Summary

Yes, existing methylation data can often support outsourced DMR analysis without generating new sequencing, provided the files, genome build, sample metadata, and intended comparisons are clear. Raw FASTQ files allow the broadest QC review; processed matrices can support downstream statistics but limit auditability. The first step is a data inventory, not a software choice.

Upload a Data Inventory for Feasibility Review

Figure 1: Existing-data DMR analysis begins with an inventory and compatibility review before statistical testing.

Key Takeaways

• Existing FASTQ, BAM, coverage, matrix, or IDAT files can support different levels of DMR reanalysis.
• A data inventory must connect file names, sample IDs, metadata, genome build, and intended contrasts.
• Processed matrices can support downstream statistics but cannot recreate every read-level QC step.
• Deliverables should distinguish rerun findings, inherited QC, and unresolved interpretation boundaries.

Start with the Question, Not the File Format

In many projects, yes. A new sequencing run is not automatically required when existing data have enough sample-level information, usable methylation measurements, and a defined comparison. The scope depends on what has been retained:

The Epigenomic Data Analysis service is the natural starting point when the analysis scope needs to be matched to the files already available. The current WGBS service and Human DNA Methylation Microarray Service pages are useful references for the types of data produced by those workflows, but the feasibility of a reanalysis project still depends on the actual files.

Inventory Existing Files Before Reanalysis

File names alone are not enough. Record the file type, sample identifier, data level, genome build, and whether each file belongs to the same processing batch.

FASTQ files

FASTQ files are the most flexible starting point for WGBS, RRBS, EM-seq, or related sequencing data. Submit both mates for paired-end data, all lanes if a sample was split across lanes, and the sample sheet linking file names to biological labels. If files were already demultiplexed or trimmed, record the software and parameters if known.

BAM files and methylation calls

BAM files can support mapping and coverage review when the reference genome is available. Include the index files when possible. Methylation calls should identify chromosome, coordinate, strand or context, methylated count, unmethylated count, and total coverage. A beta value without counts is less informative for sequencing-based reanalysis because it cannot be audited against read depth.

Coverage tables and matrices

Coverage tables are often the most useful processed input for a fast DMR review. Each row should have a stable genomic coordinate or probe ID, and each column should map to exactly one sample. Preserve missing values rather than replacing them with zero. A zero methylation value and an unmeasured CpG are not the same observation.

For array projects, provide the beta-value matrix, M-value matrix if available, detection p-values, bead counts or probe-quality fields, and the platform annotation. IDAT files are preferable when available because they permit a more complete technical QC review.

Figure 2: The retained data level determines which QC and reanalysis steps remain available.

Metadata Needed to Reconstruct the Comparison

Metadata are the bridge between a file and a valid contrast. At minimum, provide:

• A stable sample ID that matches the file names and matrix columns.
• Biological group, treatment, genotype, tissue, time point, or disease label.
• Biological replicate definition and whether samples are paired or independent.
• Sex, age, batch, extraction date, library batch, array plate, and other known covariates.
• Subject, animal, litter, donor, or matched-pair identifiers when repeated measurements exist.
• Genome build, transcript annotation, array platform, and any region annotation used previously.
• Sequencing or array technology, conversion chemistry, enrichment method, and available QC metrics.
• The intended publication or exploratory goal.

Avoid encoding multiple variables in an opaque sample name such as A1_T2_old_batch. A separate metadata table is safer. If a variable is scientifically important, it must be available as a column in the design sheet.

How Far Can Previous Outputs Be Reused?

Previous reports, scripts, parameter files, and result tables are valuable because they explain how the supplied files were produced. Include them even when the new request is limited to a different contrast. A prior DMR list without its feature table and thresholds is not enough to reproduce the result, but it can reveal the earlier genome build, annotation, filtering logic, and sample exclusions.

For public datasets, provide the accession number, download date, processed-file source, and any local modifications. Public availability does not guarantee that a dataset can be combined with a new cohort. The same sample may appear under different identifiers, and a study may contain multiple tissues, batches, or experimental arms that are not interchangeable.

Run a file-integrity check before transfer. Confirm that compressed files open, checksums are available when possible, matrix headers match the sample sheet, row coordinates are unique, and no sample appears twice under a different name. For large projects, a small manifest with file size, checksum, sample ID, data level, and expected contrast can prevent a silent mismatch at the start of analysis.

If the team wants a report audit rather than a full reanalysis, mark that explicitly. An audit can compare the prior design, filters, statistical thresholds, and annotation choices against the supplied outputs. It should not be described as a rerun of raw-read QC unless the raw data are available.

Rebuild the Contrast and Reference Groups

Write each contrast in plain language and in model-ready language. "reatment versus control"is incomplete if the samples also differ by sex, batch, tissue, or donor.

If the original analysis used different labels, do not silently reconstruct them. Provide the original design sheet and state which labels are trusted. A DMR analysis service can review ambiguous metadata, but it cannot recover a missing experimental relationship from file names.

Genome Build, Annotation, and Coordinate Compatibility

Coordinates are only meaningful within a reference frame. A BED file on hg19, a methylation matrix annotated to hg38, and a gene list based on a current transcript release cannot be combined safely without a documented lift-over or reannotation step.

Check:

• Genome assembly and chromosome naming convention.
• Whether unplaced or alternative contigs were included.
• Transcript and gene annotation release.
• Probe manifest version for array data.
• Whether the DMR coordinates are CpG-centered, region-centered, or already annotated.
• Whether the same reference was used for all samples.

Annotation is an interpretation layer, not proof of function. A DMR near a gene can be reported as a candidate regulatory association, but the nearest gene is not automatically the causal target.

What Can Be Audited Without Raw Reads?

Processed data can still be useful, but the report should distinguish reproducible QC from inherited QC.

When only a beta matrix is supplied, do not write that the raw sequencing quality "assed"unless an original QC report is available and traceable. State which checks were repeated, which were inherited, and which could not be performed.

Figure 3: Data availability defines the audit boundary for an outsourced reanalysis.

Choose a Reanalysis Scope That Matches the Evidence

A standard outsourced DMR analysis may include sample-level QC review, feature filtering, normalization appropriate to the data type, primary contrasts, DMC/DMR testing, genomic annotation, enrichment analysis, and figures such as heatmaps, volcano plots, regional plots, and principal-component analysis.

Advanced scope should be defined separately when the project needs paired or repeated-measures models, cell-composition adjustment, batch correction, time-course contrasts, allele-aware interpretation, cross-study integration, custom region sets, or methylation-expression integration. More model terms do not automatically make a result better: every added covariate needs enough information to estimate its effect.

Annotate DMRs Without Overstating Function

Annotation should follow the region definition and genome build used for the analysis. A DMR can overlap a promoter, gene body, UTR, CpG island, enhancer, or more than one category. When a region overlaps several features, retain the full overlap rather than assigning only the nearest gene.

Enrichment analysis also needs a background. If the test set contains only CpGs retained after coverage filtering or only probes represented on an array, using the whole genome as the background can create biased enrichment. The report should state whether the background was all assayed features, all tested regions, or a project-specific universe.

For a reanalysis, keep discovery and interpretation separate. First establish which DMRs are supported by the supplied data and design. Then add pathway, gene-set, regulatory-feature, or methylation-expression context. An enriched pathway can help prioritize follow-up, but it does not repair missing metadata or prove that the closest gene mediates the DMR.

Reanalysis Outputs and Decision Notes

Agree on deliverables before analysis begins. A practical report package can include:

• Data-inventory and compatibility assessment.
• Sample manifest and design matrix used for each contrast.
• QC summary with inherited versus rerun checks.
• Filtered CpG or probe tables and coverage/missingness summaries.
• DMC and DMR tables with effect size, direction, region coordinates, support metrics, and adjusted significance values.
• Annotation to promoters, gene bodies, UTRs, CpG islands, shores, enhancers, or a project-specific region set.
• Enrichment results with the tested background defined.
• Visualization files and an interpretation note describing boundaries.
• Reproducibility records for the reference genome, annotation, software, and parameters.

The final report should also state the provenance of every major result. A rerun from FASTQ or IDAT files can support a broader QC statement than a result derived from a beta matrix, but neither data level removes the need to document the contrast and sample manifest. For each DMC or DMR table, identify whether the feature was recalculated, inherited from a previous report, or supplied by the project team. Record the filtering universe, annotation release, tested background, and any missing metadata that limits interpretation. This makes an outsourced analysis auditable and gives the project team a clear basis for deciding whether to recover raw files, add validation, or proceed with the available evidence.

Before delivery, ask whether the requested output is a reanalysis, a report audit, or a focused interpretation package. A reanalysis starts from the available data level and reruns the defined processing steps. A report audit checks whether the supplied design, filters, contrasts, and reported candidates are internally consistent. A focused interpretation package may begin from a candidate table and add annotation, pathway context, or validation prioritization, but it should not be described as an independent QC rerun. Naming the scope at the start prevents a polished report from being mistaken for evidence that the retained files cannot provide.

The same scope label should appear in the handoff summary and figure captions. A reader should be able to tell whether a plot reflects raw-read processing, matrix-level analysis, or interpretation of supplied results without reconstructing the workflow from filenames.

When Existing Data Are Actionable—and When They Are Not

Suitable when

• Raw reads, methylation calls, coverage tables, IDAT files, or a well-documented matrix are available.
• Sample IDs and biological group labels can be reconciled.
• The genome build and array platform are known.
• The primary contrast and publication goal are defined.
• The team accepts a data-level-specific QC statement.

Consider recovery or new data when

• Only a screenshot or a prefiltered list of "ignificant DMRs"remains.
• Sample identities, group labels, or paired relationships cannot be reconstructed.
• The genome build is unknown and coordinate conversion would be speculative.
• There are no biological replicates for the intended population-level claim.
• The project requires raw-read QC that cannot be repeated from the available files.

Use the same provenance labels in tables, figures, and the narrative so downstream readers can distinguish rerun results from inherited project outputs. This boundary should remain explicit.

Existing-Data Review Handoff Checklist

Prepare an existing-data inventory containing:

• File type: FASTQ, BAM, coverage table, beta matrix, M-value matrix, IDAT, or prior result table.
• Genome build: assembly, chromosome naming, and annotation release.
• Sample count: total samples and samples per group.
• Group design: primary contrast, reference group, pairing, and secondary contrasts.
• Raw-data availability: which samples have FASTQ or IDAT files and which only have processed data.
• Required outputs: DMR tables, annotation, enrichment, figures, integration, or report audit.

Upload a Data Inventory for Feasibility Review with the file type, genome build, sample count, group design, raw-data availability, and required outputs.

FAQ

Conclusion

Outsourced DMR analysis works best when the project starts with a transparent inventory. Send the files you have, the design you intend, and the output you need. The feasibility review can then separate a full reanalysis from a matrix-level statistical review, identify missing QC, and define which conclusions are supportable.

Upload a Data Inventory for Feasibility Review

References

1. Hansen KD, Langmead B, Irizarry RA. BSmooth: from whole genome bisulfite sequencing reads to differentially methylated regions. Genome Biology. 2012;13:R83.
2. Feng H, Conneely KN, Wu H. A Bayesian hierarchical model to detect differentially methylated loci from single nucleotide-resolution sequencing data. Nucleic Acids Research. 2014;42(8):e69.
3. Peters TJ, Buckley MJ, Statham AL, et al. De novo identification of differentially methylated regions in the human genome. Epigenetics & Chromatin. 2015;8:6.
4. Peters TJ, Buckley MJ, Chen Y, et al. Calling differentially methylated regions from whole genome bisulphite sequencing with DMRcate. Nucleic Acids Research. 2021;49(19):e109.

Research Use Only Statement

For research purposes only. Not intended for clinical diagnosis, treatment, or individual health assessments.