Exome Panel, Whole Exome, or Whole Genome: What to Sequence and Why

Oak
exome sequencing genomics vcf coverage
A dark feathered winged creature on a steel stand, its body glowing in three nested zones of increasing brightness against black.

An “exome panel” is one of two things: a targeted gene panel (20 to 500 genes, sequenced deep) or the hybridization capture reagent used to pull down all ~20,000 protein-coding genes before sequencing. The three tiers are nested. A panel is a subset of the exome. The exome is roughly 1–2% of the genome (about 30–60 Mb of capture target depending on the kit). Whole genome sequencing covers everything, coding and not, at uniform depth. For a technically fluent person who wants to work with their own data, we recommend whole genome sequencing at 30× and treating the exome as a region filter you apply in software, not a decision you make at the wet-lab bench. If your question is clinical and specific (a family history of hereditary cancer, a suspected immune disorder), a clinician-ordered panel or clinical exome is the right instrument, and it needs to be ordered and interpreted by a clinician.

What the three tiers give you

Targeted panel. Amplicon or capture-based, 20–500 genes, typically 500–1000× mean depth on target. The depth buys you two things: confident detection of low-allele-fraction variants (mosaicism, tumor subclones) and clean copy-number calls within the panel. What you lose is everything outside the target, which you cannot recover without re-sequencing. In 878 patients with suspected primary immunodeficiency, a targeted panel and whole-exome sequencing produced comparable diagnostic yield, and the panel was cheaper per patient, with the caveat that panels miss the variants in genes nobody thought to include.1

Whole exome. Hybridization capture (IDT xGen Exome Hyb Panel v2, Twist Comprehensive Exome, Agilent SureSelect) against 33–43 Mb of coding target, sequenced to 100× mean depth. The capture step is where the problems live. Coverage is uneven across the target: GC-rich first exons, segmental duplications, and paralogous genes drop out, and 5–10% of the nominal target commonly falls below 20× even at high mean depth. Exome capture also performs poorly on copy-number variants and structural variants, because read depth is the only signal and capture efficiency varies probe to probe.2 Non-coding regions, deep intronic splice variants, promoters, and most of the mitochondrial genome are absent by design.

Whole genome. No capture, no probe bias. PCR-free library prep, 30× mean depth, roughly 100 Gb of sequence. You get uniform coverage, better structural variant and copy-number detection from paired-end and split-read signal, full mitochondrial genome at high effective depth (mtDNA copy number is high, so a 30× nuclear genome yields hundreds of × on chrM), and coverage of the regulatory and intronic space you will want in five years when the annotation improves.

The practical argument for genome over exome is not per-base coverage of coding regions. At 30× WGS versus 100× WES, the exome’s mean depth on target is higher. The argument is that WGS covers coding regions more evenly, so fewer exons fall into the sub-20× hole, and everything else comes free.

Cost stated

Consumer and research whole genome sequencing at 30× runs $200–$1,000 depending on provider and turnaround. Research-grade whole exome is $300–$600. Clinical whole exome with trio analysis, variant curation by a certified lab, and a signed report is $2,000–$5,000 in the US, and that price is mostly interpretation labor and CAP/CLIA overhead, not reagents. A clinical gene panel is $500–$2,500. Insurance coverage tracks clinical indication, not curiosity.

If you are paying out of pocket to get data you will analyze yourself, the marginal cost of genome over exome is now small enough that exome-only is hard to justify. That was not true in 2015, and much of the ranking content on this topic is written from that era’s cost structure.2

What you should ask for as deliverables

Insist on these, in writing, before you pay:

  • FASTQ, gzipped, paired-end R1/R2. This is the raw read data. Everything downstream can be regenerated from it. Roughly 80–120 GB for a 30× genome.
  • CRAM aligned to GRCh38 (or T2T-CHM13 if offered), with the reference version stated explicitly. CRAM is reference-compressed BAM, about 60% smaller. A 30× CRAM is 15–20 GB.
  • gVCF, not just VCF. The gVCF records reference blocks with confidence, so you can distinguish “reference at this position” from “no coverage here.” Plain VCFs lose that, and the difference matters when you are checking whether a specific pharmacogenomic or carrier site was interrogated.
  • A coverage report: mean depth, percent of target ≥10× and ≥20×, and a BED file of callable regions.

If a provider will only hand you a PDF report, the data is not yours in any useful sense.

Applying an exome filter to genome data

The reason we call the exome a software decision: you can subset a genome VCF to any panel or exome target in seconds.

Get a target BED. For the exome, use your capture vendor’s target file or the Ensembl/GENCODE CDS regions. For a specific panel, pull the gene list and convert to coordinates.

# subset a genome VCF to a BED of coding target
bcftools view -R exome_targets_GRCh38.bed \
  -Oz -o sample.exome.vcf.gz sample.g.vcf.gz
bcftools index sample.exome.vcf.gz

# check depth across the same regions from the CRAM
mosdepth --by exome_targets_GRCh38.bed --fasta GRCh38.fa \
  --threads 4 sample sample.cram

mosdepth writes sample.regions.bed.gz with mean depth per interval. Sort it ascending and look at the tail: those are the regions where a negative call means nothing.

For annotation, we use Ensembl VEP with the --plugin system rather than a web uploader, because you control the versions:

vep -i sample.exome.vcf.gz -o sample.vep.vcf --vcf \
  --cache --offline --assembly GRCh38 \
  --everything --pick_allele_gene \
  --plugin CADD,whole_genome_SNVs.tsv.gz \
  --plugin SpliceAI,snv=spliceai_scores.raw.snv.hg38.vcf.gz

--everything turns on SIFT, PolyPhen, gnomAD frequencies, and canonical transcript flags. --pick_allele_gene gives one consequence per allele per gene, which keeps the output joinable. Then filter on gnomAD allele frequency (AF < 0.001 for dominant conditions, < 0.01 for recessive) and consequence severity. Expect a 30× genome to yield 4–5 million variants, of which a few hundred survive a rare-and-coding filter, of which essentially all are benign or uncertain.

That last point is the one worth internalizing. Population-scale variant interpretation is dominated by variants of uncertain significance, and the base rate of a given rare missense variant being pathogenic is low. Reporting an uncertain finding as meaningful is a documented harm of broad sequencing, along with incidental findings you did not ask for and did not consent to in any considered way.3 Decide in advance which secondary findings you want to see, and take anything that looks actionable to a genetic counselor rather than to a search engine.

Where the exome stops being the right unit

Two cases where panels and exomes both fail and genome does not.

Structural and copy-number variation. Capture depth is a noisy CNV signal. WGS gives you read-pair orientation, split reads, and uniform depth, so callers like Manta, DELLY, and GRIDSS work as intended.

Somatic and tumor profiling. In advanced solid tumors, whole-exome and whole-transcriptome profiling identified clinically relevant alterations that 50-gene targeted panels missed, and adding RNA sequencing caught fusions and expression-level events invisible to DNA panels.4 The general lesson transfers: narrow targets answer only the question you framed when you chose the target.

Regulatory and splice variants. A deep intronic variant creating a cryptic splice acceptor is invisible to exome capture and findable with WGS plus SpliceAI. RNA sequencing on the same person closes the loop by showing whether the predicted splice event happens.

Questions people also ask

How much does whole exome testing cost? Research-grade whole exome with raw data is $300–$600. Clinical whole exome with a signed interpretive report, and usually parental samples for trio analysis, is $2,000–$5,000 in the US. The gap is interpretation and regulatory overhead, not sequencing chemistry.

What’s the difference between exome and genome testing? The exome is the protein-coding fraction, about 1–2% of the genome, isolated by hybridization capture and typically sequenced to 100×. The genome is everything at uniform 30× with no capture step, which gives better copy-number and structural variant detection, full mitochondrial coverage, and coverage of intronic and regulatory sequence. Exome files are smaller (6–10 GB CRAM versus 15–20 GB) and cheaper to store and process.

Should I order a panel or an exome? If you have a specific clinical question with a well-characterized gene set, a panel is cheaper and produces fewer uncertain variants to adjudicate, with comparable diagnostic yield in at least one large cohort.1 If the phenotype is ambiguous, broader sequencing avoids serial re-testing. Either way, that order goes through a clinician.

Can I get exome results from whole genome data? Yes, and the subset is exact once you have a target BED, which is why we sequence the genome and filter in software. What you cannot do is go the other direction.

Oak builds longitudinal molecular profiles of individuals: whole-genome sequencing, RNA sequencing, proteomics, blood biomarkers, and continuous glucose data, integrated into one model of you. Build your profile.

Footnotes

  1. Craig D. Platt, Fatima Zaman, Wayne Bainter, et al. Efficacy and economics of targeted panel versus whole-exome sequencing in 878 patients with suspected primary immunodeficiency. Journal of Allergy and Clinical Immunology, 2021. https://doi.org/10.1016/j.jaci.2020.08.022 ↩ ↩2

  2. Amanda Warr, Christelle Robert, David Hume, et al. Exome Sequencing: Current and Future Perspectives. G3 Genes|Genomes|Genetics, 2015. https://doi.org/10.1534/g3.115.018564 ↩ ↩2

  3. Wim Pinxten, Heidi Carmen Howard. Ethical issues raised by whole genome sequencing. Best Practice & Research Clinical Gastroenterology, 2014. https://doi.org/10.1016/j.bpg.2014.02.004 ↩

  4. Jean-Paul De La O, David W Hall, Jess R Hoag, et al. Comparison of whole-exome, whole-transcriptome genomic profiling and targeted sequencing with 50-gene panels in advanced solid tumors. Scientific Reports, 2026. https://doi.org/10.1038/s41598-026-64008-7 ↩