# More sequencing reads do not answer every quality question

Draft • Research checked 4 October 2026 • AI-assisted DomDNA editorial content. No independent clinical review or publication is claimed. General education, not individual medical advice.

A sequencing report says “30 times coverage”. Another says “100 times”. The second number looks more reassuring, but you first need to know what it summarises. A genome-wide average and the number of useful reads at one difficult location are not the same description.

Read depth is an important technical measure. It is not a stand-alone percentage certainty that a variant is real, clinically important or absent.

## Count the evidence at the location

The [NHS Genomics Education glossary](https://www.genomicseducation.hee.nhs.uk/glossary/read-depth/) defines read depth as the number of times a base is represented in sequencing reads. Repeated observations can help distinguish a real signal from a sequencing error. That is the limited claim being used here, not a universal threshold for any test.

A hypothetical example makes the average problem visible. Imagine that a small target region has several easily read stretches with many observations and one stretch with very few. An average can look satisfactory while concealing the sparse section. That section matters if it contains the location of interest.

This does not mean that a higher-depth test is pointless. It means the headline statistic should lead to a site-specific question: what usable evidence was available where this particular result was called?

A consumer should not attempt to repair sparse data by manually declaring a genotype from a handful of displayed letters. A pile-up image is a technical view, not a substitute for the laboratory's validated calling and review process.

## The reads must be useful, not merely numerous

A read can contain a mistaken base or be difficult to place in the genome. [GATK's base-quality documentation](https://gatk.broadinstitute.org/hc/en-us/articles/360035890531-Base-Quality-Score-Recalibration-BQSR) describes base-quality scores as estimates related to the accuracy of individual base calls. That is a different question from how many observations exist.

The [NHS short-read sequencing explanation](https://www.genomicseducation.hee.nhs.uk/genotes/knowledge-hub/short-read-sequencing/) also notes that short fragments can be hard to map to the region they came from. More observations do not, by themselves, resolve every ambiguity in repetitive or similar-looking sequence.

Think of three headings in a hypothetical technical note: amount of evidence, accuracy of the observed letters, and confidence about their location. They belong together. None is the same as whether a detected change causes a condition.

That distinction helps avoid a common leap. Even a well-supported sequence difference can have uncertain biological significance. Technical confidence and clinical interpretation are separate stages.

## Ask a question the laboratory can answer

If a report concerns a particular location, ask whether the laboratory considered the result adequately supported under its validated method. If a test did not find an explanation, ask whether relevant difficult regions or variant types were outside its reliable detection.

These are method questions, not demands for a larger number. A different technology or confirmation process may sometimes answer a specific problem better than simply increasing the average depth of the same approach. The appropriate choice depends on the question and the service.

Keep any report wording about limitations intact when making a summary. Dropping the paragraph because it looks technical can make a result sound broader than it is. A short summary can still say that the method has limitations at some locations.

## A practical comparison note

Before comparing two offers or reports, record the stated depth measure and whether it is an average, minimum or location-specific figure. Record the technology, the question the test addresses and the laboratory's explanation of difficult regions. Do not create a ranking from depth alone.

If one provider supplies only a headline average, write that down as such. A useful request is “How do you assess whether the relevant sites are adequately read?” rather than “Is 100 better than 30?” This directs attention to test performance where it matters.

The aim is not to make every reader a sequencing analyst. It is to recognise when one impressive number has been asked to carry several different meanings.

A useful export therefore keeps the quality fields alongside the result, rather than saving only the letters. In a hypothetical comparison, two files may agree on the genotype while differing in read support or filtering. That difference is a reason to preserve the technical record, not to invent a personal confidence percentage.

For general lifestyle education, [DomDNA’s quiz](https://domdna.com/quiz) uses question answers; it does not inspect sequencing reads.