What Whole Exome Sequencing (WES) Measures, and When to Choose It Over a Genome

Whole exome sequencing (WES) is a next-generation sequencing method that enriches and sequences only the protein-coding portion of the genome, roughly 1 to 2 percent of the 3.1 gigabase human sequence, before alignment and variant calling. In genetic testing it is used to find small variants (single-nucleotide variants and short insertions or deletions) in the exons of about 19,000 genes, which is where the large majority of currently interpretable disease-associated variants sit ¹. Our view, if you are paying for your own data and intend to analyze it yourself: sequence the genome at 30x rather than the exome, unless you specifically need very deep coverage of coding bases (mosaicism, tumor work, family studies at high depth) or you are constrained to a few hundred dollars of sequencing. The reasoning below is about coverage uniformity, callable regions, and what you can re-analyze later.

How the enrichment works and what the target really is

WES is a hybrid-capture experiment. Genomic DNA is fragmented to roughly 150 to 250 bp, adapters are ligated, and the library is hybridized to a pool of biotinylated 80- to 120-mer oligonucleotide probes complementary to exonic sequence. Probe-bound fragments are pulled down on streptavidin beads, washed, amplified, and sequenced, usually 2x150 bp on an Illumina instrument. The product is a library in which exonic fragments are enriched perhaps 50- to 100-fold relative to the rest of the genome, though never to the exclusion of everything else.

The word “exome” is a vendor definition. It is not a biological constant. Twist’s Exome 2.0 targets about 36.8 Mb, IDT’s xGen Exome Research Panel v2 about 34 Mb, and older Agilent SureSelect designs range from 38 to over 60 Mb depending on whether untranslated regions and non-coding RNAs are included. A “clinical exome” is a different product again: capture over the whole exome but interpretation restricted to a curated disease gene list ². Before any analysis, download the manufacturer’s target BED file for your exact kit and lot, and check that the gene you care about is in it. Two kits can differ by several thousand exons, and a gene absent from the probe set will appear only as sparse off-target coverage.

Coverage statistics that determine whether a result means anything

Depth is reported as a mean because means look good. What matters is uniformity. A typical WES run at 100x mean depth places 95 to 98 percent of target bases above 20x. The remaining 2 to 5 percent is not random. The shortfall sits in GC-rich first exons, promoter-proximal coding sequence, and regions with high homology to pseudogenes. That structure is the central limitation of capture, and it is why comparisons with PCR-free genome libraries consistently favor the genome even inside the exome itself. Belkadi and colleagues sequenced the same individuals by both methods and found that 30x WGS recovered exonic variants that 70x to 100x WES missed, with better coverage uniformity and less GC bias ³. Meienberg and colleagues reached the same conclusion using PCR-free WGS, reporting more complete coverage of coding regions than any of the capture kits they tested ⁴.

Compute these numbers yourself rather than trusting a summary PDF. With mosdepth:

mosdepth --by targets.bed --thresholds 1,10,20,30,50 \
  --fast-mode sample sample.cram

Then look at sample.thresholds.bed.gz per exon. A per-genome view is not what you want here. We care about four metrics. The first is the fraction of target at ≥20x. The second is the fold-80 base penalty, the extra sequencing required to bring 80 percent of targets to the mean, which should be under 2.0. The third is duplicate rate, 5 to 20 percent for a PCR-amplified exome. The fourth is on-target rate, 60 to 80 percent for a well-run capture. A fold-80 above 3 means the library was over-hybridized or input DNA was degraded, and no amount of downstream filtering will recover the dropout.

A pipeline that produces data you can reuse

The formats are the same as for any NGS application. Raw reads come as FASTQ and alignments as CRAM or BAM. Per-sample calls come as gVCF, and the final joint call set as VCF. Insist on FASTQ or CRAM from any vendor. A VCF alone is a derived artifact, and you will want to recall variants in three years with better software.

Alignment with bwa-mem2 mem -t 16 -K 100000000 -R '@RG\tID:1\tSM:sample\tPL:ILLUMINA' against GRCh38 with ALT contigs, followed by samtools sort and gatk MarkDuplicates, is still the reference-grade route. For calling, restrict to the padded target so you do not spend hours genotyping off-target noise:

bedtools slop -i targets.bed -g genome.txt -b 100 > targets.padded.bed
gatk HaplotypeCaller -R GRCh38.fa -I sample.cram \
  -L targets.padded.bed -ERC GVCF -O sample.g.vcf.gz

The 100 bp pad matters: capture fragments extend past probe boundaries, so canonical splice sites two to ten bases into the intron are well covered even when the BED file stops at the exon edge. Normalize before annotation (bcftools norm -m -any -f GRCh38.fa) because indel representation differences will silently break joins against ClinVar and gnomAD. Annotate with Ensembl VEP offline (--cache --offline --everything --pick_allele_gene) or with bcftools csq if you want something faster and simpler.

Expect on the order of 20,000 to 25,000 coding variants per exome. Roughly 10,000 to 12,000 of those are nonsynonymous, a few hundred are rare and predicted loss-of-function, and one or two might be individually interpretable. Filtering to allele frequency below 0.1 percent in gnomAD and to high-confidence LoF or ClinVar pathogenic entries is the standard first pass ¹. Interpretation of anything that survives that filter is clinical work and belongs with a board-certified medical geneticist or genetic counselor, who will also want orthogonal confirmation of any variant that would change medical management.

What WES cannot see

Capture defines a boundary, and everything outside it is invisible or unreliable. Deep intronic and regulatory variants, most structural variants, and balanced rearrangements are out of reach. Copy number can be inferred from read-depth ratios across samples using GATK gCNV, ExomeDepth, or CNVkit. The resolution is coarse, and the false-positive rate depends on having a matched panel of normals processed with the same kit and the same batch. Short tandem repeat expansions of the kind that matter in neurology require genome data and tools like ExpansionHunter. Pseudogene-heavy loci are poorly resolved by short-read capture and typically need targeted assays such as MLPA or long-read sequencing. Examples are PMS2, SMN1/SMN2, GBA, CYP21A2, and HBA1/HBA2. HLA typing from short reads is possible but demands specialized allele-aware callers rather than a standard reference-based pipeline ⁵.

This is why a focused gene panel is sometimes better than an exome. Panels deliver higher and more even depth over their targets, cost less, and have lower incidental-finding burden. The Saudi Mendeliome Group found that comprehensive panels outperformed clinical exome sequencing for Mendelian disease diagnosis on coverage and yield grounds ⁶, and panel-first strategies remain defensible in well-delineated clinical presentations ⁷.

When we would still choose WES

Three cases. First, depth: if you need 500x over coding bases to detect low-frequency mosaic or somatic variants, capture is the economical way to get there. Second, cohort scale: joint calling across hundreds of exomes remains cheaper in both sequencing and storage than the genome equivalent (about 6 to 10 GB of FASTQ per exome against 90 to 120 GB per 30x genome). Third, budget. Otherwise the genome gives strictly more information for a modest multiple of the price, covers the exome better than the exome does ³⁴, and does not require re-sequencing when a new kit or a new gene becomes relevant. For a healthy person sequencing themselves, the interpretive caveats are the same either way: most of what comes back is uncertain, and the value accumulates over years as databases improve ⁸⁹.

Questions people also ask

What is a WES test used for in genetic testing?

It is used to search the protein-coding regions of about 19,000 genes for small variants that could explain a suspected Mendelian condition, to characterize tumor coding mutations, and in research to scan cohorts for rare-variant associations. In diagnostic settings it is typically ordered after the clinical picture has narrowed the differential and a single-gene test or panel has been considered and rejected as too narrow ¹².

What is the WES test used for in NGS?

WES is one application of next-generation sequencing, defined by the capture step that precedes sequencing. The NGS instrument and chemistry are identical to those used for whole genome sequencing, RNA sequencing, or amplicon panels. What distinguishes WES is that library preparation includes hybrid capture with exon-targeting probes, so sequencing capacity is concentrated on roughly 1 to 2 percent of the genome ¹⁰.

How does WES compare with WGS for finding exonic variants?

Counterintuitively, whole genome sequencing at 30x detects exonic variants that exome sequencing at higher nominal depth misses, because PCR-free genome libraries give far more uniform coverage across GC content and do not suffer probe hybridization dropout ³⁴. WGS also captures copy number, structural, and non-coding variation that capture cannot reach.

Does a WES report tell me my disease risk?

A WES report lists variants and their classifications against curated databases. Translating that into risk for you as an individual requires family history, phenotype, and clinical judgment. It should be done with a genetics professional. Variant classifications are revised over time, so a report is a snapshot rather than a final answer ⁸.

Footnotes

1. Jacek Majewski, Jeremy Schwartzentruber, Emilie Lalonde, et al. What can exome sequencing do for you?. Journal of Medical Genetics, 2011. https://doi.org/10.1136/jmedgenet-2011-100223 ↩ ↩² ↩³

2. Hanns-Georg Klein, Peter Bauer, Tina Hambuch. Whole genome sequencing (WGS), whole exome sequencing (WES) and clinical exome sequencing (CES) in patient care. LaboratoriumsMedizin, 2014. https://doi.org/10.1515/labmed-2014-0025 ↩ ↩²

3. Aziz Belkadi, Alexandre Bolze, Yuval Itan, et al. Whole-genome sequencing is more powerful than whole-exome sequencing for detecting exome variants. Proceedings of the National Academy of Sciences, 2015. https://doi.org/10.1073/pnas.1418631112 ↩ ↩² ↩³

4. Janine Meienberg, Rémy Bruggmann, Konrad Oexle, et al. Clinical sequencing: is WGS the better WES?. Human Genetics, 2016. https://doi.org/10.1007/s00439-015-1631-9 ↩ ↩² ↩³

5. Kazuyoshi Hosomichi, Takashi Shiina, Atsushi Tajima, et al. The impact of next-generation sequencing technologies on HLA research. Journal of Human Genetics, 2015. https://doi.org/10.1038/jhg.2015.102 ↩

6. Saudi Mendeliome Group. Comprehensive gene panels provide advantages over clinical exome sequencing for Mendelian diseases. Genome Biology, 2015. https://doi.org/10.1186/s13059-015-0693-2 ↩

7. Christopher J. Klein, Tatiana M. Foroud. Neurology Individualized Medicine: When to Use Next-Generation Sequencing Panels. Mayo Clinic Proceedings, 2017. https://doi.org/10.1016/j.mayocp.2016.09.008 ↩

8. Noralane M. Lindor, Stephen N. Thibodeau, Wylie Burke. Whole-Genome Sequencing in Healthy People. Mayo Clinic Proceedings, 2017. https://doi.org/10.1016/j.mayocp.2016.10.019 ↩ ↩²

9. A.J. Marian. Sequencing Your Genome: What Does it Mean?. Methodist DeBakey Cardiovascular Journal, 2014. https://doi.org/10.14797/mdcj-10-1-3 ↩

10. Chee-Seng Ku, Dimitrios H Roukos. From next-generation sequencing to nanopore sequencing technology: paving the way to personalized genomic medicine. Expert Review of Medical Devices, 2013. https://doi.org/10.1586/erd.12.63 ↩