How to Choose a Whole Genome Sequencing Company

There is no single best whole genome sequencing company, but there is a best way to choose one. Rank vendors on five things. First, the sequencing platform and read length. Second, mean depth plus the fraction of the genome above a depth threshold. Third, which files you receive and own. Fourth, whether the lab is CLIA/CAP accredited. Fifth, the data deletion and re-sequencing policy. If a vendor will not tell you all five before you pay, that is your answer. Everything else in the marketing copy is downstream of these.

For a technically fluent buyer who wants to analyze their own data, the practical shortlist is short. Nebula Genomics and Sequencing.com sell 30x short-read WGS with raw file access at consumer prices. Dante Labs sells similar coverage with a longer history of turnaround complaints. Nucleus Genomics sells a consumer product with a polished interpretation layer. Full Genomes Corporation is the usual choice when you want long reads or deep Y-chromosome work. For anything clinical, order through a physician at a diagnostic lab rather than through a direct-to-consumer storefront. Variantyx, GeneDx, and Invitae are examples.

The five criteria that matter

Platform and read length determine what classes of variation you can see at all. Almost every consumer product is Illumina short-read sequencing, typically 2x150 base pair paired-end reads. Short reads call single-nucleotide variants and small insertions and deletions well. They handle structural variants, repeat expansions, and segmental duplications poorly. Those limits are inherent to the read length rather than to the vendor ¹.

Depth is the number most often misrepresented. “30x” usually means mean depth across the genome, which is a weak guarantee. Ask instead for the fraction of callable bases at 10x and 20x. A well-executed 30x library typically leaves more than 90% of the genome above 10x. A poor library with the same nominal mean can drop well below that because coverage is concentrated in GC-neutral regions. Population-scale sequencing programs settled on roughly 30x precisely because it is the point where single-sample genotype accuracy becomes adequate for most analyses ².

File access is the criterion that separates a data product from an app. You want the FASTQ files (raw reads), the aligned BAM or CRAM, and a VCF or gVCF of called variants. FASTQ lets you realign to a new reference build later. BAM or CRAM lets you inspect any locus yourself. A VCF alone is a dead end. You cannot re-call variants, check allele balance at a surprising call, or recover reads the vendor’s pipeline discarded.

Accreditation tells you what the result may be used for. CLIA certification and CAP accreditation cover the laboratory process for clinical testing in the United States. A research-use-only consumer genome can be technically excellent and still be unusable by a clinician, who will normally require a confirmatory test in an accredited lab before acting on any finding. Expect that, and budget for it.

Policy is the last criterion and the easiest to skip. Ask three questions in writing. Will you delete my sample and my data on request? Will you re-sequence free if coverage misses the advertised threshold? Who else receives my data, and under what consent? Several consumer genomics firms have changed hands or wound down, and the data-sharing terms you signed travel with the asset.

What you should expect to pay

Consumer 30x short-read WGS with raw data access sits roughly between $300 and $1,000, and promotional pricing dips lower. Long-read WGS on PacBio HiFi or Oxford Nanopore runs several thousand dollars for a comparable depth, because throughput per run is lower. Clinical WGS ordered through a physician is typically several thousand dollars billed to insurance, and includes a reviewed report and genetic counseling.

That price range is the end of a two-decade collapse. The first human genome was a multi-institution, multi-year effort with a budget in the hundreds of millions, and the commercial genomics sector of 2000 was built around selling access to sequence data rather than around sequencing individuals ³. Massively parallel sequencing broke the cost curve, and the per-genome price fell by several orders of magnitude within about a decade ⁴. National programs now sequence hundreds of thousands of participants as routine infrastructure ⁵.

Long reads versus short reads

We would buy short reads first and long reads only for a specific question. Short-read 30x is the cheapest path to a complete, reusable variant call set, and nearly all published interpretation tools and population frequency databases are built on short-read data. It is the correct default.

Long reads earn their premium when the question is structural. Reads of 10 to 100 kilobases span repeat regions, resolve large insertions and inversions, and phase variants across long haplotypes. A long-read genome typically reports several times more structural variants than a short-read genome of the same sample, most of them in repetitive regions that short reads cannot anchor into. Pay for long reads if you care about a known repeat expansion locus, a pharmacogene with a complex structure such as CYP2D6, or de novo assembly. Otherwise the money is better spent on a second timepoint of other data.

Verify what you received

Treat delivery as untrusted input. The first check is the reference build, because mixing GRCh37 and GRCh38 coordinates silently corrupts every downstream annotation.

samtools view -H sample.cram | grep -E '@SQ|@PG' | head -40

Look at the sequence names and lengths, and confirm whether alt contigs are present. Then measure depth directly rather than trusting the report.

mosdepth --fast-mode --by 1000 --threads 4 sample sample.cram
# then read sample.mosdepth.summary.txt and sample.mosdepth.global.dist.txt

The global distribution file gives you the fraction of bases at or above each depth. That is the number to compare against the vendor’s claim. Next check library quality and contamination.

samtools stats sample.cram > stats.txt   # duplicate rate, insert size, error rate
verifybamid2 --SVDPrefix resource/1000g.phase3.100k.b38.vcf.gz.dat \
             --Reference GRCh38.fa --BamFile sample.cram

Duplicate rates above roughly 15% indicate an over-amplified library and inflate apparent coverage. A contamination estimate above about 2% should prompt a conversation with the vendor. Finally sanity-check the variant call set.

bcftools stats sample.vcf.gz | head -40

A single human genome against GRCh38 typically yields on the order of four to five million variants, the large majority single-nucleotide. Counts far below that suggest hard filtering or a partial callset. Counts far above suggest contamination or a missing filter step. If you have prior array or exome data on yourself, use somalier to confirm the two files are the same person before you trust anything.

Reanalyze the data yourself

If you have the FASTQs, you are not bound to the vendor’s pipeline. Our default is bwa-mem2 for alignment and DeepVariant for small-variant calling, with GATK HaplotypeCaller as the alternative when you need a gVCF for joint calling later. Use manta or dysgu for structural variants from short reads, and accept that recall will be modest. Annotate with Ensembl VEP or with bcftools csq when you only need consequence terms.

Expect the compute bill. A 30x genome is roughly 100 GB of gzipped FASTQ, 15 to 20 GB as CRAM, and about 1 GB as a gVCF. Alignment and calling take a few hours on 32 cores, or under an hour on a DRAGEN or GPU-accelerated setup. Analysis at any scale beyond one genome becomes a data-engineering problem rather than a biology problem ².

One interpretive caution. Most of the variants in your file have no established clinical meaning, and consumer report generators vary widely in how conservatively they annotate the ones that do. Any finding that would change a medical decision needs confirmation in an accredited laboratory and discussion with a clinician or genetic counselor.

Questions people also ask

How much does it cost to get your entire genome sequenced? Consumer 30x short-read WGS with raw file access runs roughly $300 to $1,000. Long-read WGS costs several thousand dollars. Clinical WGS ordered by a physician is typically several thousand dollars and often billed to insurance.

Why is Illumina sequencing preferred over Sanger sequencing for whole genomes? Sanger sequencing reads one amplicon at a time at high accuracy, which makes whole-genome work impractically slow and expensive. Illumina sequencing reads billions of short fragments in parallel in a single run, which is what made per-genome costs fall by orders of magnitude ⁴.

What is the most reputable DNA testing company? Reputation depends on the use. For accredited clinical sequencing, established diagnostic laboratories ordered through a physician are the standard. For a raw data product you intend to analyze yourself, judge vendors on coverage evidence, file formats, and deletion policy rather than on brand.

What was unique about Ozzy Osbourne’s DNA? His genome was sequenced in 2010 by Knome as a publicity-oriented research project, and the reported findings included novel variants near genes involved in alcohol and drug metabolism plus some Neanderthal-derived ancestry segments. Those results were exploratory and were never a clinical interpretation.

Does whole genome sequencing replace ancestry testing? It supersedes it technically, since a 30x genome contains every marker an array measures. Ancestry inference depends on the reference panel and the vendor’s method, so a genome from a lab with a weak panel can produce a less informative ancestry breakdown than a dedicated consumer array.

Footnotes

1. Jianhua Zhao, Struan F.A. Grant. Advances in Whole Genome Sequencing Technology. Current Pharmaceutical Biotechnology, 2011. https://doi.org/10.2174/138920111794295729 ↩

2. Rachel L Goldfeder, Dennis P Wall, Muin J Khoury, et al. Human Genome Sequencing at the Population Scale: A Primer on High-Throughput DNA Sequencing and Analysis. American Journal of Epidemiology, 2017. https://doi.org/10.1093/aje/kww224 ↩ ↩²

3. Kathryn Brown. The Human Genome Business Today. Scientific American, 2000. https://doi.org/10.1038/scientificamerican0700-50 ↩

4. Sang Tae Park, Jayoung Kim. Trends in Next-Generation Sequencing and a New Era for Whole Genome Sequencing. International Neurourology Journal, 2016. https://doi.org/10.5213/inj.1632742.371 ↩ ↩²

5. Jan Smetana, Petr Brož. National Genome Initiatives in Europe and the United Kingdom in the Era of Whole-Genome Sequencing: A Comprehensive Review. Genes, 2022. https://doi.org/10.3390/genes13030556 ↩