What 30x Whole Genome Sequencing Gives You

30x whole genome sequencing means the sequencer produced enough reads that, averaged across the genome, each base is covered by about 30 independent reads. That depth is the standard for germline variant calling because it makes heterozygous sites easy to distinguish from sequencing error. At 30x you should expect roughly 4 to 5 million variants relative to the reference, of which around 3.5 million are single-nucleotide variants and several hundred thousand are short insertions and deletions. What you do not get is complete coverage: a meaningful slice of the genome stays hard to call at any short-read depth, and structural variants remain only partially visible.

What the number 30 refers to

Depth is a mean. It is not a floor. If you sequence 150 base-pair paired-end reads to 30x, coverage at any given position is roughly Poisson-distributed around 30. Real data shows more spread than Poisson because of GC content, mappability and library biases. The practical consequence is that a genome with 30x mean depth still contains millions of bases at 10x or below, and a smaller set at zero.

That matters because calling a heterozygous site requires seeing both alleles. At 10x depth, the chance of drawing zero copies of one allele is about one in a thousand per site under binomial sampling. Multiply that across three billion bases and the low-depth tail becomes the main source of missed calls. Depth and breadth of coverage are separate quantities, and the relationship between them depends on read length and library complexity as much as on total sequencing output.¹

Going deeper helps, with diminishing returns. Work on comprehensive personal genome sequencing found that sensitivity for both SNVs and indels keeps improving past the conventional 30x mark before flattening, with the largest gains in repetitive and low-complexity regions.² Above roughly 40x the additional variants are mostly in regions where short reads are a poor tool anyway. We think 30x is the right operating point for a personal baseline, and that money beyond it is better spent on other data types than on more of the same reads.

Where 30x short reads fail

The failures are systematic rather than random, which is good news because you can predict and audit them. Coverage drops in GC-rich and GC-poor sequence, in segmental duplications and in homopolymer runs. A comparison of four sequencing platforms showed platform-specific coverage bias that translates directly into differences in variant calling sensitivity. So the vendor and chemistry matter as well as the depth number. Depth is not the only thing to compare.³ Analysis of coding-region coverage found that GC content and mappability dominate the variation in per-exon depth. That leaves some clinically interesting genes persistently under-covered even in whole-genome data.⁴

Practical list of things a 30x short-read genome handles poorly. Repeat expansions such as those in HTT or FMR1 need specialized callers and are unreliable at best. Large structural variants and copy-number changes are detectable in aggregate but with poor breakpoint resolution. Paralogous gene families confound alignment, including PMS2, SMN1 and the HLA region. Mitochondrial heteroplasmy at low fraction needs far more depth than 30x on the mitochondrial contig, though mtDNA is usually oversampled and often lands above 1000x anyway.

Contamination is a failure mode people underestimate. A study of whole genome sequences from 1000 families catalogued bacterial, viral and computational contamination present in the raw reads. Some of it was introduced by the laboratory workflow rather than the sample.⁵ If you plan to use unmapped reads for anything, treat them with suspicion.

The files to insist on

Ask for raw and processed data, and check what you received before deleting anything. The minimum useful set looks like this.

• FASTQ, gzipped, paired-end. Roughly 40 to 60 GB per read file at 30x. This is the only artifact that lets you realign to a future reference.
• CRAM aligned to GRCh38 with alternate contigs handled explicitly, plus the .crai index. A 30x CRAM runs about 15 to 25 GB, versus 80 to 120 GB for the equivalent BAM. CRAM is reference-dependent, so record the exact reference FASTA and its MD5.
• gVCF as well as the filtered VCF. The gVCF records reference confidence at non-variant positions, which is how you distinguish “reference” from “no data” at any locus you later care about. The filtered VCF does not record it.
• A per-base or per-window coverage file, and the caller’s own QC metrics.

If a provider will only hand you a curated report or a proprietary viewer, the data is not yours in any useful sense. A reference point for what a well-processed 30x cohort looks like is the expanded 1000 Genomes Project resequencing, which produced 30x data on 3202 samples with published per-sample variant counts and quality metrics you can compare your own numbers against.⁶

The QC pass we would run first

Before interpreting anything, spend an hour establishing that the data describes one human at the depth you paid for, and that it is the right human.

Compute coverage with mosdepth, which is fast and gives you both distribution and per-region output:

mosdepth --by 1000 --fast-mode --no-per-base sample data/sample.cram

Read the sample.mosdepth.global.dist.txt file and check the fraction of the genome at 10x and above. On good 30x data we expect around 90% of the primary assembly at 10x or better and a mean near 30. If the fraction at 10x is in the 70s, the mean depth is hiding a wide coverage distribution.

Check alignment and duplication with samtools stats and look at the duplicate rate. Above roughly 15% suggests a low-complexity library, meaning you paid for reads that carry no new information. Check contamination and sample identity with verifyBamID2 or somalier, and confirm reported sex against X and Y coverage ratios. Then compare your variant counts against expectation: a European-ancestry sample should land near 4 to 5 million variants, and an African-ancestry sample noticeably higher. A count far outside that range points at a pipeline problem rather than an unusual genome.

For variant calling we would use DeepVariant on the aligned CRAM for SNVs and small indels, because it is straightforward to run and performs well without hand-tuned filters. GATK HaplotypeCaller in gVCF mode remains the right choice if you need a gVCF that joins cleanly with public cohorts. Evaluate your own pipeline against a Genome in a Bottle reference sample using hap.py with the GIAB high-confidence BED. That gives you a real precision and recall number for your setup instead of a vendor claim.

When 30x is the wrong purchase

Two alternatives are worth knowing about, because for specific goals they are better value than a 30x genome.

If your interest is polygenic scores and common-variant ancestry, low-depth whole genome sequencing with imputation gets most of the way there for a fraction of the cost. Combining low-depth genome data with high-depth exome sequencing has been shown to be a cost-effective design for genetic studies. It captures both common-variant imputation and rare coding variation.⁷ A related approach, blended genome-exome sequencing, was developed explicitly as a cheaper substitute for deep genomes or arrays in large cohorts.⁸ Low-depth genomes also add value as a complement to targeted panels. That includes structural and copy-number variants missed by panel sequencing in hearing loss.⁹

If your interest is a specific diagnostic question, a clinician-ordered test with a validated pipeline and an interpreted report is the correct route. Research-grade 30x data is not a clinical diagnosis, and any variant you find that appears medically relevant should be confirmed in a clinical laboratory and discussed with a genetic counselor or physician before you act on it.

Questions people also ask

What does 30x whole genome sequencing coverage mean? It means the total number of sequenced bases equals about 30 times the genome length, so each position is covered by roughly 30 reads on average. Coverage varies widely around that mean, so a 30x genome typically has 90% or so of the primary assembly at 10x or deeper and some fraction at zero.

What was unique about Ozzy Osbourne’s DNA? His genome was sequenced in 2010 by Knome and discussed publicly as an early celebrity whole-genome project. The reported findings included variants near alcohol-metabolism genes and segments of archaic ancestry. The scientific value was limited, and it mostly demonstrated how hard single-genome interpretation is without a comparison cohort.

Is full genome sequencing worth it? It is worth it if you intend to work with the data: run your own callers, track variants against new literature, and combine the genome with other molecular measurements over time. If you want a fixed set of consumer reports, a genotyping array costs far less and answers most of the same questions.

How much will whole genome sequencing cost in 2026? Sequencing reagents and instrument time for a 30x human genome now sit in the low hundreds of dollars at scale. Consumer and research offerings that include library preparation, alignment and variant calling typically price between a few hundred and roughly a thousand dollars. The variable cost is analysis and data delivery. The reads are not the variable part.

Footnotes

1. David Sims, Ian Sudbery, Nicholas E. Ilott, et al. Sequencing depth and coverage: key considerations in genomic analyses. Nature Reviews Genetics, 2014. https://doi.org/10.1038/nrg3642 ↩

2. Subramanian S. Ajay, Stephen C.J. Parker, Hatice Ozel Abaan, et al. Accurate and comprehensive sequencing of personal genomes. Genome Research, 2011. https://doi.org/10.1101/gr.123638.111 ↩

3. Nora Rieber, Marc Zapatka, Bärbel Lasitschka, et al. Coverage Bias and Sensitivity of Variant Calling for Four Whole-genome Sequencing Technologies. PLoS ONE, 2013. https://doi.org/10.1371/journal.pone.0066621 ↩

4. Yury A. Barbitoff, Dmitrii E. Polev, Andrey S. Glotov, et al. Systematic dissection of biases in whole-exome and whole-genome sequencing reveals major determinants of coding sequence coverage. Scientific Reports, 2020. https://doi.org/10.1038/s41598-020-59026-y ↩

5. Brianna Chrisman, Chloe He, Jae-Yoon Jung, et al. The human “contaminome”: bacterial, viral, and computational contamination in whole genome sequences from 1000 families. Scientific Reports, 2022. https://doi.org/10.1038/s41598-022-13269-z ↩

6. Marta Byrska-Bishop, Uday S. Evani, Xuefang Zhao, et al. High-coverage whole-genome sequencing of the expanded 1000 Genomes Project cohort including 602 trios. Cell, 2022. https://doi.org/10.1016/j.cell.2022.08.004 ↩

7. Claude Bhérer, Robert Eveleigh, Katerina Trajanoska, et al. A cost-effective sequencing method for genetic studies combining high-depth whole exome and low-depth whole genome. npj Genomic Medicine, 2024. https://doi.org/10.1038/s41525-024-00390-3 ↩

8. Matthew DeFelice, Jonna L. Grimsby, Daniel Howrigan, et al. Blended Genome Exome (BGE) as a Cost Efficient Alternative to Deep Whole Genomes or Arrays. 2024. https://doi.org/10.1101/2024.04.03.587209 ↩

9. Sun Young Joo, Iyosaphit Endrias Essay, Seung Hyun Jang, et al. Low-depth whole genome sequencing as a complementary approach for detecting clinically relevant variants in hearing loss. Scientific Reports, 2026. https://doi.org/10.1038/s41598-026-67227-0 ↩