Sequencing depth and coverage

In one sentence

Sequencing depth counts usable observations at a position, while coverage breadth describes how much of a defined target meets an explicitly stated measurement criterion.

The intuition

Think of rain falling on a garden. Depth describes the water at one spot; breadth asks how much of the garden received enough water under a stated rule. A wet average can hide a dry corner. Sequencing differs because observations have qualities and may come from copied fragments, so counting them requires additional rules.

How it works

At a DNA position, depth commonly means the number of aligned reads covering that base after specified filters. A value of 40× means forty counted observations under those rules. Overlapping read pairs, duplicate fragments, low-quality bases and uncertain alignments may be handled differently by different pipelines. A depth without its counting rule is incomplete.

Mean depth averages positions. Coverage breadth reports a fraction of a named target, often the percentage of bases reaching a stated minimum usable depth. Neither tells you which important positions were missed. Ask for the per-position view or a list of low-coverage regions when the question depends on one location.

More sequencing can increase observation counts, but cannot recover a region absent from the library or create new original molecules from duplicates. Library complexity describes the diversity of represented fragments. Input amount, preservation, capture and alignment affect the usable evidence.

Why it matters in cancer

A variant present in only a small fraction of sampled DNA needs enough usable evidence to distinguish it from error. Variant allele fraction depends on tumor mixture, copy state and subpopulations, not just depth. Even deep sequencing can struggle with damage artifacts or an untested rearrangement class.

A negative report is therefore a statement about tested regions, detectable change classes and validated sensitivity. It cannot become “this gene has no relevant change” when a consequential location was poorly covered or never tested. The Association for Molecular Pathology (AMP) and College of American Pathologists (CAP) panel guideline connects these quantities to end-to-end validation rather than a single universal depth requirement.

Assay card

Field	What to retain
Measures and method	Observations at each defined position, then summaries across a specified target
Input and tissue cost	Existing alignment files need no new tissue; obtaining more diverse library molecules may require another DNA aliquot and consumes material
Output and units	Per-position depth in ×, mean or median depth, and breadth as a percentage of named bases meeting a specified rule
Thresholds	Report the chosen depth criterion separately from the assay's validated limit of detection for each relevant change class and specimen
Failure modes	Uneven capture, repeats, damaged input, duplicate inflation and misleading averages
Limits	Depth does not establish variant origin, protein function or clinical actionability
Validation context	Check low-fraction reference materials, reproducibility, quality filters and reportable regions for the intended assay

Common confusions

Coverage as a vague word: establish whether the author means regions targeted, depth, or breadth.
Deep versus broad: a narrow panel can be deep; a genome assay can be broader with less depth per position.
Depth versus sensitivity: the same depth can support different detection limits under different errors and variant classes.

Try it

A fictional report has mean depth 100×. Of 1,000 target bases, 900 reach its illustrative reporting criterion of 20×; the position you care about has four usable observations.

Answer: Breadth at that criterion is 90%. Mean depth does not establish adequate evidence at your position. The 20× rule here is a fictional summary criterion, not a recommended diagnostic cutoff. Ask whether the assay supports the proposed positive or negative claim there.

Sequencing reads: observations versus independent fragments.
Variant calling: evidence, error models and filters.

Sources

Source check: October 9, 2026. Expert and learner review remain pending. Examples are fictional.