What a Genome Tester Gives You, and How to Judge One

If you want your own genome and you intend to work with the data, the answer is a 30x whole-genome sequencing service that returns three things. It should return FASTQ or unaligned BAM, a CRAM aligned to GRCh38 with an explicit reference build and decoy set, and a gVCF alongside the filtered VCF. Anything that returns only a curated PDF of “wellness insights” is a report subscription with a saliva tube attached. The distinction that matters is not the marketing tier but the deliverable: raw reads plus variant calls you can re-call, re-annotate, and re-interpret in five years when the annotation databases have improved and the sequencer has not changed a thing about your DNA.

That framing also explains why “genome tester” returns two unrelated categories of result. One is GenomeTester, a small bioinformatics tool from the University of Tartu for counting PCR primer binding sites in a template sequence. The other is consumer and clinical genome sequencing. This page is about the second, and specifically about how to evaluate a provider like an engineer evaluating a data vendor.

Genotyping arrays versus whole-genome sequencing

Most inexpensive DNA tests are genotyping arrays. They do not sequence your DNA. An array interrogates a fixed panel of roughly 600,000 to 900,000 pre-selected positions with allele-specific probes. It is cheap and accurate at those positions, and blind everywhere else. That blind fraction is about 99.97 percent of your genome. Array output is a tab-delimited text file of rsID, chromosome, position, and genotype. It cannot report a novel variant, because novelty is outside the probe set by construction.

Whole-genome sequencing shears the genome into fragments, sequences them in massive parallel, and reconstructs genotypes by aligning reads to a reference. The library preparation step determines much of what you get. Adapter ligation, indexing, and amplification choices set duplicate rates and coverage uniformity. The protocols underlying standard Illumina libraries are well documented ¹. The practical consequences are that sequencing sees small insertions and deletions, can be reanalyzed with better callers, and supports structural variant discovery that an array cannot approach. The tradeoff is cost, data volume (roughly 100 GB for a 30x CRAM plus FASTQ), and the fact that you now own a genuine analysis problem rather than a lookup table.

Exome sequencing sits between the two. It captures the roughly 1 to 2 percent of the genome that codes for protein, at 100x or deeper, and it remains a reasonable choice when the question is purely about coding variants. It misses deep intronic splice variants, most regulatory variation, and mitochondrial and repeat-expansion content unless the panel is deliberately extended. The conceptual ancestor of exome-focused approaches is expressed-sequence-tag sequencing, which sampled transcribed regions to find genes long before whole genomes were tractable ².

Coverage, platform, and what “30x” hides

Mean coverage is the number everyone quotes and the weakest single descriptor of quality. A 30x genome with poor uniformity can leave thousands of exons below 10x, which is where genotype calls become unreliable. Ask for the fraction of the callable genome at 10x and at 20x, the duplicate rate, and the insert size distribution. For short-read Illumina data at 30x, a reasonable set of targets is over 95 percent of the genome at 10x. The PCR duplicate rate should stay under 10 percent for a PCR-free library, with median insert size around 350 to 450 bp. The reason this level of detail matters is that error modes in high-throughput sequencing are systematic rather than random: GC-extreme regions, homopolymers, and segmental duplications fail in predictable places, and coverage summaries hide exactly those places ³.

Read length is the other axis. Short reads of 2x150 bp cannot uniquely align in long repeats. Short-read genomes therefore under-call structural variants, large repeat expansions, and paralogous genes. Long-read platforms (PacBio HiFi, Oxford Nanopore) resolve those regions and phase haplotypes over tens of kilobases, at higher cost per sample. Our view is that a PCR-free 30x short-read genome is the sensible default if the purpose is a durable personal baseline and you can afford one sequencing run. The ecosystem of callers, annotation, and reference cohorts is built around it. Add long reads later if a specific question requires phasing or a repeat locus. Interpreting any structural call also requires knowing what is common in the population, which is what curated catalogs of structural variation exist to tell you ⁴.

The files to demand, and what to do with them

Insist on these, in writing, before you pay:

• FASTQ (gzipped) or unaligned BAM with read groups intact.
• CRAM or BAM aligned to a named reference, ideally GRCh38 with ALT contigs and decoys, or a T2T-CHM13 alignment if the provider offers it.
• A gVCF as well as the filtered VCF. The gVCF records reference confidence at non-variant positions, which is the only way to distinguish “you are homozygous reference here” from “we could not see this position.” The filtered VCF alone does not show that difference.
• QC metrics: the output of samtools stats, Picard CollectWgsMetrics, and per-sample bcftools stats.

A reasonable first pass on your own machine or a cloud instance with 16 cores and 64 GB of RAM looks like this. Verify integrity and identity first:

samtools quickcheck -v sample.cram
samtools stats sample.cram > stats.txt
bcftools gtcheck -g array_genotypes.vcf.gz sample.vcf.gz

That third command catches sample swaps if you also have an old array file, which is the single most consequential failure mode in personal genomics and the easiest to detect. Then re-call from the CRAM with an independent pipeline so you are not trusting one caller’s idiosyncrasies. DeepVariant is our default for short reads:

run_deepvariant --model_type=WGS --ref=GRCh38.fa \
  --reads=sample.cram --output_vcf=dv.vcf.gz \
  --output_gvcf=dv.g.vcf.gz --num_shards=16

Annotate with Ensembl VEP, pulling in gnomAD allele frequencies and ClinVar, and then filter hard on frequency before you read a single gene name:

vep -i dv.vcf.gz -o annotated.vcf --cache --assembly GRCh38 \
  --everything --vcf --custom gnomad.genomes.vcf.gz,gnomADg,vcf,exact,0,AF

The discipline that separates useful analysis from anxiety is filtering. A single genome contains roughly 4 to 5 million variants against GRCh38, of which perhaps 10,000 to 12,000 alter protein coding sequence and tens are genuinely rare and predicted to be damaging. Most of those tens are still benign. Any variant you take seriously should have a ClinVar assertion with a submitter you can name, a review status of at least two stars, and a gnomAD allele frequency consistent with the claimed penetrance. Rare-variant burden methods exist precisely because single rare variants are individually uninformative in one person and require aggregation across a gene and a cohort to yield a statistical signal ⁵.

Where a clinician is not optional

A research-grade genome is a measurement. Turning a measurement into a medical decision requires a clinical laboratory report and a genetics professional, and we will not pretend otherwise. If your analysis surfaces a plausible pathogenic variant in a gene with actionable consequences, the correct next step is confirmatory testing in a CLIA-certified or equivalently accredited clinical laboratory. Order it through a clinician and follow it with genetic counseling. Sanger confirmation of a single position costs a few hundred dollars and resolves false positives that arise from alignment artifacts in paralogous regions.

Two further points deserve saying plainly. Whole-genome data reveals information you did not ask for, including carrier status, unexpected family relationships, and risk information about relatives who never consented, and these are recognized ethical features of the technology rather than edge cases ⁶. Professional guidance on reporting findings unrelated to the original indication is deliberately cautious, and European Society of Human Genetics recommendations argue against broad opportunistic screening outside a considered framework ⁷. Decide in advance what you want to know, write it down, and revisit it later rather than in the middle of reading a variant table.

Privacy is an engineering question you control. Keep the CRAM and FASTQ encrypted at rest, avoid uploading raw data to third-party interpretation sites whose terms permit research sharing, and prefer providers who will delete their copy on request and say so contractually. The legal terrain around sequence analysis has shifted over the past decade, and gene patents no longer constrain analysis in the United States the way they once did ⁸.

Questions people also ask

How much does a whole genome DNA test cost? Consumer 30x whole-genome sequencing with raw data typically runs a few hundred to about a thousand dollars in 2026. Clinical-grade sequencing with an accredited laboratory report and clinician involvement runs considerably higher. A longitudinal profile that adds transcriptomics, proteomics, and biomarkers is a different product at a different price.

Which DNA test is best for health issues? For a specific suspected condition, targeted clinical testing ordered by a clinician, because it comes with an accredited interpretation and confirmation. For a general-purpose personal baseline you intend to reanalyze, 30x whole-genome sequencing with raw files, since it subsumes every panel that might be ordered later.

Which DNA company is the safest? Judge on contract terms rather than brand. Does the provider delete your sample and data on request? Does it refuse law-enforcement requests absent a warrant? Does it sell or share de-identified data, and is it accredited for the claims it makes? Providers that return raw data and let you leave with it reduce your dependence on their long-term behavior.

What does genome testing mean? It means examining DNA sequence to identify variants, either at selected positions (genotyping) or across the whole sequence (sequencing). The interpretive framework connecting variants to biology is the direct product of the Human Genome Project and the reference maps it produced ⁹.

Footnotes

1. Matthias Meyer, Martin Kircher. Illumina Sequencing Library Preparation for Highly Multiplexed Target Capture and Sequencing. Cold Spring Harbor Protocols, 2010. https://doi.org/10.1101/pdb.prot5448 ↩

2. Mark D. Adams, Jenny M. Kelley, Jeannine D. Gocayne, et al. Complementary DNA Sequencing: Expressed Sequence Tags and Human Genome Project. Science, 1991. https://doi.org/10.1126/science.2047873 ↩

3. 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 ↩

4. Jeffrey R. MacDonald, Robert Ziman, Ryan K. C. Yuen, et al. The Database of Genomic Variants: a curated collection of structural variation in the human genome. Nucleic Acids Research, 2013. https://doi.org/10.1093/nar/gkt958 ↩

5. Michael C. Wu, Seunggeun Lee, Tianxi Cai, et al. Rare-Variant Association Testing for Sequencing Data with the Sequence Kernel Association Test. The American Journal of Human Genetics, 2011. https://doi.org/10.1016/j.ajhg.2011.05.029 ↩

6. Wim Pinxten, Heidi Carmen Howard. Ethical issues raised by whole genome sequencing. Best Practice & Research Clinical Gastroenterology, 2014. https://doi.org/10.1016/j.bpg.2014.02.004 ↩

7. On behalf of the European Society of Human Genetics, Guido de Wert, Wybo Dondorp, et al. Opportunistic genomic screening. Recommendations of the European Society of Human Genetics. European Journal of Human Genetics, 2020. https://doi.org/10.1038/s41431-020-00758-w ↩

8. Robert Cook-Deegan, Subhashini Chandrasekharan. Patents and Genome-Wide DNA Sequence Analysis: Is it Safe to Go into the Human Genome?. Journal of Law, Medicine & Ethics, 2014. https://doi.org/10.1111/jlme.12161 ↩

9. Francis S. Collins. Implications of the Human Genome Project for Medical Science. JAMA, 2001. https://doi.org/10.1001/jama.285.5.540 ↩