Every tumour sequencing project starts with the same budget question: sequence the whole genome, sequence just the coding regions, or use a targeted panel. The answer determines what you can conclude, and it cannot be revisited later without sequencing again.
This article sets out what each approach actually measures, what it costs in practice, and how to choose.
What each one covers
Whole-genome sequencing (WGS) covers essentially everything — roughly 3 billion bases, coding and non-coding. Tumour WGS is typically run at around 30–60× coverage, sometimes higher for low-purity samples.
Whole-exome sequencing (WES) captures the protein-coding regions, about 1–2% of the genome. Because the target is small, the same sequencing budget buys much deeper coverage — commonly 100–200× for tumours.
Targeted panels cover a chosen gene set, from a few dozen to a few hundred genes, at very high depth — often 500–1000× or more.
The pattern is a trade: breadth against depth. That trade is what makes the choice consequential, because in cancer, depth is not a luxury.
Why depth matters more in cancer than elsewhere
In germline sequencing, a variant is present in essentially 50% or 100% of reads. In a tumour it may be present in a small fraction, for two reasons that compound each other.
First, purity: a biopsy contains stroma, immune cells and normal tissue alongside tumour. A 40%-pure sample halves the observed allele fraction of a clonal mutation before anything else happens.
Second, subclonality: a mutation present in only a quarter of the tumour cells is quartered again.
Combine them and a subclonal mutation in a moderately pure sample can sit at a 5% variant allele fraction. At 30× coverage you expect roughly one or two supporting reads — indistinguishable from sequencing error. At 200× you expect ten, which is callable.
This is the single most important practical consideration, and it is why low-purity or heterogeneous tumours often argue for exome or panel sequencing over genome sequencing at the same cost.
What only WGS can give you
Structural variants and rearrangements
Breakpoints usually fall in intronic or intergenic sequence. Exome capture does not cover those regions, so WES detects structural variants poorly and unreliably. If gene fusions, chromothripsis or complex rearrangements matter to your question, WGS is the honest choice — or a fusion-specific assay such as RNA-Seq.
Mutational signatures
Signature decomposition needs a large, unbiased mutation catalogue. WGS typically yields thousands to tens of thousands of somatic mutations; a whole exome yields far fewer, and they are drawn from a functionally constrained subset of the genome.
Exome-based signature analysis is possible but noisier, and rarer signatures are frequently missed. Where homologous recombination deficiency status or a mismatch-repair signature is the point of the study, WGS is substantially more reliable.
Non-coding drivers and regulatory mutations
The TERT promoter mutations found in many cancers are the standard example: they are non-coding, recurrent and functionally important, and an exome cannot see them. Enhancer and promoter mutations more broadly are invisible to WES by construction.
Accurate copy number and ploidy
WGS gives even coverage across the genome, so copy-number segmentation is clean. Exome capture efficiency varies substantially between targets, which introduces waves and biases into copy-number estimates. Exome copy-number calling works, but is noisier and less reliable for subtle changes.
What WES and panels do better
Depth per pound. For detecting low-frequency coding variants in impure or heterogeneous samples, the depth advantage is decisive.
Analytical tractability. A WGS tumour-normal pair generates a very large amount of data and needs meaningful compute and storage. Exome analysis is comfortable on modest infrastructure, which matters for large cohorts.
Clinical maturity. Targeted panels dominate clinical practice for good reasons: they are validated, accredited, fast, cheaper, and report on a defined actionable gene set. Their limitation is equally clear — they only ever find what they were designed to look for.
FFPE tolerance. Archival formalin-fixed material is degraded and carries deamination artefacts. Higher depth helps distinguish real low-frequency variants from artefact, which favours exome and panel approaches for retrospective cohorts.
Tumour mutational burden: a caveat worth knowing
TMB is used as a biomarker for immunotherapy response, and it is measured differently by different assays. WGS measures mutations per megabase across the genome; exome measures it across coding regions; panels extrapolate from a much smaller territory.
These do not produce interchangeable numbers. Panel-based TMB has wider confidence intervals and can vary meaningfully with panel size and the bioinformatic filters applied. A TMB threshold validated on one assay should not be applied to another without care — this is a common and consequential error.
The design decision that matters more than any of this
Whichever you choose, sequence a matched normal if you possibly can.
Every genome carries millions of inherited variants. Without a matched normal, separating somatic mutations from rare germline ones depends on population databases — and rare germline variants, especially in under-represented ancestries, are systematically misclassified as somatic. That inflates TMB, contaminates driver lists and distorts signature analysis.
A tumour-only exome with a matched normal is more informative than a tumour-only genome without one. If budget forces a choice between breadth and pairing, pair.
Choosing, in short
- WGS when structural variants, mutational signatures, non-coding drivers or accurate copy number are central; when samples are reasonably pure; and when compute and budget allow.
- WES when coding drivers are the question, when samples are impure or heterogeneous, when the cohort is large, or when working with FFPE material.
- Targeted panel when the question is clinical actionability against a known gene set, when turnaround and validation matter more than discovery, or when input material is very limited.
A common and sensible design combines them: panel or exome across a full cohort to establish the mutational landscape, WGS on a selected subset where structural and signature questions need answering. That usually delivers more per pound than committing everything to one approach.
