What Whole-Genome Sequencing Is Used For, and What You Can Do With Your Own Data

Whole-genome sequencing reads essentially all 3.1 billion bases of your nuclear genome plus the mitochondrial genome. Its applications fall into four buckets. These are diagnosing rare and undiagnosed disease, profiling tumors (tumor/normal somatic calling), pharmacogenomics and carrier status, and research-grade personal analysis. That last bucket covers polygenic scores, ancestry, phasing and structural variation. A fifth bucket, pathogen typing and outbreak tracing, uses the same chemistry on bacterial genomes and has nothing to do with your genome. Most search results conflate the two. This page is about the human application, and specifically about what you can do with the files once you have them.

What WGS covers that exomes and arrays don’t

An exome (WES) captures roughly 1–2% of the genome, about 35 Mb of coding sequence, using hybridization probes. Capture efficiency varies with GC content, so exons in GC-rich first exons and promoters drop out, and coverage across the target is uneven enough that you need 100x mean to get the same callable fraction that WGS gets at 30x. A PCR-free WGS library has near-uniform depth, which is what makes copy-number and structural variant calling from read depth and split reads possible at all.

WGS makes these practical additions over WES.

• non-coding and deep intronic variants
• balanced translocations and inversions
• mobile element insertions
• mitochondrial variants at useful heteroplasmy resolution
• short tandem repeat expansions
• uniform enough depth for CNV calling

Reviews of clinical implementation consistently put WGS ahead of exome on diagnostic yield in undiagnosed disease, mostly from these classes of variant rather than from coding SNVs the exome missed.¹² A genotyping array, by contrast, interrogates ~700k pre-chosen positions and imputes the rest. It cannot find a variant nobody put on the chip, which is most of what matters in rare disease.

The clinical applications, ranked by how well they work

Rare and undiagnosed disease is the application with the strongest evidence. A trio (proband plus both parents) at 30x lets you call de novo variants and resolve compound heterozygotes by phase, and it is the design used in most successful genetics clinics.³ Filtering is phenotype-driven, with HPO terms in and a gene panel out. Rare variants inside that panel come next, then everything else.

Oncology WGS is a tumor/normal design at higher depth (60–100x tumor, 30x normal) used to call somatic SNVs, indels, copy number, and mutational signatures. It is ordered and interpreted by an oncology team. It is not run as a personal project.

Pharmacogenomics is where short-read WGS is genuinely useful and genuinely hard. Star-allele calling for CYP2D6 requires resolving a segmental duplication with a near-identical pseudogene (CYP2D7) plus hybrid alleles and copy number. Generic variant callers get this wrong. Use a purpose-built caller.

Secondary findings, the ACMG-style scan of medically actionable genes, is the application people most often overestimate. Dewey and colleagues sequenced adults and curated the results by hand. They found that a substantial fraction of positions inside inherited-disease genes were not covered well enough to call confidently, while manual interpretation of candidate variants took many hours per genome.⁴ The year was 2014, and coverage has improved, but the shape of the problem has not. The bottleneck is interpretation. Sequencing is not the limiting step.

What you get, and in what format

A 30x PCR-free human genome delivers roughly:

• FASTQ: two gzipped files, ~40–50 GB each for 2x150 bp reads
• CRAM aligned to GRCh38: 15–25 GB depending on compression settings (BAM is roughly 3x larger)
• gVCF: the per-base record you need if you ever want to joint-call with relatives
• VCF: ~4–5 million variants against GRCh38, of which ~600k are small indels

Insist on FASTQ or CRAM in your deliverable. A VCF alone is a derived product of somebody else’s pipeline choices, and you cannot re-call structural variants, repeat expansions, or mitochondrial heteroplasmy from it.

The pipeline we would run

Reference: GRCh38 no-alt analysis set. Use it even though T2T-CHM13 is a better assembly, because ClinVar, gnomAD, and every annotation source you will touch are keyed to GRCh38 coordinates. Re-align to CHM13 as a second pass if you care about centromeric and acrocentric regions.

bwa-mem2 mem -t 32 -K 100000000 -Y \
  -R '@RG\tID:L1\tSM:sample\tLB:lib1\tPL:ILLUMINA' \
  GRCh38_no_alt.fa R1.fq.gz R2.fq.gz \
  | samtools sort -@ 8 -m 4G -o sample.bam
gatk MarkDuplicates -I sample.bam -O sample.md.bam -M dup.txt

-K 100000000 fixes the batch size so the output is deterministic across thread counts. -Y uses soft clipping for supplementary alignments, which SV callers need.

Small variants: DeepVariant with --model_type=WGS is our default. It outperforms GATK HaplotypeCaller on indels on Illumina WGS and needs no base quality score recalibration. Keep the gVCF output (--output_gvcf) so you can joint-call later with GLnexus if a relative sequences.

Then, in rough order of yield per hour spent:

• Structural variants: Manta plus a read-depth caller, intersected. Expect hundreds of false positives per genome from either alone.
• Repeat expansions: ExpansionHunter with the published variant catalog. Short reads underestimate expansions longer than the fragment size, so treat a positive as a screen. A positive here is not a result.
• Mitochondrial: gatk Mutect2 --mitochondria-mode against chrM, then filter NUMT contamination.
• HLA typing: HLA-LA or similar graph-based typer.
• Annotation: vep --offline --cache --assembly GRCh38 --everything --plugin SpliceAI --plugin CADD, plus gnomAD v4 allele frequencies and a current ClinVar VCF.

For a first pass at your own genome, filter to ClinVar pathogenic and likely-pathogenic with gnomAD popmax AF below 0.1%, drop anything with a conflicting-interpretation status, and look at every remaining call in IGV. You will typically be left with a few dozen rows. Most of them will be carrier-state recessive variants.

The failure modes that matter

A pathogenic-looking variant in your VCF is not a diagnosis, and confirming or refuting one requires a clinical geneticist and a CLIA-accredited lab. Clinical WGS best practices exist precisely because research pipelines and clinical reporting diverge: orthogonal confirmation, phenotype-anchored filtering, and explicit variant classification are all part of the clinical workflow and none of them are part of a bcftools command line.⁵

Penetrance is the deepest problem. Variant classifications were largely built from families ascertained because they were sick. When the same alleles are looked up in unselected population cohorts, the probability of disease given the genotype is often far lower than the family studies implied, and interpretation at population scale needs allele frequency and functional data rather than pedigree data alone.⁶ Reading your own ClinVar hits without that context reliably produces alarm about variants that will never do anything.

Coverage gaps are structural. They do not fall at random. Segmental duplications cannot be resolved with 150 bp reads. Examples are SMN1/SMN2, PMS2/PMS2CL, CYP2D6/CYP2D7 and the alpha-globin cluster. If your question lives in one of those loci, you need a targeted assay or long reads.

Then there is the volume of uncertainty. Every genome carries variants of uncertain significance in disease-associated genes, and the burden of incidental information WGS creates is the main reason professional societies have argued for pre-test counseling and explicit opt-in choices about what gets returned.⁷ Decide before you look what you want to know.

Questions people also ask

What does WGS stand for? Whole-genome sequencing: reading the full nuclear genome and the mitochondrial genome, typically to a mean depth of 30x on short-read Illumina instruments.

What’s the difference between WGS and WES? WES sequences only the protein-coding exons (~1–2% of the genome) after hybridization capture. WGS sequences everything, without capture bias, which is what makes non-coding variants, structural variants, repeat expansions, and mitochondrial calls accessible.¹

Is WGS the same as NGS? No. NGS (next-generation sequencing) is the class of massively parallel sequencing technology. WGS is one application of it, alongside exomes, RNA-seq, targeted panels, and microbial sequencing.⁸

Is WGS covered by insurance? Coverage generally applies to diagnostic indications with a clinician’s order, typically for an undiagnosed suspected genetic condition or in oncology, and it varies by payer and country.² Elective sequencing of a healthy adult is usually paid out of pocket. Ask your insurer for the specific CPT code before assuming.

Can I interpret my own genome? You can run the pipeline, call the variants, and understand the statistics better than most clinicians. Classification of a specific variant as the cause of a specific condition is a different job, done against ACMG criteria with lab-grade confirmation, and it requires a clinical geneticist.⁵

Footnotes

1. 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

2. Frederik Otzen Bagger, Line Borgwardt, Andreas Sand Jespersen, et al. Whole genome sequencing in clinical practice. BMC Medical Genomics, 2024. https://doi.org/10.1186/s12920-024-01795-w ↩ ↩²

3. Pamela Fraser, Michael F. Gutzeit, Jeremy M. Harris, et al. Successful Application of Whole Genome Sequencing in a Medical Genetics Clinic. Journal of Pediatric Genetics, 2016. https://doi.org/10.1055/s-0036-1593968 ↩

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. 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

6. 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 ↩

7. 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 ↩

8. 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 ↩