How to Find Out Your Own Genetics

If you want to know your genetics and you are willing to do the analysis yourself, get a 30x PCR-free whole-genome sequence and insist on the raw FASTQ or an aligned CRAM plus a per-sample gVCF. That single dataset subsumes every consumer array, every ancestry panel, and most clinical gene panels, and it does not expire: as annotation databases improve, you re-run the interpretation on files you already own. Arrays cost less and answer far less. Exomes are a reasonable middle if you only care about coding variation and want a clinical-grade report. Nothing you do on your own laptop is diagnostic, and any result you would act on needs confirmation in a CLIA/CAP lab and a conversation with a genetic counselor or physician.

Array, exome, or whole genome

A consumer genotyping array (what 23andMe and AncestryDNA run) is a fixed probe set, typically 600k to 900k sites out of roughly 3.1 billion bases. It measures the sites the chip designer chose, mostly common variation, cataloged in principle since the first large-scale SNP maps two decades ago ¹. It cannot see a variant that is not on the chip, and the raw text file it gives you is genotype calls with no read evidence behind them. Array data is also the wrong substrate for anything rare: a de novo missense in a single gene, a 40-base deletion, a repeat expansion, all invisible.

Exome capture sequences the ~1.5% of the genome that codes for protein, usually at 100x mean depth. It catches most known pathogenic variation because most curated pathogenic variants are coding. Its weakness is capture bias: GC-rich first exons drop out, and coverage across clinically relevant genes is uneven in ways that depend on the kit.

Whole genome at 30x with a PCR-free library is what we run. Coverage is far more uniform than capture, you get structural variants and intronic splice variants, and you can revisit the data when a gene gets curated in 2029. Library preparation matters more than people expect: fragmentation method measurably changes coverage uniformity across clinically relevant genes, which is the difference between a callable exon and a gap you have to Sanger-fill ². Ask the lab whether the library is PCR-free and how they fragment (enzymatic versus acoustic).

Depth is the other lever. 30x is the standard for germline SNV and indel calling. Deeper sequencing mainly buys you rare and structural variant sensitivity, which is the point of large deep-coverage cohorts like the 1,070-genome Japanese reference panel that recovered millions of novel variants absent from existing databases ³. For one person, 30x is the right price-performance point.

Files to insist on before you pay

Get this in writing from the provider. Many consumer sequencing companies sell you a report and treat the underlying data as an upsell or withhold it entirely.

• FASTQ (gzipped, R1/R2) or a CRAM aligned to GRCh38 with the reference explicitly named, including whether ALT contigs and decoys were used.
• A gVCF, not only a filtered VCF. The gVCF records reference-confident blocks, so you can distinguish “reference at this position” from “no coverage here.”
• The BAM/CRAM index and a QC report: mean depth, percent of genome at ≥10x and ≥20x, duplicate rate, insert size distribution, contamination estimate.
• The exact pipeline name and version (DRAGEN 4.x, GATK 4.5, etc.). Reprocessing is easier when you know what was done.

A 30x CRAM is roughly 15-20 GB, FASTQs are 40-60 GB gzipped. Budget storage accordingly and keep two copies.

The pipeline we would run

Start from FASTQ if you have it, because you control the reference build.

bwa-mem2 mem -t 32 -K 100000000 -Y -R '@RG\tID:s1\tSM:me\tPL:ILLUMINA' \
  GRCh38_full_analysis_set_plus_decoy_hla.fa r1.fq.gz r2.fq.gz \
| samtools sort -@ 8 -o me.bam -
samtools index me.bam
gatk MarkDuplicates -I me.bam -O me.md.bam -M dup.txt

For calling, use DeepVariant rather than GATK HaplotypeCaller on a single sample. It has better indel precision on Illumina data out of the box and needs no joint-calling cohort:

run_deepvariant --model_type=WGS --ref=GRCh38.fa \
  --reads=me.md.bam --output_vcf=me.vcf.gz --output_gvcf=me.g.vcf.gz \
  --num_shards=32

Then normalize before you annotate anything, or half your ClinVar matches will silently fail:

bcftools norm -m -any -f GRCh38.fa me.vcf.gz -Oz -o me.norm.vcf.gz
bcftools index me.norm.vcf.gz

QC with mosdepth --by 1000 --fast-mode me me.md.bam and check contamination with VerifyBamID2. If your cross-sample contamination estimate is above ~2%, stop and ask the lab to re-run.

Annotate with Ensembl VEP against GRCh38, pulling ClinVar, gnomAD v4 allele frequencies, and a splice predictor:

vep -i me.norm.vcf.gz --cache --assembly GRCh38 --everything \
  --custom clinvar.vcf.gz,ClinVar,vcf,exact,0,CLNSIG,CLNREVSTAT \
  --plugin SpliceAI --vcf -o me.vep.vcf

Short-read callers fail predictably in specific places. CYP2D6 needs a dedicated caller (Cyrius or Aldy) because of the CYP2D7 pseudogene. SMN1/SMN2 copy number needs SMNCopyNumberCaller. Repeat expansions (HTT, FMR1, C9orf72) need ExpansionHunter with a variant catalog. HLA typing needs HLA-LA or similar. If you skip these, you will conclude “nothing found” in regions your pipeline never covered.

What to read first, and what to ignore

Filter by evidence quality, not by effect size claims. A pragmatic ordering:

1. ClinVar pathogenic and likely pathogenic with two or more stars of review status, in genes with established disease association. Cross-check each hit against gnomAD: a “pathogenic” allele at 3% frequency in any population is almost certainly miscurated.
2. Pharmacogenetics: star alleles in CYP2C19, CYP2D6, DPYD, TPMT, SLCO1B1, and the corresponding CPIC guideline level. Report to a clinician, do not act on it yourself.
3. Carrier status for recessive conditions, which matters for reproductive planning and is the thing genetic counselors are best at contextualizing.
4. Polygenic scores, last and skeptically. They are population-relative, poorly transferable across ancestries, and sensitive to which weights file you use. plink2 --score will happily produce a number that means very little for you individually.

One category confusion worth naming: germline sequencing of your blood or saliva tells you what you inherited. It says nothing about somatic mutations in a tumor, which is a separate assay on tumor tissue, and the mutational picture there can be substantially different from what germline testing implies ⁴.

Where a clinician is required

Everything above is measurement. Interpretation that changes medical decisions is not something to do alone. A variant you find in a research-grade pipeline needs orthogonal confirmation in a CLIA/CAP laboratory before any clinician will act on it, and a genetic counselor should walk you through penetrance, family implications, and what the finding means for relatives. Findings in the ACMG secondary-findings gene set (about 80 genes covering hereditary cancer syndromes, cardiomyopathies, arrhythmias, and familial hypercholesterolemia) are exactly the class where a self-directed read is insufficient.

Questions people also ask

How can I get free genetic testing? Research programs are the main route. The NIH All of Us Research Program returns hereditary disease risk and pharmacogenetic results at no cost to participants, and regional population genomics programs (Geisinger MyCode, HealthPartners myGenetics) do similar. The tradeoff is that you get a curated report on a limited gene set, usually not raw data, and the timeline is months.

What qualifies you for genetic testing under insurance? Typically a personal or family history meeting published criteria (NCCN criteria for hereditary cancer panels, for example), an unexplained phenotype consistent with a genetic condition, or a specific drug decision with a pharmacogenetic indication. Curiosity does not qualify. Coverage decisions hinge on the ordering clinician’s documentation of medical necessity, so a referral to a genetics clinic is usually the first step.

How much does genetic testing cost for autoimmune disease? There is no single test. Monogenic immune dysregulation panels ordered through a clinical lab commonly run in the low thousands billed, often much less self-pay. For common autoimmune conditions the genetic contribution is polygenic plus HLA, and no test currently changes management, which is why insurers rarely cover it.

How do I find out my genetics without a doctor? Buy a direct-to-consumer 30x whole genome from a provider that ships FASTQ or CRAM, run the pipeline above, and treat every result as a hypothesis to confirm clinically. You will learn a great deal about ancestry, carrier status, and pharmacogenetics. You will not get a diagnosis, and you should not try to make one.

Oak builds longitudinal molecular profiles of individuals: whole-genome sequencing, RNA sequencing, proteomics, blood biomarkers, and continuous glucose data, integrated into one model of you. Build your profile.

Footnotes

1. The International SNP Map Working Group, Cold Spring Harbor Laboratories:, Ravi Sachidanandam, et al. A map of human genome sequence variation containing 1.42 million single nucleotide polymorphisms. Nature, 2001. https://doi.org/10.1038/35057149 ↩

2. Vanessa Process, Madana M.R. Ambavaram, Sameer Vasantgadkar, et al. Optimization of DNA Fragmentation Techniques to Maximize Coverage Uniformity of Clinically Relevant Genes Using Whole Genome Sequencing. Diagnostics, 2025. https://doi.org/10.3390/diagnostics15182294 ↩

3. Masao Nagasaki, Jun Yasuda, Fumiki Katsuoka, et al. Rare variant discovery by deep whole-genome sequencing of 1,070 Japanese individuals. Nature Communications, 2015. https://doi.org/10.1038/ncomms9018 ↩

4. Australian Pancreatic Cancer Genome Initiative, Nicola Waddell, Marina Pajic, et al. Whole genomes redefine the mutational landscape of pancreatic cancer. Nature, 2015. https://doi.org/10.1038/nature14169 ↩