachar, Sergey Viukov, Roni Winkler, et al. "Deciphering the 'm6A Code' via Antibody-Independent Quantitative Profiling." Cell, vol. 178, no. 3, 2019, pp. 731–747.e16. DOI: 10.1016/j.cell.2019.06.013
4. Dominissini, Dan, Sharon Moshitch-Moshkovitz, Schraga Schwartz, Mali Salmon-Divon, Lior Ungar, Sivan Osenberg, Karen Cesarkas, et al. "Topology of the human and mouse m6A RNA methylomes revealed by m6A-seq." Nature, vol. 485, 2012, pp. 201–206. DOI: 10.1038/nature11112

CD Genomics provides epitranscriptomics services for research use only and they are not intended for clinical diagnosis or treatment.