Sequencing Coverage Explained: Depth, Uniformity, and Cost

Sequencing coverage is the average number of times each base in a target region is read during a sequencing run, expressed as a multiple such as 30X, where 30X means each base was read about thirty times on average. Coverage is the run's statistical strength: higher coverage separates true variants from sequencing errors with more confidence, at the price of more sequencing and more cost.

For research teams planning a sequencing project, coverage is the first design decision after choosing the technology, because it drives both data quality and budget. This guide explains what coverage measures, how it is calculated, what levels different experiment types typically need, and why the average alone is never the whole story.

Coverage, Depth, and Breadth: Three Terms That Are Not Synonyms

Coverage and depth are often used interchangeably for the average read multiplicity across a region, but they become distinct when a project spans a large genome: coverage can describe the breadth of a genome that has any reads at all, while depth describes how many times a given base was read. A targeted panel can reach very high depth across a tiny breadth of genome, which is why panel and exome numbers look large compared with whole-genome numbers.

Breadth is the fraction of the target that is actually represented. A run that averages 30X but covers only 80% of the genome has gaps where variants are invisible, and the missing fifth matters more than the average suggests. Asking for all three numbers, breadth, average depth, and uniformity, gives a complete picture of what the data can support.

How Coverage Is Calculated

Average coverage is calculated from the run's total output: the number of reads multiplied by the read length, divided by the size of the target region. A run producing 600 million bases of usable sequence against a 20-megabase target gives an average depth of 30X. The calculation uses usable, mapped reads after quality filtering, so the raw output number overstates what the analysis actually receives.

The calculation also explains the two levers that control coverage: sequencing more reads or using a smaller target. A lab that needs deeper coverage can either buy more sequencing or focus the same output on a smaller region, which is exactly the trade-off between whole-genome, exome, and targeted approaches.

How Much Coverage an Experiment Needs

These levels are planning references, not rules: the required coverage depends on the variant frequency to be detected, the sample quality, and the confidence the project must reach. The correct process is to define the question first, the lowest detectable variant frequency it requires, then the coverage that supports it, and to record that reasoning with the run.

Uniformity: Why the Average Hides the Gaps

An average of 30X can be produced by uniform coverage, where every base sits near 30 reads, or by a mix of deep peaks and near-empty valleys. The valleys are where variants are missed or called with low confidence, so uniformity is what determines how much of the target is genuinely interrogated at the promised depth. GC-rich regions, repetitive sequence, and library preparation bias all produce systematic valleys.

Practical depth thresholds should therefore be stated as minimums rather than averages: a project requiring 20X minimum depth will reject regions below that level and report them as gaps. Reviewing coverage plots across the target, and documenting which regions fell short and why, turns the run's blind spots from silent omissions into known limitations.

The Cost Trade-Off and Where to Spend It

Coverage and cost scale together through data volume, so the cheapest improvement is usually to shrink the target rather than buy more sequencing. A research question that only concerns a few hundred genes is answered far more cheaply at high depth with a panel than at whole-genome scale, which is why matching the target to the question is the first cost control.

Spending on coverage instead of replicates is another common misallocation: deeper coverage of one sample does not substitute for biological replicates when the question is about variability across samples. Coverage buys confidence in reads, not confidence in biology, and the budget should be split with that distinction in mind. Recording the target, thresholds, and achieved coverage in the project's records keeps that design decision visible, and the Zettalab workspace links structured records with team file storage so coverage reports stay attached to the samples they describe.

Documenting Coverage Decisions With the Run

The coverage plan belongs with the sequencing record: target region, expected and achieved average depth, minimum thresholds, and the regions that failed them. When a variant call is questioned later, the record shows what depth supported it and where the data could not speak. For teams that want sequencing context and analysis documentation connected, the Zettalab workspace links structured experiment records with team file storage, so the run's coverage report stays attached to the samples and analyses that depend on it.

FAQ

What does 30X coverage actually mean?

It means that on average, each base in the target region was sequenced about thirty times. The thirty independent reads vote on the true base at each position, which is what separates a genuine variant from a single read's sequencing error. The number is an average, so individual positions can sit above or below it, and the regions below a minimum threshold are where calling confidence drops.

What is the difference between sequencing coverage and sequencing depth?

In practice the terms are used interchangeably for the average read multiplicity across a region. Where they separate, coverage refers to how much of the target has reads at all, while depth refers to how many reads sit on a given base. A targeted panel illustrates the distinction: enormous depth across a small covered fraction of the genome.

How do I calculate the coverage I will get from a run?

Divide the run's usable output by the target size: multiply the number of mapped, quality-filtered reads by the read length, then divide by the number of bases in the target region. For example, 600 million usable bases against a 20-megabase target gives about 30X average depth. The calculation should use mapped reads, because unmapped or failed reads contribute nothing to coverage.

Why is my sequencing coverage uneven across the target?

GC-rich regions amplify poorly in library preparation, repetitive regions map ambiguously, and some sequence contexts are systematically under- or over-represented. These biases create valleys and peaks that the average hides. Reviewing the coverage plot identifies the affected regions, and those below the project's minimum threshold should be reported as gaps rather than silently included.

Does higher coverage fix sequencing errors?

Higher coverage does not remove errors, it outvotes them: with more independent reads per base, a sporadic error in one read is overwhelmed by the agreement of the others. Coverage therefore improves variant calling confidence up to a point, beyond which additional depth adds cost faster than it adds certainty. It cannot fix systematic problems such as library bias or contaminated samples.

Should I increase coverage or add biological replicates?

The two buy different things. Coverage buys confidence in the reads of one sample; replicates buy confidence that a finding generalizes across samples. When the question is about biological variability, replicates answer it and extra coverage does not. Budgets should follow the question, and in most expression and comparative studies, replicates are the more informative purchase.

Conclusion

Sequencing coverage is the project's statistical budget: the average depth sets how confidently variants are called, uniformity determines how much of the target is truly covered, and both are bought with sequencing volume. Choosing coverage starts from the question, is checked against the coverage plot rather than the average, and is recorded so the data's limits stay visible. To keep sequencing context connected to the lab's records and analyses, explore Zettalab's cloud-based R&D lab platform.