What a WGS Report Is, and Why the Files Matter More

A WGS report is the interpretation document a clinical lab produces after sequencing your genome: usually two to six pages listing a small number of variants the lab classified as pathogenic or likely pathogenic and judged relevant to the question that was asked, plus methods, coverage statistics, and disclaimers. WGS stands for whole genome sequencing. The report is not your genome. Your genome, as data, is a set of files: FASTQ or CRAM with the reads, a VCF with small variants, usually separate VCFs for structural variants and copy number, and a QC bundle. The report describes maybe five variants. The small-variant VCF holds roughly four to five million. If you want to work with your own molecular data, the report is the smallest artifact you will receive and the files are where nearly everything lives.

What is in a clinical WGS report

A report follows a fairly standardized shape, and the best-practices guidance from the Medical Genome Initiative describes it well. It has primary findings tied to the indication and a variant table with these columns.¹

• gene
• transcript
• HGVS nomenclature
• zygosity
• inheritance
• an ACMG/AMP classification with the evidence codes that produced it

A well-built report states the reference build (GRCh38 in anything recent), the transcript used for each call (MANE Select, ideally), the mean depth, and the fraction of the genome or of a defined gene list covered at ≥10x and ≥20x.

Two categories usually sit below the primary findings. Secondary or incidental findings, reported on an opt-in basis against the ACMG gene list (around 80 genes covering hereditary cancer syndromes, cardiomyopathies and arrhythmias, familial hypercholesterolemia, malignant hyperthermia, and a few others). And carrier status for recessive conditions. Whether to return these has been argued over for a decade, and European professional guidance has been notably more conservative than American, favoring targeted analysis and explicit consent over broad opportunistic screening ².

What ends up in the report is a filtering decision. The content is not a discovery. The lab applied a phenotype-driven gene list, filtered by population frequency, kept variants meeting classification criteria, and manually curated the survivors. Curation is expensive human work, which is why the output is short.

What the report leaves out

Everything classified as a variant of uncertain significance, in most reports. Everything in a gene not on the ordered panel. Nearly all non-coding variation, because we still cannot interpret most of it. The interpretive bottleneck is not sequencing but the mapping from variant to function, and that gap is what large-scale functional genomics, expression QTL maps, and saturation mutagenesis are trying to close ³.

Coverage gaps get understated. Dewey and colleagues sequenced twelve adults on two platforms. They found that a meaningful share of genes known to cause inherited disease were not covered well enough for confident variant calling, and that concordance between platforms broke down badly for insertions and deletions ⁴. Longer reads and better alignment have improved this, but the structural problem persists. Segmental duplications, GC-extreme promoters, and genes with pseudogenes (PMS2, GBA1, SMN1/SMN2 and CYP2D6) are hard for short-read WGS and are frequently reported without an explicit caveat. If a report says “negative,” read the coverage appendix before you believe it about a specific gene.

Population reference matters too. Allele frequency filters only work if your ancestry is represented in the reference database. Deep sequencing of 1,070 Japanese individuals produced millions of variants absent from then-current references and a population-specific imputation panel that substantially improved calls for that group ⁵. If you are from an underrepresented population, expect more variants of uncertain significance and more frequency-based misclassification.

The files to ask for, and what to do with them

Ask for, in order of value:

1. CRAM plus its .crai index, aligned to GRCh38 with alt contigs, roughly 15–25 GB at 30x. CRAM is reference-compressed, so keep the exact reference FASTA and its .fai alongside it or the file is unreadable.
2. FASTQ, ~100 GB gzipped. Only worth storing if you plan to realign, for example to T2T-CHM13 or with a pangenome reference.
3. Small-variant VCF or gVCF, bgzipped and tabix-indexed. The gVCF is what lets you distinguish “reference at this position” from “no data at this position,” which is the difference between a real negative and a coverage gap.
4. SV and CNV VCFs, from Manta/Delly/GRIDSS and a read-depth caller.
5. QC: the samtools stats or Picard CollectWgsMetrics output, cross-sample contamination estimate (VerifyBamID FREEMIX), and sex/ancestry checks.

A workable first pass:

# sanity: depth and callability
mosdepth --by 1000 --fast-mode sample sample.cram

# how many variants, and do they look like a real human genome?
bcftools stats sample.vcf.gz | grep -E '^SN|^TSTV'

Expect roughly 4–5 million small variants against GRCh38, a transition/transversion ratio near 2.0–2.1 genome-wide, and heterozygous/homozygous ratios that vary by ancestry. A Ti/Tv of 1.6 means you are looking at noise. Then annotate:

vep -i sample.vcf.gz --cache --assembly GRCh38 --everything \
    --mane_select --pick_allele_gene --plugin CADD \
    --custom gnomad.genomes.v4.1.sites.vcf.gz,gnomAD,vcf,exact,0,AF_grpmax \
    --vcf -o annotated.vcf.gz --compress_output bgzip

Then filter hard before you look at anything: gnomAD popmax AF below 0.1% for dominant conditions, restrict to a gene list you chose in advance, require FILTER=PASS, and require DP ≥ 10 and GQ ≥ 20. Without a pre-specified gene list you will find something alarming in every genome, because everyone carries dozens of rare protein-truncating variants.

Two specialist tools are worth running separately. ExpansionHunter for repeat expansions, which standard callers miss entirely. And Cyrius or StellarPGx for CYP2D6, since the gene’s structural complexity defeats generic pipelines. Pharmacogenomic star alleles from a naive VCF parse are frequently wrong.

Nothing you compute at home should change what you take, stop, or do medically. Confirmatory testing in a CLIA/CAP laboratory and interpretation with a genetic counselor or medical geneticist is the path for any finding that matters. Research-grade calls have real false-positive rates, and phasing, penetrance, and family history all change what a variant means.

Where the report is still the right tool

For a specific diagnostic question, a clinical lab with curated evidence, orthogonal confirmation, and a reportable ACMG classification does something you cannot replicate with a laptop ¹. WGS as a first-tier test has a diagnostic yield in rare disease that exceeds exome and panel testing, and its clinical implementation is now practical at scale ⁶. Sequencing cost is no longer the constraint ⁷. Interpretation is.

The useful model: treat the report as a supervised, legally accountable summary of a narrow question, and treat the files as a durable asset you will re-analyze in five years when the annotation databases and functional maps are better. The reads do not expire. The interpretation does.

Questions people also ask

What does WGS stand for? Whole genome sequencing. In practice it means short-read sequencing of the full genome at a target mean depth, typically 30x for germline work, 100x or more for somatic and cancer applications. Long-read WGS (PacBio HiFi, ONT) is a different method with much better structural variant and repeat resolution.

What does WGS tell you? Nearly all single-nucleotide variants and small indels, most copy number changes, many structural variants, mitochondrial variants, and repeat expansions if you run the right caller. It does not tell you methylation (unless long-read), does not tell you gene expression, and does not tell you what most non-coding variants do ³.

Is WGS covered by insurance? Sometimes, for a diagnostic indication. Examples are a child with a suspected genetic disorder, an undiagnosed condition after a negative panel, and certain cancers. Elective baseline sequencing in a healthy adult is generally not covered. Coverage policy varies by payer and country and changes often, so check the specific policy rather than a general claim.

What is a WGS test? It starts with blood or saliva, then DNA extraction, library prep and sequencing. Alignment, variant calling and filtering follow, and then curation and a report. The wet-lab portion is commoditized. The variation in quality between providers is almost entirely in the bioinformatics and the interpretation, plus whether they give you the primary data ⁶.

Footnotes

1. Christina Austin‐Tse, Vaidehi Jobanputra, Denise Perry, et al. Best practices for the interpretation and reporting of clinical whole genome sequencing. npj Genomic Medicine, 2022. https://doi.org/10.1038/s41525-022-00295-z ↩ ↩²

2. Carla van El, Martina C. Cornel, Pascal Borry, et al. Whole-genome sequencing in health care. European Journal of Human Genetics, 2013. https://doi.org/10.1038/ejhg.2013.46 ↩

3. Tuuli Lappalainen, Alexandra J. Scott, Margot Brandt, et al. Genomic Analysis in the Age of Human Genome Sequencing. Cell, 2019. https://doi.org/10.1016/j.cell.2019.02.032 ↩ ↩²

4. Frederick E. Dewey, Megan E. Grove, Cuiping Pan, et al. Clinical Interpretation and Implications of Whole-Genome Sequencing. JAMA, 2014. https://doi.org/10.1001/jama.2014.1717 ↩

5. Masao Nagasaki, Jun Yasuda, Fumiki Katsuoka, et al. Rare variant discovery by deep whole-genome sequencing of 1,070 Japanese individuals. Nature Communications, 2015. https://doi.org/10.1038/ncomms9018 ↩

6. Petar Brlek, Luka Bulić, Matea Bračić, et al. Implementing Whole Genome Sequencing (WGS) in Clinical Practice: Advantages, Challenges, and Future Perspectives. Cells, 2024. https://doi.org/10.3390/cells13060504 ↩ ↩²

7. Sang Tae Park, Jayoung Kim. Trends in Next-Generation Sequencing and a New Era for Whole Genome Sequencing. International Neurourology Journal, 2016. https://doi.org/10.5213/inj.1632742.371 ↩