Exome sequencing reads only the protein-coding regions of DNA, roughly 1 to 2 percent of the genome, while genome sequencing reads nearly all of it, including the 98 percent that does not code for proteins. The exome is cheaper and faster to interpret; the genome captures deep intronic variants, structural rearrangements, and copy-number changes the exome cannot see. Choosing between them depends on what question you are asking and what the test can actually detect.

Two capture strategies, one shared chemistry

Both tests begin the same way. DNA is fragmented, and the resulting library is sequenced on a short-read platform, typically producing reads of 100 to 150 base pairs. The difference is what gets sequenced, and that difference is created before the instrument ever runs.

Exome sequencing uses a capture step. A set of biotinylated oligonucleotide probes, designed against known exons, hybridizes to complementary fragments in the library. Streptavidin beads pull the hybridized fragments out of solution, and everything else is washed away. The captured pool is then amplified and sequenced. Because roughly 180,000 exons across about 20,000 protein-coding genes occupy only 1 to 2 percent of the 3.1 billion base pair genome, the sequencer spends its capacity on a small, dense target. That concentration is why a 100x average depth is achievable at a fraction of genome-scale cost.

Genome sequencing skips capture entirely. Libraries are sequenced directly, and coverage arrives in two flavors. Paired-end reads give roughly 30x coverage across the unique portion of the genome. Long-insert or large-insert libraries add a second layer, with inserts of several kilobases producing what is called long-range coverage, which is estimated rather than counted directly. That second layer is what makes structural variant detection possible at genome scale, because a 5 kilobase deletion is invisible to a 150 base pair read but obvious when a long insert fails to align.

The mechanism matters because it explains the failure modes. Probes capture what they were designed to capture. Exons with extreme GC content, first exons, and repetitive paralogous regions often capture poorly, leaving gaps that look like negative results but are really coverage failures. The plain-language version: the exome test is a fishing net with a specific mesh size, and some fish swim through.

What each test actually misses

The numbers here are the whole argument. Exome sequencing covers roughly 1.5 to 2 percent of the genome at diagnostic depth. Genome sequencing covers 90 percent or more at 30x, though the reliably callable fraction is smaller because of repeats, centromeres, and segmental duplications.

The diagnostic yield gap is real but narrower than the coverage gap suggests. For rare Mendelian disease, exome sequencing has historically produced a diagnosis in roughly 25 to 35 percent of cases in specialty genetics cohorts. Genome sequencing adds roughly 5 to 10 percentage points on top, with the gain concentrated in specific categories: deep intronic splice variants, balanced structural rearrangements, small copy-number changes below exon resolution, and mitochondrial genome variants. A meta-analysis of exome and genome sequencing across genetic disorders found an overall added diagnostic yield of roughly 8 percent for genome sequencing over exome, with the largest gains in conditions with known noncoding mechanisms.

That asymmetry is the counterintuitive part. Genome sequencing covers 60 times more DNA, yet the incremental diagnostic yield is measured in single digits. Most disease-causing variation in known Mendelian genes sits in exons and splice sites, which the exome already sees. The genome's advantage is not breadth for its own sake; it is access to the specific noncoding and structural classes of variation that exome capture structurally cannot reach.

Cost tracks the same asymmetry. At list prices, clinical exome sequencing runs roughly USD 1,000 to USD 3,000, while clinical genome sequencing runs roughly USD 2,500 to USD 6,000, though negotiated and research pricing sits well below these figures. Sequencing cost per base has fallen far faster than interpretation cost, which is why the price gap between the tests is narrower than the coverage gap would predict. Interpretation, not chemistry, dominates the bill.

Feature Exome sequencing Genome sequencing
Fraction of genome covered 1 to 2 percent 90 percent or more
Typical depth 80x to 100x on exons 30x plus long-range coverage
Deep intronic variants Not reliably detected Detectable
Structural variants Limited, breakpoint dependent Core strength
Copy-number changes Exon-level resolution Near base-pair resolution
Mitochondrial genome Captured with targeted probes Covered
List price, clinical USD 1,000 to USD 3,000 USD 2,500 to USD 6,000
Interpretation burden Lower, coding-focused Higher, noncoding uncertainty

Where exome wins outright

Depth is the underappreciated argument. Because the sequencer concentrates reads on 1 to 2 percent of the genome, exome sequencing routinely achieves 80x to 100x mean coverage on captured exons, and often more. That depth buys sensitivity for low-level somatic variants in tumor samples and for mosaicism, where a variant might be present in 5 to 10 percent of cells. A 30x genome cannot match that depth on the same budget. For mosaic disorders and cancer panels, exome is often the more sensitive instrument.

Variant interpretation is the second argument. A genome returns roughly 4 to 5 million variants per person. An exome returns roughly 20,000 to 50,000 coding variants. The coding set can be filtered against population frequency, predicted consequence, and clinical databases with relative speed. The noncoding set cannot, because the effect of a variant in an intron or an intergenic region is usually unknown. More data is not automatically more information, and the variant of uncertain significance rate climbs accordingly.

The noncoding genome is not junk, and that cuts both ways

The historical framing of the 98 percent noncoding fraction as junk was wrong, but the correction has been oversold. Roughly 10 to 20 percent of the genome shows evidence of biochemical activity in ENCODE-style assays, and a meaningful fraction of that is regulatory. Known noncoding pathogenic mechanisms include deep intronic splice activation, where a variant creates a cryptic splice site far from the exon, and enhancer disruption in genes with well-characterized long-range regulation.

The professional caveat is that most noncoding variants are still uninterpretable. Finding a rare variant in a conserved intronic region is not a diagnosis; it is a hypothesis. Genome sequencing shifts the bottleneck from detection to interpretation, and that shift is why clinical genome reports carry larger uncertain-variant sections and why genetic counseling before testing matters as much as the test itself.

Red flags when reading a report or a marketing claim:

A negative exome is not a negative genome; coverage gaps and undetected variant classes remain. A "whole genome" marketed at exome prices is usually low-pass or capture-based, not 30x long-range. Variant of uncertain significance counts rise with genome sequencing; plan for follow-up. Mitochondrial and repeat-expansion disorders may require dedicated assays regardless of test choice. Noncoding findings without functional evidence should not be treated as diagnostic. Coverage statistics should be reported per gene, not as a single genome-wide mean.

How the workflow diverges in time and follow-up

The two tests share front-end steps and diverge at capture and analysis, which is where turnaround differences accumulate.

Stage Exome sequencing Genome sequencing
Sample and library prep 1 to 3 days 1 to 3 days
Capture or no capture 1 to 2 days None
Sequencing run 1 to 2 days 2 to 4 days
Alignment and variant calling 1 to 2 days 2 to 5 days
Interpretation and reporting 3 to 10 days 7 to 21 days
Typical total turnaround 1 to 3 weeks 2 to 5 weeks

The interpretation row is the honest one. Genome analysis requires noncoding annotation, structural variant review, and often additional family studies to phase variants, all of which add calendar time. A faster chemistry does not produce a faster answer if the analysis queue is longer.

Choosing between them

The practical algorithm is question-first, not price-first.

Step 1: Define the phenotype. For a suspected Mendelian disorder with a known coding mechanism, exome sequencing is a reasonable first tier and is supported by major genetics society guidance as a first-line test for rare disease.

Step 2: Check prior testing. If an exome was negative and the phenotype fits a noncoding or structural mechanism, genome sequencing is the logical next step rather than a repeat exome.

Step 3: Consider sample type. For mosaic or somatic questions, prioritize depth over breadth and favor exome or a targeted high-depth panel.

Step 4: Budget for interpretation. Add the cost of genetic counseling and follow-up functional studies to any price comparison; the sticker price understates total cost.

Step 5: Confirm coverage metrics. Ask for per-gene coverage and the laboratory's reported limit of detection for structural variants before ordering.

Step 6: Reconcile thresholds explicitly. A 30x genome is the standard for germline single-nucleotide variant calling, while 80x to 100x exome depth is the standard for mosaic detection. These are different thresholds for different questions, and neither is universally correct. Match the depth to the variant class you need to find.

Limits of what the comparison can settle

Neither test sequences everything. Both miss large repeat expansions, which require dedicated assays such as repeat-primed PCR. Both struggle with highly homologous regions where reads map ambiguously. Both leave pharmacogenomic and polygenic risk questions largely outside the diagnostic frame. Variant interpretation standards evolve, so a variant classified as uncertain today may be reclassified later, which is why reanalysis policies matter more than the original test choice.

The deepest limit is epistemic. A genome sequence contains far more data than current knowledge can interpret, and an exome sequence contains less data but a higher fraction that current knowledge can act on. The right question is not which test sees more, but which test produces more interpretable answers for this patient, this phenotype, and this point in time. For most rare disease evaluations, exome first remains defensible; for negative exomes with a persistent noncoding or structural hypothesis, genome sequencing is the appropriate escalation, and the roughly 8 percent incremental yield is worth the added cost and uncertainty when the question fits.