The Human Genome Revolution: From $3 Billion to $200 and What We Found

In 1989, a committee report to the U.S. Congress estimated that mapping and sequencing the entire human genome would require 15 years and cost $3 billion — roughly $1 per base pair across the 3 billion base pairs in the human genome. Critics called it the most ambitious project in biology since the moon landing. Supporters called it the moon landing. Both groups underestimated what would actually be found inside. The Human Genome Project finished in 2003, roughly on budget and two years ahead of schedule. Twenty years later, a second group completed what the first could not: a gapless, telomere-to-telomere sequence of the human genome that revealed almost 200 million base pairs the original project had missed. What those sequences do, and what we still don't understand about them, is reshaping medicine, evolutionary biology, and the ethics of who gets to know what about whom.

The Human Genome Project: What It Was and What It Cost

The Human Genome Project (HGP) was an international scientific research effort formally launched in October 1990, coordinated primarily by the U.S. National Institutes of Health (NIH) and the U.S. Department of Energy, with major contributions from the United Kingdom, France, Germany, Japan, and China. Its goal was straightforward to state and fiendishly difficult to execute: determine the complete sequence of the approximately 3.2 billion base pairs (adenine, thymine, guanine, cytosine) that constitute the chemical code of the human genome, and identify all the protein-coding genes within it.

Base pair — a complementary pair of nucleotide bases held together by hydrogen bonds in the DNA double helix. Adenine always pairs with thymine; guanine always pairs with cytosine. The sequence of base pairs along a DNA molecule constitutes genetic information.

The sequencing technology of 1990 — gel-based Sanger sequencing, developed by Frederick Sanger in the 1970s — could read approximately 500–800 base pairs per reaction. Sequencing a genome of 3 billion base pairs with this technology required millions of individual reactions, automated but still limited by throughput. The project processed the genome in fragments: chromosomes were cut into smaller segments, each segment was sequenced multiple times from different starting points, and the overlapping sequences were assembled like a jigsaw puzzle using computational algorithms.

The publicly funded HGP was not the only effort. In 1998, Craig Venter's company Celera Genomics announced it would sequence the human genome faster and cheaper using a different method — "whole-genome shotgun sequencing," which skipped the step of physically mapping chromosome regions before sequencing them. The resulting race accelerated both efforts. In June 2000, U.S. President Bill Clinton and British Prime Minister Tony Blair jointly announced the completion of a "working draft" of the genome — about 85 percent complete. The final reference sequence, covering approximately 92 percent of the genome, was published in Nature in April 2003.

Source: International Human Genome Sequencing Consortium (2001), "Initial sequencing and analysis of the human genome," Nature, Vol. 409 — the landmark paper describing the working draft of the human genome sequence.

Source: Collins, F.S. et al. (2003), "A vision for the future of genomics research," Nature, Vol. 422 — the NIH/NHGRI roadmap for genomics following the HGP's completion.

The Cost Collapse: Steeper Than Moore's Law

Moore's Law — the observation that the number of transistors on a microchip roughly doubles every two years — has long been used as the standard example of exponential technological improvement. The cost of genome sequencing has beaten it.

The NHGRI has tracked sequencing costs at its funded centers since 2001. In 2001, sequencing a human genome cost approximately $95 million. By 2007, costs had fallen to around $10 million — following Moore's Law pace, a respectable but unremarkable decline. Then the "next-generation sequencing" (NGS) revolution hit.

Starting in 2007 and accelerating through 2008–2011, new high-throughput sequencing platforms (from companies including Illumina, 454 Life Sciences, and Applied Biosystems) replaced the Sanger method. Instead of reading one sequence at a time, these platforms processed millions of short DNA fragments in parallel. Cost fell by three orders of magnitude in roughly four years — a pace roughly ten times faster than Moore's Law.

Year	Cost per genome (NHGRI data)
2001	~$95,000,000
2006	~$14,000,000
2008	~$1,000,000
2011	~$21,000
2015	~$4,000
2020	~$1,000
2023	~$200 (commercial, consumer-grade)

The consequence is that genome sequencing has transitioned from a once-per-decade research achievement to a routine clinical tool. By 2023, Illumina's NovaSeq X system can sequence a human genome at high accuracy in under 24 hours at a reagent cost approaching $200. For context: a complete blood count panel — a basic clinical test — costs $30–150, and provides information about a handful of parameters. A full genome sequences all 3 billion base pairs and, in principle, encodes everything heritable about a person's biology.

Source: NHGRI (2023), "The Cost of Sequencing a Human Genome," genome.gov — NHGRI's official tracking of sequencing costs from 2001 to 2022, updated approximately annually.

Source: Wetterstrand, K.A. (2023), "DNA Sequencing Costs: Data from the NHGRI Genome Sequencing Program," National Human Genome Research Institute — the dataset underlying NHGRI's cost tracking.

The 8 Percent Gap: What the Original Project Couldn't Read

The 2003 reference human genome, despite being called "complete," was not. Approximately 8 percent of the genome — concentrated in the most repetitive regions — remained unsequenced. These regions include:

• Centromeres: the constriction point of each chromosome, where spindle fibers attach during cell division. Centromeres consist almost entirely of highly repetitive sequences called alpha-satellite DNA, which repeat the same ~171 base-pair unit thousands of times in tandem.
• Telomeres: the end caps of chromosomes, consisting of thousands of repeats of the sequence TTAGGG.
• Segmental duplications: large blocks of sequence that appear in multiple nearly identical copies at different locations, making it impossible to determine which copy each sequence fragment came from.

Short-read sequencing technology — which reads fragments of 100–300 base pairs — cannot assemble repetitive regions because there is no way to determine the order of fragments that are all nearly identical to each other. The problem is like assembling a jigsaw puzzle where every piece has the same picture: the fragments fit everywhere, so they can't be placed with certainty.

Telomere — the protective repetitive DNA sequence at the end of each chromosome. Telomeres shorten with each cell division and are associated with cellular aging. They consist of thousands of repeats of the six-base sequence TTAGGG.

The solution required a different approach: "long-read" sequencing technology, which can read individual DNA molecules tens of thousands of base pairs long. Two platforms — PacBio's HiFi sequencing and Oxford Nanopore's nanopore sequencing — can read through repetitive regions because even if the sequence is repetitive, a single read spanning 20,000 base pairs can establish the context of each repeat within the full array.

Source: Nurk, S. et al. (2022), "The complete sequence of a human genome," Science, Vol. 376 — the T2T-CHM13 paper completing the human genome sequence, published by the Telomere-to-Telomere Consortium.

The T2T Genome: What the Missing 8 Percent Contains

On March 31, 2022, the Telomere-to-Telomere (T2T) Consortium — a group of 99 researchers led by teams at the University of California Santa Cruz (UCSC) and the National Human Genome Research Institute — published a gapless sequence of the human genome in Science. The new reference, designated T2T-CHM13, added approximately 200 million base pairs of novel sequence not present in the previous reference (GRCh38), including:

• 115 million base pairs of satellite DNA in centromeric regions
• Complete sequences of all 22 autosomes (non-sex chromosomes) and chromosome X
• 2,226 new gene copies, including 99 predicted to code for proteins
• Nearly 2,000 additional candidate genes requiring further study
• Over 2 million new genetic variants discovered by using T2T-CHM13 as a reference

The centromere sequences alone had been essentially unknown for decades. Their composition reveals that human centromeres are organized into distinct layers: an outer alpha-satellite repeat array flanked by different types of repetitive sequence, with a specific set of alpha-satellite units (the "active" centromere) that nucleates kinetochore assembly — the protein complex that physically attaches the chromosome to the spindle during cell division. Understanding centromere structure is essential for understanding chromosome segregation errors, which cause aneuploidy (wrong chromosome number in cells) and are the leading cause of miscarriage and a contributor to cancer.

Source: Altemose, N. et al. (2022), "Complete genomic and epigenetic maps of human centromeres," Science, Vol. 376 — the T2T Consortium companion paper specifically analyzing centromere structure and function.

"Junk DNA" Is Not Junk

One of the most persistent misconceptions in popular genetics is that the majority of the human genome is "junk" — non-functional remnants of evolution that serve no purpose. The term came from a 1972 paper by Susumu Ohno, who applied it specifically to noncoding DNA to contrast it with protein-coding genes. It was never meant as a formal classification, but it stuck.

Protein-coding genes occupy approximately 1.5–2 percent of the human genome. The ENCODE project (Encyclopedia of DNA Elements), a large-scale effort launched by NIH in 2003 and still ongoing, has produced evidence that at least 80 percent of the human genome is transcribed or shows biochemical activity of some kind — though what fraction of that activity is functionally significant remains debated.

What was "junk" has turned out to contain:

Regulatory elements — short DNA sequences that control when, where, and how much a nearby gene is expressed. Enhancers can act on genes millions of base pairs away, looping through three-dimensional space to interact with promoters. Many disease-causing genetic variants map not to protein-coding genes but to nearby regulatory sequences.

Transposable elements (TEs) — mobile genetic elements that have copied themselves throughout the genome over evolutionary time. Roughly 45–50 percent of the human genome consists of TE-derived sequences. Many are now "domesticated" — their sequences have been repurposed by the host genome for regulatory functions. The syncytins, proteins essential for forming the placenta in mammals, are derived from retroviral envelope genes that integrated into the mammalian genome approximately 25 million years ago.

Long non-coding RNAs (lncRNAs) — RNA molecules transcribed from DNA but not translated into protein, numbering in the tens of thousands. Functions have been identified for relatively few, but established lncRNAs regulate X-chromosome inactivation (XIST), chromatin structure, and immune responses.

The T2T genome revealed that the previously unsequenced centromeric regions contain massive arrays of satellite repeats that are not transcriptionally inert: they are transcribed into RNA molecules that play structural roles in centromere function. The "junk" was regulating the machinery that keeps chromosomes intact during every cell division.

Source: ENCODE Project Consortium (2012), "An integrated encyclopedia of DNA elements in the human genome," Nature, Vol. 489 — the landmark ENCODE paper claiming biochemical activity across 80 percent of the genome, sparking ongoing debate about functional significance.

Source: Dupressoir, A. et al. (2012), "From ancestral infectious retroviruses to bona fide cellular genes: role of the captured syncytins in placentation," Placenta, Vol. 33 — evidence that placental syncytin proteins are derived from endogenous retroviral sequences.

Pharmacogenomics: Matching Drugs to Genomes

One of the most immediately practical consequences of the genomic revolution is pharmacogenomics — the study of how genetic variation affects an individual's response to drugs.

Pharmacogenomics — the field that examines how a person's genetic makeup influences their response to medications, including whether a drug will work, what dose is needed, and whether adverse effects are likely.

The metabolic rate of most drugs is controlled by enzymes in the liver, encoded by genes in the cytochrome P450 (CYP) family. Common variants in these genes — variants carried by tens of millions of people — dramatically alter how fast or slowly a drug is metabolized. The most-studied example is CYP2C19, which metabolizes the blood-thinner clopidogrel (Plavix). Approximately 25–30 percent of people of East Asian ancestry, and 2–5 percent of people of European ancestry, carry loss-of-function variants in CYP2C19 that prevent them from converting clopidogrel to its active form. For these individuals, the standard dose of clopidogrel does not work as a blood thinner — a clinically significant difference that affects the standard of care after heart attack or stroke.

Similar pharmacogenomic effects have been documented for antidepressants, warfarin (a blood thinner), codeine (a painkiller that can be lethally toxic in "ultra-rapid" metabolizers, who convert it to morphine too quickly), and certain cancer chemotherapies. The FDA has added pharmacogenomic information to the labels of more than 300 approved drugs.

Breast cancer provides the most widely known example of genomic precision medicine. BRCA1 and BRCA2 — genes involved in DNA repair — when carrying certain harmful variants, confer a lifetime risk of breast cancer of approximately 69 percent (BRCA1) and 45 percent (BRCA2), compared to the general population risk of around 12 percent. This information enables risk-reduction surgery, enhanced surveillance, and selection of chemotherapy agents that exploit BRCA-deficient cancer cells' inability to repair DNA damage.

Source: Relling, M.V. & Evans, W.E. (2015), "Pharmacogenomics in the clinic," Nature, Vol. 526 — comprehensive review of clinical pharmacogenomics, CYP variants, and drug-label guidance.

Source: Kuchenbaecker, K.B. et al. (2017), "Risks of Breast, Ovarian, and Contralateral Breast Cancer for BRCA1 and BRCA2 Mutation Carriers," JAMA, Vol. 317 — the large study establishing lifetime breast cancer risk estimates for BRCA1/2 carriers.

The Ethics of Reading the Genome

The ability to read the human genome rapidly and cheaply does not settle questions about who should read it, under what circumstances, and who should have access to the results.

Genetic discrimination — the use of genetic information to make decisions about employment, insurance, or other aspects of a person's life — is prohibited in the United States for health insurance and employment under the Genetic Information Nondiscrimination Act (GINA) of 2008. GINA does not cover life insurance, disability insurance, or long-term care insurance. In the UK, the Association of British Insurers maintains a moratorium on requesting genetic test results for most insurance purposes, but the legal landscape varies significantly across countries.

Incidental findings — genomic variants discovered during clinical sequencing that are unrelated to the original clinical question but have significant health implications — present a disclosure dilemma. A patient sequenced to diagnose a suspected rare disorder might also have variants associated with hereditary cancer risk or cardiac arrhythmia. The American College of Medical Genetics and Genomics (ACMG) recommends that laboratories report pathogenic variants in 73 specified genes regardless of the original clinical indication — but patients can opt out of receiving some categories of results.

Ancestry databases and privacy — consumer genomics companies (23andMe, AncestryDNA) have collected genetic data from tens of millions of individuals. This data is valuable for research but also creates privacy risks: genetic data can identify individuals even when names are removed, because the genome is inherently linked to biological relatives, who have not consented to be identified. In 2023, 23andMe disclosed a data breach that exposed the genetic and ancestry data of approximately 6.9 million users. In 2024, facing bankruptcy, the company filed for Chapter 11 protection, raising urgent questions about what happens to customer genetic data when a consumer genomics company fails.

The framework for governing genomic data — who owns it, who can access it, who profits from discoveries made using it — remains incomplete in every jurisdiction. The science has outpaced the regulatory infrastructure by at least a decade.

Source: Hudson, K.L. et al. (2008), "Keeping pace with the times — the Genetic Information Nondiscrimination Act of 2008," New England Journal of Medicine, Vol. 358 — analysis of GINA's scope, limitations, and significance.

What Sequencing Revealed About Human Evolution

The complete human genome contains a detailed record of human evolutionary history, and since the HGP, genomics has rewritten several chapters of that history.

Ancient DNA sequencing — extracting and sequencing DNA from archaeological specimens — revealed that modern humans interbred with Neanderthals before Neanderthals went extinct approximately 40,000 years ago. People of non-African ancestry carry approximately 1–2 percent Neanderthal DNA in their genomes. People of Melanesian and Papua New Guinean ancestry carry an additional 4–6 percent DNA from another archaic human group, the Denisovans, known from a single cave in Siberia and a tooth fragment from Tibet.

These findings, enabled by advances in ancient DNA recovery and next-generation sequencing, overturned the simple "out of Africa" model in which modern humans replaced other hominids without interbreeding. The genome is a historical document — and the more completely we read it, the more complex the history it records.

Comparative genomics — comparing the human genome with those of other species — has been equally productive. The human and chimpanzee genomes differ by approximately 1.2 percent in single base-pair substitutions, consistent with a common ancestor approximately 6–7 million years ago. But structural differences (insertions, deletions, inversions of large chromosomal segments) add substantially to the functional divergence. Human Accelerated Regions (HARs) — parts of the genome that are conserved across mammals but show unusually rapid evolution in humans — include sequences that regulate brain development genes and may underlie distinctively human cognitive capacities.

Source: Prüfer, K. et al. (2014), "The complete genome sequence of a Neanderthal from the Altai Mountains," Nature, Vol. 505 — the high-quality Neanderthal genome enabling fine-scale analysis of human-Neanderthal introgression.

The Takeaway

The Human Genome Project delivered a first reading of the human genome. The T2T consortium delivered a complete one. The cost collapse from $3 billion to $200 transformed sequencing from a once-per-civilization project to a routine clinical tool. What the complete sequence has revealed — that "junk DNA" regulates the machinery of life, that centromeres are structurally complex, that our genome carries traces of archaic human relatives — has complicated every simple story told about the genome. Pharmacogenomics is beginning to translate this complexity into clinical benefit, one patient at a time. The ethical frameworks needed to govern what we know, and who gets to know it, are lagging behind the science — a gap that will require as much careful work as the sequencing itself.

Frequently Asked Questions

Q: How many genes does the human genome contain?
A: The current best estimate is approximately 19,000–20,000 protein-coding genes — far fewer than the 100,000 once predicted. The T2T-CHM13 sequence added 99 new predicted protein-coding genes and nearly 2,000 additional candidates. Non-protein-coding genes (lncRNAs, microRNAs, etc.) number in the tens of thousands additional.

Q: What is the difference between the Human Genome Project and the T2T genome?
A: The HGP (completed 2003) sequenced approximately 92 percent of the human genome, leaving the most repetitive regions — centromeres, telomeres, segmental duplications — unread. The T2T Consortium (2022) used long-read sequencing technology to fill all remaining gaps, adding approximately 200 million new base pairs and producing the first truly gapless human genome sequence.

Q: Why did genome sequencing costs fall so fast?
A: The introduction of next-generation sequencing platforms between 2007 and 2011 replaced gel-based Sanger sequencing with massively parallel short-read sequencing, reducing cost by three orders of magnitude in roughly four years. Further engineering improvements in optics, chemistry, and computational assembly have continued to reduce costs, reaching under $200 per genome commercially by the mid-2020s.

Q: What is pharmacogenomics and why does it matter?
A: Pharmacogenomics studies how genetic variants affect drug metabolism and response. Variants in liver enzyme genes (CYP family) can make standard drug doses ineffective or toxic depending on how fast a person metabolizes the drug. FDA labels for more than 300 drugs now include pharmacogenomic information. Knowing a patient's relevant genetic variants before prescribing can avoid serious adverse drug reactions.

Q: Is "junk DNA" really junk?
A: No. "Junk DNA" was a colloquial label for the ~98 percent of the human genome that does not code for proteins. Research since the HGP has shown that large fractions of this sequence regulate gene expression, define chromosome structure, contribute RNA molecules with regulatory functions, and contain evolutionary records of ancient viral infections that have been repurposed for biological functions. The term is now considered scientifically inaccurate.

Q: Can anyone have their genome sequenced today?
A: Consumer genomic services (23andMe, AncestryDNA) sequence portions of the genome (typically 600,000–700,000 specific variants) for $100–$200. Clinical whole-genome sequencing is available through healthcare providers for diagnostic purposes, typically at costs of $1,000–$5,000 including interpretation. Direct-to-consumer whole-genome sequencing (reading all 3 billion base pairs) is available from several companies for under $300, though the clinical interpretation of most variants remains uncertain.

---

Read next:

• CRISPR Gene Editing Explained: How It Works
• How Nuclear Fission Works Explained
• How the Nervous System Works