Coverage depth, explained: why 30× isn't the ceiling — and when 1× is enough
"30× coverage" is the most-quoted number in consumer whole genome sequencing, and the most frequently misunderstood. It is an average, not a guarantee. It tells you how many times each base in your genome was read by the sequencer on the way to calling a variant — more reads, more confidence. For most of the genome, 30× is plenty. For some stretches, it isn't. And for specific use cases, 1× is genuinely enough.
The short version: at 30× average depth, roughly 96% of the genome lands in the "called confidently" bucket for a short-read Illumina instrument. Another 3% falls into low-confidence calls — positions where the caller saw enough reads to have an opinion but not enough to be sure. The remaining 1% is effectively uncovered: repetitive regions, low-complexity stretches, and segmental duplications where short reads can't tell one copy from another.
If you are buying WGS for carrier screening, ancestry, or pharmacogenomics, 30× is the right number. Those use cases depend on well-characterized regions where the reference genome is reliable and the variant-call question is binary. If you are looking for structural variants, de novo mutations, or anything in a repetitive region (some of which matter for neurological and developmental conditions), 30× short-read is not going to be enough on its own. That's where 100× — or long-read sequencing, which solves the problem differently — earns its keep.
The news in April 2026 is at the other end of the curve. MyHeritage's October 2025 move to low-pass WGS (~1× average depth) is the first time a mainstream consumer-DNA brand has shipped whole-genome sequencing at a genealogy-kit price. At 1×, you can't call rare or clinical variants confidently — the single-position confidence numbers are too low. But you absolutely can do what MyHeritage's product is built for: identify ancestry proportions across ~2,100 regions and do DNA-matching across millions of relatives. LP-WGS is not a cheap 30× WGS; it is a different product, optimized for statistical genealogy rather than clinical variant calling, and it is a genuine advance for readers whose question is family-history-shaped.
The widely-published confidence curve looks roughly like this: 1× gets you ancestry and nothing else (about 30% confident variant calls), 5× is trait-survey territory (~55%), 15× is "most SNPs confident" (~82%), 30× is the consumer standard (~96%), 60× is clinical-confirmation depth (~99%), and 100× is the practical floor for structural-variant discovery (~99.7%). Numbers vary by pipeline and by region, but the shape is robust.
If you take one thing away: 30× is not a ceiling. It's the right floor for clinical-style reporting, and the wrong product if all you want is family-history matching.