What it costs to sequence a human genome, and what that price does not include

A 30x whole human genome costs roughly $200 to $700 today if you buy it as a research service and accept raw data with no interpretation. University core facilities advertise 30x with FASTQ delivery in the $200 to $400 range. Direct-to-consumer providers charge $400 to $1,000 for the same coverage with a web report attached. A CLIA/CAP-certified clinical genome with a report signed by a molecular geneticist runs $1,500 to $5,000. That gap is mostly regulatory and labor. Sequencing chemistry is not what drives it. The $100 genome that gets discussed is a reagent figure at high instrument utilization. It is not a price anyone quotes you.

The number people usually want is the one that includes the parts nobody advertises: sample handling, analysis compute, storage for a decade, and the confirmatory test you need before any result touches a clinical decision.

Where the money goes in a $300 genome

Break a 30x research genome into its parts:

• Library prep: $20 to $60 per sample in reagents (PCR-free TruSeq or equivalent), plus a few minutes of tech time. PCR-free matters. PCR-based prep introduces GC bias and duplicate reads that cost you effective coverage in exactly the promoter and first-exon regions you care about.
• Sequencing: on a NovaSeq X with a 25B flow cell at 2x150, a full run yields several terabases and amortizes to roughly $200 per 30x genome when the flow cell is filled. The core facility prices you see reflect that amortization. A half-empty flow cell doubles your cost.
• QC and demultiplexing: small, but it is where samples get swapped. Ask whether the lab runs a fingerprint concordance check.
• Data delivery: usually free download for 30 days, then you are on your own.

Coverage is the lever with the largest price effect. 15x costs about half of 30x and is adequate for common-variant genotyping and polygenic scores, but single-sample germline calling degrades noticeably: heterozygote sensitivity drops, and small indels in homopolymer runs become unreliable. 30x is the standard because it is where per-base heterozygous sensitivity plateaus for short reads. Going to 60x buys you very little on short reads. If you want more, the next real step is long reads (PacBio HiFi or ONT), which costs $800 to $2,000 per genome and resolves structural variants, phasing, and repeat expansions that short reads systematically miss.

”Mapping” means two different things

When Smith and Hood laid out the strategy in 1987, mapping and sequencing were separate problems solved in sequence. The plan was to build a physical map of ordered clones and then sequence along it ¹. That sense of the word is obsolete for individuals. Nobody maps your genome de novo. Your reads are aligned to a reference that already exists.

The second sense of “mapping” is read alignment, which is what happens to your data. That is a compute job, and it is cheap. The third sense, the one marketing departments use, is SNP genotyping on an array: 600,000 to 900,000 probed positions for $50 to $200. An array is a sparse sample of known common variation, descended conceptually from the 1.42-million-SNP map published in 2001 ². It cannot see your rare variants, cannot call de novo mutations, and cannot report anything outside its probe set. If someone sells you “gene mapping” for under $200, check whether the deliverable is a genotype table or a FASTQ.

What you pay after the lab

The lab invoice is not the total. Budget these:

Analysis compute. Aligning 30x paired-end reads with bwa-mem2 mem -t 32 -K 100000000 -Y takes about two to three hours on a 32-vCPU instance. Sorting and duplicate marking with samtools sort -@ 16 -m 4G and samtools markdup adds an hour. DeepVariant on the same machine, --model_type=WGS, runs one to three hours. On spot instances at $0.40 to $0.80 per hour, the full FASTQ-to-gVCF path costs $5 to $15 of compute. Illumina DRAGEN on an FPGA instance does the same work in about 30 minutes for a license fee that only makes sense at volume. If you are doing one genome, use the open tools.

Storage. A 30x BAM is 60 to 80 GB. The same data as CRAM 3.0 with the reference specified (samtools view -C -T GRCh38.fa -@ 8) is 15 to 20 GB, lossless. Gzipped FASTQ is 40 to 60 GB and you should not keep it once you have a verified CRAM. At S3 Standard pricing (~$0.023/GB/month), 20 GB is about $5.50 a year. In Glacier Deep Archive it is about 25 cents a year, with a retrieval delay measured in hours. Keep two copies in different places. The single most common way people lose their genome is letting the provider’s download link expire and never checking the checksum on the local copy.

Confirmation. Any variant you would act on clinically needs orthogonal confirmation in a CLIA laboratory, typically Sanger sequencing at $200 to $500 per site, ordered through a clinician. Research-grade calls, including those from a $221 core facility genome, are not a clinical result regardless of how clean the read pileup looks. If a variant in your data appears in ClinVar as pathogenic, the next step is a genetic counselor. It is not a purchase.

What the price has and has not done

The $1,000 genome was set as an explicit engineering target in the mid-2000s, when the state of the art was cyclic reversible-termination chemistry producing short reads at costs orders of magnitude above that goal ³. That target was met around 2014 and then stalled for most of a decade, which is the frustration behind the Reddit threads. The stall was an instrument-monopoly story. Physics did not cause it. Prices moved again in 2023 to 2024 when patterned flow cells at higher density and competing platforms arrived.

What has not fallen is interpretation cost. A 30x genome yields roughly 4 to 5 million variants relative to GRCh38, of which perhaps 50,000 to 100,000 are rare and coding-adjacent, and a few dozen are annotated with any clinical assertion. Filtering that down is analyst time, and analyst time has gotten more expensive. It has not gotten cheaper. The marginal value of your genome comes from the annotation layer (gnomAD allele frequencies, ClinVar, SpliceAI, pharmacogenomic star-allele callers like PharmCAT) and from having other measurements on the same person to condition on.

What we would buy

Buy 30x PCR-free short-read WGS from a lab that delivers FASTQ or CRAM plus an MD5 manifest, aligned to GRCh38 with alt contigs or to T2T-CHM13 if offered. Insist on raw data in the contract before you pay. Verify checksums the day you download, convert to CRAM, and put one copy in cold storage and one on a disk you control. Run DeepVariant yourself rather than trusting a vendor’s GATK settings you cannot inspect. Expect to spend $300 to $600 all-in for the data and under $50 a year to keep it.

Skip arrays if you can afford WGS. Skip 15x unless you are genotyping common variants only. Add long reads later if a specific question requires phasing or a repeat locus. Do not add them as a default.

Questions people also ask

Can I map my own genome? You can buy the sequencing and do the alignment and variant calling yourself with open tools. What you cannot do at home is the wet lab, and what you should not do is treat your own calls as a clinical result. Interpretation that would change a medical decision belongs with a genetic counselor or clinical geneticist.

How much does a gene mapping test cost? If “gene mapping” means a SNP array, $50 to $200 for several hundred thousand probed positions. If it means whole-genome sequencing, $200 to $700 research-grade and $1,500 to $5,000 clinical. If it means a targeted panel for a specific set of genes, $300 to $2,000 depending on gene count and whether it is clinical.

Is the human genome fully mapped? The reference is. The 2001 draft had gaps, GRCh38 still has some, and the T2T-CHM13 assembly published in 2022 closed them for one haploid genome. Your genome is a different matter: short reads leave segmental duplications, large repeat expansions, and many structural variants unresolved, which is the main argument for long reads.

Why did the price stop falling between 2015 and 2022? Market structure. The chemistry improvements existed but there was little competitive pressure to price them through. The move from the $1,000 genome toward the $200 genome tracked the arrival of higher-density flow cells and rival platforms. It did not track a step change in sequencing physics ³.

Footnotes

1. L. C. Smith, Leroy Hood. Mapping and Sequencing the Human Genome: How to Proceed. Nature Biotechnology, 1987. https://doi.org/10.1038/nbt0987-933 ↩

2. 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 ↩

3. Simon T. Bennett, C.L. Barnes, Anthony J. Cox, et al. Toward the $1000 Human Genome. Pharmacogenomics, 2005. https://doi.org/10.1517/14622416.6.4.373 ↩ ↩²