Human Accelerated Regions
The coherence
that remains.
Some stretches of our DNA changed unusually fast after we split from other apes. They almost never rewrite proteins. They rewrite the switches that time a developing brain. This atlas shows those switches — and a mathematical keep-set they are hypothesized to tune.
- EstablishedExternal historical or biological literature
- Machine verifiedExact or numerical result inside the E47 / QEGT corpus
- HypothesisProposed map — estimable, testable, rejectable
Field and Formal are two registers of the same facts. 47/125 is exact on the E47 carrier — not a biological cognition threshold. Education, not a clinical report.
Live contraction · 125 modes
Ω_c
0.376
47/125
0.376
V
125
carrier
0.376
ker
47
keep
0.376
⊥
78
fade
0.376
Invariant
What survives contraction
Imagine 125 possible notes. A rule plays the music over and over, fading everything that is not in a special 47-note chord. The fraction that remains is always 47/125.
| j | j(j+1) | dim | copies × irrep | K² | sector |
|---|---|---|---|---|---|
| 0 | 0 | 1 | V₀ | 32400 | transient |
| 1 | 2 | 9 | 3 V₁ | 12544 | transient |
| 2 | 6 | 25 | 5 V₂ | 0 | kernel |
| 3 | 12 | 28 | 4 V₃ | 11664 | transient |
| 4 | 20 | 27 | 3 V₄ | 19600 | transient |
| 5 | 30 | 22 | 2 V₅ | 0 | kernel |
| 6 | 42 | 13 | V₆ | 186624 | transient |
Genome
Novelty through recombination
Like the chimera of myth — lion, goat, serpent — HARs make new developmental programs by recombining old parts. The proteins stay familiar. The wiring diagram does not.
2,600–3,100
HARs in major functional studies
96%
noncoding (Whalen 2023 set)
43%
opposing chromatin effects
HAR1
7.62chr20 · 20q13.33
The loudest human-specific stretch of DNA ever found — an RNA gene that turns on while the fetal cortex is being layered.
HACNS1
6.91chr2 · 2q14.2
A limb-and-hand volume knob. This enhancer is one of the best-tested examples of a human-specific developmental switch.
HAR2 / NPAS3
6.34chr14 · 14q13.1
NPAS3 hosts the largest HAR cluster in the genome — a transcription-factor locus that helps build and balance neural circuits.
HAR21
5.94chr14 · 14q13.1
A human-accelerated enhancer sitting inside NPAS3 itself — a switch embedded in the gene it helps control.
HARE5
6.05chr10 · 10p11.21
A brain-growth enhancer of FZD8. In mice, the human version makes a slightly larger, more proliferative cortex.
AUTS2 HARs
5.55chr7 · 7q11.22
AUTS2 is a neurodevelopmental gene with a HAR cluster. Protein sequence is conserved; the schedule of use is human-shifted.
Six-cluster operators
How the switches actually change
Acceleration is not one trick. Sometimes the volume goes up. Sometimes a letter is swapped to keep the circuit working. Sometimes the 3D fold of the chromosome moves a switch to a new gene.
01
Enhancer activity
ΔE
These stretches of DNA do not code for proteins. They act as volume knobs — turning nearby brain-development genes up or down at the right moment.
02
Compensatory substitution
C
When one letter in the sequence changes, another often changes to keep the whole switch working. Evolution rewired the circuit without breaking it.
03
Three-dimensional rewiring
T
The genome folds in 3D. Human-specific changes can re-route which switches sit next to which genes, even if the letters look similar in a straight line.
04
Copy-number expansion
D
Sometimes the switch is duplicated. Extra copies give evolution more material to experiment with — one copy keeps the old job, another can specialize.
05
Deletion and loss
L
Human evolution also deleted ancestral DNA. Removing an old brake can let a developmental program run longer or start earlier.
06
Population spread
P
Once a change is useful, it spreads. Many HAR variants rose quickly in ancestral humans and are now nearly fixed across living populations.
Codex
Operator pipeline
Select, shrink, project, measure. The lab lets you run that loop on the 125-mode carrier until only the keep-set glows.
Selector
K = (C − 6I)(C − 30I)
A filter that keeps two special families of patterns and rejects the rest.
Dissipation
A = K†K = K² ≥ 0
Turn the filter into a one-way shrink rule. Only the keep-set is untouched.
Step
Γ_ε = I − ε A, 0 < ε < 2 / λ_max(A)
Each round multiplies the leftovers by a number smaller than one.
Limit
lim Γ_εⁿ = P₄₇
After enough rounds you are looking only at the keep-set.
Invariant
Ω_c = dim E₄₇ / dim V = 47/125 = 0.376
The fraction that remains. On this carrier it is exact, not fitted.
Algebraic certificates: E47 reconstructed from spin-2 SU(2) (dim V = 125, dim ker K = 47, gap(G) = 11664). Genomic layer from Pollard 2006 through Mosti 2026. The HAR→QEGT map is a typed hypothesis. Ω_c is exact on the declared carrier, not a fitted biological threshold.