DNA Complete vs Sequencing: What Differs Between Whole-Genome Products
If you are comparing DNA Complete (Nebula Genomics) against Sequencing.com, the underlying laboratory product is close to identical: paired-end short-read whole-genome sequencing on an Illumina instrument at roughly 30x mean coverage. The chemistry is a commodity. What separates the two offerings is the deliverable and the access model. That covers which files you can download and which reference build they were aligned to. It also covers whether a variant call file comes with quality metadata and how much of the interpretation sits behind a recurring subscription. If you plan to do your own analysis, pick on file access and reference build. Do not pick on the marketing claim about reading 100% of your DNA.
What each name refers to
“Sequencing” is the generic laboratory process. “DNA Complete” is a brand name for a consumer whole-genome sequencing package. Comparing them as if they were two different technologies is a category error that the marketing invites. Both DNA Complete and Sequencing.com resell short-read whole-genome sequencing generated in a contract lab, then wrap it in a web app that runs reports over your variants.
Both are distinct from the genotyping arrays sold by ancestry companies. An array measures a fixed panel of roughly 600,000 to 900,000 predetermined positions and infers nothing about the rest. Whole-genome sequencing generates reads across the genome and lets you call variants at positions nobody chose in advance. That difference is real and it is the reason to pay for sequencing over an array.
The specification questions that matter
Ask these before you buy. First, do you get raw FASTQ, and is it the full unfiltered set of reads? FASTQ is the text format holding each read sequence and its per-base quality scores. If you have FASTQ you can realign to any reference, forever, with any aligner. Everything downstream is reproducible from it.
Second, do you get an aligned BAM or CRAM, and against which reference build? GRCh38 is the sane default for anything current. Many consumer pipelines still ship hg19/GRCh37 because older annotation databases assume it. Coordinates are not interchangeable between builds, and lifting over is lossy near indels and segmental duplications.
Third, do you get a VCF with full quality annotations, or a stripped list of genotypes? A usable VCF carries per-sample DP for read depth and GQ for genotype quality. It also carries AD for allelic depth and the filter status from the caller. A two-column table of rsIDs and genotypes is not enough to judge whether a call is trustworthy.
Fourth, what happens after the promotional period ends? Some products gate reports and even file download behind an annual membership. Anything you cannot export is a rental.
What “100% of your DNA” leaves out
The claim is not literally true for any short-read product, and the reasons are structural rather than a matter of vendor quality. Reads of 150 base pairs cannot be placed uniquely inside long repeats. Centromeres, most of the acrocentric short arms, and large segmental duplications collapse into ambiguous alignments with mapping quality near zero. Callers then filter those regions out.
The reference itself is a construct with a history. The physical maps that made assembly possible were built on a shared coordinate vocabulary agreed by the community in the late 1980s1. The public project’s timeline was pulled forward by a shift in strategy in the mid-1990s2, and the genome was declared finished in 2003 with known euchromatic gaps remaining3. Those gaps were only closed much later by long-read assemblies. So when a vendor says 100%, read it as “unbiased shotgun coverage of the mappable genome,” which is a genuinely useful thing and still not everything.
Short reads also under-call structural variation. Large deletions, inversions, and mobile element insertions are inferred from read-pair and split-read signals rather than observed directly. Tandem repeat expansions of clinical interest are usually beyond reach at 150 bp. If those are your target, you want long reads (PacBio HiFi or Oxford Nanopore), which neither product sells.
Verify what you received
Do not trust the coverage number on the dashboard. Compute it. Start with the raw reads.
seqkit stats -a sample_R1.fastq.gz sample_R2.fastq.gz
Multiply total bases across both files by the expected alignment rate and divide by 3.1e9. For 30x you want roughly 93 to 100 gigabases of sequence, which is about 310 to 330 million read pairs at 2x150.
Then measure depth on the alignment. mosdepth is fast and gives you the distribution, which a mean alone does not show:
mosdepth --by 1000 --fast-mode --threads 4 sample sample.cram
Look at sample.mosdepth.global.dist.txt. The number to care about is the fraction of the genome at 10x or more. A healthy 30x library puts well above 90% of the genome past that threshold. A mean of 30x with a fat low tail means uneven coverage from library bias or duplication.
Check duplicates and contamination too:
samtools flagstat sample.cram
verifyBamID2 --SVDPrefix resource/1000g.phase3.100k.b38.vcf.gz.dat --BamFile sample.cram
Optical and PCR duplicate rates above roughly 15% waste sequencing you paid for. A contamination estimate (FREEMIX) above about 0.02 should make you distrust heterozygous calls.
Finally, sanity-check the variant file:
bcftools stats sample.vcf.gz | grep -E "^SN|^TSTV"
A single human genome against GRCh38 yields on the order of 4 to 5 million SNVs plus several hundred thousand small indels. The transition/transversion ratio should land near 2.0 to 2.1 genome-wide. Materially fewer variants usually means the file was filtered to a reportable subset. A Ti/Tv well below 2.0 suggests a permissive filter admitting noise.
What we would do with the files
If the vendor supplies FASTQ, realign it yourself. Our default is bwa-mem2 to GRCh38 with alt contigs handled properly, then DeepVariant for small variants, then Ensembl VEP for annotation. GATK HaplotypeCaller is the alternative and is better documented for joint calling across a family. Producing a gVCF rather than a VCF is worth the extra disk, because a gVCF records confidence at non-variant positions. That lets you distinguish “reference at this site” from “we had no reads here,” a distinction that matters constantly and that plain VCFs destroy.
Budget for storage. FASTQ for a 30x genome is roughly 100 GB compressed. A BAM lands near 60 to 80 GB, while CRAM with reference-based compression comes in around 15 to 20 GB. We keep CRAM plus the gVCF and regenerate everything else on demand.
One limit is worth stating plainly. Consumer whole-genome sequencing is generally not returned as a clinical-grade result. If you find a variant that appears medically meaningful, the next step is confirmation in an accredited laboratory and discussion with a clinical geneticist or genetic counselor. Do not act on a research-grade call.
Questions people also ask
Is DNA Complete different technology from regular sequencing? No. It is a consumer package built on standard Illumina short-read whole-genome sequencing at about 30x. The differentiation is in the app and the subscription terms and the price. The instrument is not what differs.
Which is cheaper? Prices move often enough that any figure here would be stale. Compare total cost over three years including membership renewals, and check whether file download survives cancellation.
Do I get my raw data with both? Both advertise raw data access. Confirm specifically that FASTQ is included rather than only a VCF, and confirm there is no extra fee to retrieve it.
Is 30x enough? For germline small variant calling, yes. It is the standard for a reason, and it gives enough depth to call heterozygous sites reliably across the mappable genome. It is not enough for somatic work or for confident structural variant detection.
What does 30x mean exactly? It is the mean number of reads covering each base. The mean hides variance, which is why the fraction of the genome above 10x is the more informative statistic.
Footnotes
-
Maynard Olson, Leroy Hood, Charles Cantor, et al. A Common Language for Physical Mapping of the Human Genome. Science, 1989. https://doi.org/10.1126/science.2781285 ↩
-
Eliot Marshall. A Strategy for Sequencing the Genome 5 Years Early. Science, 1995. https://doi.org/10.1126/science.7846520 ↩
-
Tabitha M Powledge. Human genome project completed. Genome Biology, 2003. https://doi.org/10.1186/gb-spotlight-20030415-01 ↩
Pothos builds longitudinal molecular profiles of individuals: whole-genome sequencing, single-cell RNA sequencing, and plasma proteomics, delivered as a detailed report, every data file, and stored cells. Build your profile.