Search bioRxiv⌕ Search
Home/Archive/bioRxiv

bioRxiv · 10.64898/2026.03.09.710612

Evolution as Active Geometry: The Geometric State Equation of the Tree of Life

Fenn, R.·Fenn, A.
View original source ↗Export metadata ↓

Abstract

Any process that generates information at a constant rate into a branching hierarchy must embed into hyperbolic space: exponentially growing lineages cannot pack into polynomial-growth Euclidean geometry. We derive a geometric state equation,{kappa} = (h ln 2/(n-1))2, relating the curvature{kappa} of the embedding manifold to the entropy rate h and dimension n, with zero adjustable parameters. Back-solving the dimension across every system tested--from decade-old viral outbreaks to 3.8-billion-year cellular lineages to domain-level species phylogenies--yields n = 2.00 {+/-} 0.05: evolution is two-dimensional. Curvature, by contrast, is scale-dependent. At the inter-domain scale, a neural encoder trained on 5,550 genomes with no phylogenetic supervision finds an optimal curvature range{kappa} {approx} 1.28-1.34, set by the Kolmogorov complexity profile of the biosphere; at the intra-domain scale, direct [H]2 embeddings of the complete GTDB bacterial (107,000 tips), archaeal (5,900 tips), and fungal (1,600 tips) species trees yield{kappa} = 3-16. Both regimes obey the state equation at the scale-appropriate entropy rate. Fifteen viral families trace the predicted curvature-entropy curve at Pearson r = 0.996; fifteen protein families confirm the predicted 3.1x curvature increase from a 4-letter to a 20-letter alphabet. The universal invariant is the dimension, not the curvature. The geometry of the tree of life is not a historical accident but a constraint imposed by the information capacity of the genetic code. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=174 HEIGHT=200 SRC="FIGDIR/small/710612v2_ufig1.gif" ALT="Figure 1"> View larger version (40K): org.highwire.dtl.DTLVardef@148fbf6org.highwire.dtl.DTLVardef@95f2e9org.highwire.dtl.DTLVardef@7ea052org.highwire.dtl.DTLVardef@14f8999_HPS_FORMAT_FIGEXP M_FIG C_FIG The tree of life embeds into 2D hyperbolic space with curvature determined by the geometric state equation{kappa} = (h ln 2)2. Top: Voronoi tessellation of 5,550 genome embeddings in the Poincare disk, colored by domain (Bacteria, Archaea, Eukarya). LUCA occupies the center; cell boundaries are hyperbolic geodesics (circular arcs orthogonal to the disk boundary). Bottom: The state equation predicts curvature from entropy alone across a 13-fold range--from the compressed inter-domain hierarchy ({kappa} {approx} 1.3) through domain-level species trees ({kappa} = 3-16)--and across both DNA and protein alphabets (3.1 x increase), with zero adjustable parameters. Fifteen viral families confirm the curvature-entropy curve at r = 0.996.

Explore related subjects

evolutionary biologyPreprint

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Fenn, R., Fenn, A.. 2026-03-13. Evolution as Active Geometry: The Geometric State Equation of the Tree of Life. https://doi.org/10.64898/2026.03.09.710612

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related preprints

bioRxiv2026

RELAX does not reproduce its own estimates at default settings, and its output does not show it

Selection-intensity estimates from RELAX are reported as a point value of K with a likelihood-ratio P. We report that, at default settings and on data of ordinary size, the program does not reproduce its own fits. Of 27 enzyme entries refitted under two optimiser configurations, none reproduced its log-likelihood to within 0.01 units; the median change was 103 units, the largest over 3,400, and four verdicts reversed. Eighty null orthologues reproduced none. A byte-identical command returned a distinct likelihood on every repetition, single-threaded, across three releases, and on alignments simulated under the fitted model, where 3.3 per cent of replicates reproduced. The documented random-number seed never reaches the generator when assigned on the command line, yet reads back as the value supplied. PAML localises the cause: its two-ratio model, without site classes, reproduced its log-likelihood for all 288 genes; its site-class models agreed for 27 to 67 per cent. The instability follows the mixture over sites, not the program. The output does not show it: 46 of 410 fits ended with a negative likelihood-ratio statistic, impossible under convergence, and 123 of 410 report a K re-estimated under a domain restriction rather than the unconstrained maximum. Of 234 published studies using RELAX, none reported a seed. Seeding while holding the thread count at one reproduced sixty of sixty runs on twenty genes under two releases; the seed alone reproduced none of five, and no documentation states the second condition. We recommend that fits be repeated and their dispersion published.

evolutionary biology↗
bioRxiv2026

Sequential accumulation of adaptive alleles forms an inversion supergene in deer mice

Supergenes are clusters of co-inherited loci that affect multiple or complex phenotypes. Despite the growing number of chromosomal inversions identified as supergenes in natural populations, their molecular basis and evolutionary history often remain obscure. Here, we identified two candidate genes, Slc45a2 and Npr3, within a 41-Mb inversion supergene in the deer mouse (Peromyscus maniculatus) that respectively drive darker coats and longer tails - two traits associated with forest adaptation. Mice homozygous for the inversion (inv/inv) exhibit elevated Slc45a2 expression in melanocytes relative to the congenic standard genotype (std/std), disrupting pheomelanin production. In parallel, downregulation of Npr3 in inv/inv mouse growth plates prolongs postnatal growth of caudal vertebrae, resulting in tail elongation. Population-level analyses further implicate that this supergene arose through the subsequent accumulation of the Npr3 allele within the inversion, rather than by capturing all beneficial mutations at its origin.

evolutionary biology↗
bioRxiv2026

Toxin structure shapes palatability in a chemically defended butterfly

The toxicity of chemical defences is well studied, but the potential contribution of compound structure to predator deterrence remains largely unexplored. Whether predation acts more strongly on toxicity or unpalatability remains largely untested, partly because few systems allow toxin structure to vary independently of quantity. Heliconius sara larvae provide such a system: those reared on Passiflora auriculata sequester cyclopentenyl cyanogenic glucosides (CGs), while those reared on P. biflora biosynthesise comparable quantities of aliphatic CGs. Using two invertebrate predators, Camponotus floridanus ants and Hierodula membranacea mantids, we tested whether this structural difference affects palatability independent of toxicity. Mantids rejected larvae with cyclopentenyl CGs more often than larvae with aliphatic CGs, despite no detectable difference in total CG content. This pattern was mirrored in extract-based assays with ants, independently of cyanide release: extracts with cyclopentenyl CGs remained deterrent, while extracts with aliphatic CGs did not differ in deterrence from water. Live larvae, by contrast, elicited similar responses from ants regardless of CG structure. These results show that variation in toxin structure can strongly affect palatability, with some compounds conferring greater protection than others. This demonstrates the importance of chemical structural diversity in the evolution of chemical defences.

evolutionary biology↗