Skip to content

science.md

Latest commit

 

History

History
619 lines (512 loc) · 31.6 KB

science.md

File metadata and controls

619 lines (512 loc) · 31.6 KB
Raw

The science behind cellauto

cellauto is structured around a five-stage approximation of abiogenesis — the chemistry-to-life transition. The original v1.0 README called it "natural selection," but read carefully, the four rules sketch the prebiotic-chemistry chapter of the origin-of-life story: random mixing, condensation, activated intermediates, compartmentalization. v3.0 honors that intuition by implementing each stage with real (or toy-but-real-concept) scientific machinery and citing the canonical literature.

This document is the academic appendix: what each stage means, where the math comes from, and what corners are cut.

Stage 0 — Primordial soup

"Each cell takes on a random color [from its surroundings]." — original Rule 1

Discrete molecular species mixing by random Brownian-style neighbor sampling. Same-species cells in proximity condense; freshly-formed species are reactive (is_new); a successful condensation yields a primitive protocell (amoeba) that persists for a fixed lifetime, then dissolves.

The honest science here is the primordial soup hypothesis — first articulated by Alexander Oparin (1924, The Origin of Life) and J. B. S. Haldane (1929, Rationalist Annual). Both independently argued that early Earth's oceans contained dissolved organic monomers (amino acids, simple sugars, nucleobases) that mixed and reacted to produce more complex molecules. Stanley Miller and Harold Urey's famous 1953 experiment demonstrated that lightning-driven discharges in a reducing atmosphere produce amino acids, validating the soup framework.

Real data: the 16 molecular "species" are not abstract — they are the 16 most abundant products Miller actually recovered in 1953, and the initial soup is sampled weighted by his reported yields (formic acid ≈ 49%, glycine ≈ 13%, glycolic acid ≈ 12%, alanine ≈ 7%, with everything else in the tail). So the starting grid reflects the measured composition of a spark-discharge soup rather than a uniform rainbow — a real soup is dominated by a few simple molecules. (See MILLER_UREY_SPECIES in natural_selection.py.)

What this implementation captures: mass-action mixing, condensation of like monomers, kinetic activation lifetime. What it doesn't: real thermodynamics, real reaction kinetics, the specific organic-chemistry pathways.

Citations

  • Oparin, A. I. (1924). Происхождение жизни (The Origin of Life). Moscow.
  • Haldane, J. B. S. (1929). The Origin of Life. Rationalist Annual, 148, 3–10.
  • Miller, S. L. (1953). A production of amino acids under possible primitive Earth conditions. Science, 117(3046), 528–529. doi:10.1126/science.117.3046.528

Stage 1 — Reaction-diffusion

A two-species reaction-diffusion PDE on a continuous concentration field. The mathematics:

∂u/∂t  =  D_u ∇²u  -  u v²  +  F (1 - u)
∂v/∂t  =  D_v ∇²v  +  u v²  -  (F + k) v

Two reactants u and v diffuse with coefficients D_u, D_v. The non-linear term u v² represents an autocatalytic reaction (one molecule of v catalyses its own production from u). F is the feed rate of u; F+k is the kill rate of v. Despite having only four parameters, this system produces a remarkable parameter landscape — Pearson (1993) catalogued five qualitatively different regimes:

Regime Behaviour (F, k) example
Spots Self-replicating spots, mitosis-like division (0.035, 0.065)
Stripes Long meandering stripes (0.04, 0.06)
Mitosis Cells split and migrate (0.0367, 0.0649)
Waves Travelling wavefronts (0.014, 0.045)
Labyrinth Maze-like patterns (0.039, 0.058)

Numerics. We integrate with forward Euler on a unit grid (dx = 1, dt = 1) and a 5-point Laplacian — the standard non-dimensional "xmorphia" parameterization with Du = 0.16, Dv = 0.08. These are rescaled values, not Pearson's original Du = 2×10⁻⁵, Dv = 10⁻⁵ on the unit square; space and time are non-dimensionalized, so the patterns are faithful while the raw constants differ. The explicit-diffusion stability bound for a 2D 5-point stencil is D·dt/dx² ≤ 1/4; here the largest term is 0.16 ≤ 0.25, so the scheme is stable with margin. (The integrator clips u, v to [0, 1], which keeps a demo well-behaved but would also mask a true instability if you pushed dt past the bound.)

The Du : Dv = 2 : 1 ratio is a modelling choice, not a measured chemical ratio: Turing patterns require the inhibitor to diffuse faster than the activator, and a ratio around 2 reliably lands in the pattern-forming regime. Real small-molecule aqueous diffusion coefficients are all of order 10⁻⁹ m²/s (glycine D ≈ 1.06×10⁻⁹ m²/s; most metabolites cluster in 0.5–1×10⁻⁹), differing by far less than 2× — so the qualitative requirement "inhibitor diffuses faster" is the physically meaningful part, while the exact factor is tuned for the demonstration.

Reaction-diffusion is the mechanism Alan Turing proposed in 1952 to explain biological pattern formation (zebra stripes, leopard spots, fingerprint ridges). The same math drives the BZ chemical oscillator, the Liesegang ring phenomenon, and the spatial dynamics of many real catalytic systems.

We include it as Stage 1 because self-replicating spots in Gray-Scott visually resemble protocell division and constitute the simplest dynamical system the project ships that produces emergent self-organizing structure.

Citations

  • Turing, A. M. (1952). The chemical basis of morphogenesis. Phil. Trans. Royal Society B, 237(641), 37–72. doi:10.1098/rstb.1952.0012
  • Gray, P., & Scott, S. K. (1985). Sustained oscillations and other exotic patterns of behaviour in isothermal reactions. J. Phys. Chem., 89(1), 22–32.
  • Pearson, J. E. (1993). Complex patterns in a simple system. Science, 261(5118), 189–192. doi:10.1126/science.261.5118.189

Stage 2 — Autocatalytic sets (Kauffman RAFs)

Stuart Kauffman's 1986 insight: above a connectivity threshold, random reaction networks spontaneously contain closed catalytic cycles. A small random sample of plausibly-pre-biotic reactions, with each reaction catalysed by some other species, will — with high probability — include a subset whose products collectively catalyse each other in a closed loop.

Kauffman's argument was qualitative until Hordijk & Steel (2004) gave it rigorous form: the RAF algorithm finds the maximal Reflexively Autocatalytic Food-generated set of reactions in a network. RAF detection is now the canonical way to formalize "Kauffman-style autocatalysis":

  1. Start with the full reaction set.
  2. Compute which species are producible from the food set ∪ current reaction products.
  3. Prune reactions whose reactants OR catalyst are not producible.
  4. Repeat until stable. What remains is the maximal RAF.

A non-empty RAF means: given the food, the network can sustain itself indefinitely; every reaction can keep firing because its prerequisites are always being regenerated by other reactions in the same set.

In this stage we generate a random reaction network, find its RAF, then simulate the chemistry on a 2D grid with diffusion. The RAF reactions fire whenever their reactants are colocated, producing visible autocatalytic "ignition" — bright regions where the closed loop is amplifying.

Citations

  • Kauffman, S. A. (1986). Autocatalytic sets of proteins. J. Theor. Biol., 119(1), 1–24. doi:10.1016/S0022-5193(86)80047-9
  • Kauffman, S. A. (1993). The Origins of Order: Self-Organization and Selection in Evolution. Oxford University Press.
  • Hordijk, W., & Steel, M. (2004). Detecting autocatalytic, self-sustaining sets in chemical reaction systems. J. Theor. Biol., 227(4), 451–461. doi:10.1016/j.jtbi.2003.11.020
  • Hordijk, W., Steel, M., & Kauffman, S. A. (2012). The structure of autocatalytic sets: evolvability, enablement, and emergence. Acta Biotheoretica, 60(4), 379–392.

Stage 3 — Vesicle formation (lipid self-assembly)

Chemistry compartmentalizes. Amphiphilic molecules (lipids with a hydrophilic head and hydrophobic tail) spontaneously self-assemble into bilayer membranes above their critical micelle concentration (CMC). The first stable membrane around a patch of self-sustaining chemistry is a protocell — the moment the universe acquires bounded chemical individuals capable of having their own internal state.

This stage runs a Gray-Scott-like reaction-diffusion for a lipid-precursor species, then marks regions where lipid concentration exceeds the CMC as membrane. Connected high-lipid regions become discrete vesicles tracked by flood-fill connected-component labelling.

Real data: the membrane is identified with a named fatty acid whose measured critical micelle concentration anchors the threshold (decanoic acid (C10) ≈ 85 mM, octanoic (C8) ≈ 250 mM, oleic (C18:1) ≈ 0.1 mM; see AMPHIPHILE_CMC_MM). C8–C10 monocarboxylic acids are the prebiotic sweet spot — the species Deamer extracted from the Murchison meteorite that form vesicles under early-Earth conditions. The simulation field is normalized so 1.0 corresponds to the chosen amphiphile's CMC.

What this captures: threshold-based self-assembly, vesicle counting. What it cuts: real lipid bilayer dynamics involve Helfrich curvature elasticity, surface tension, and fluid mechanics (Lipowsky & Sackmann 1995). Real vesicle growth/division has been studied experimentally by Hanczyc & Szostak (2004). A faithful implementation would be a separate research project.

Citations

  • Helfrich, W. (1973). Elastic properties of lipid bilayers: theory and possible experiments. Z. Naturforsch. C, 28(11), 693–703.
  • Deamer, D. W. (2008). Origins of life: How leaky were primitive cells? Nature, 454(7200), 37–38. doi:10.1038/454037a
  • Szostak, J. W., Bartel, D. P., & Luisi, P. L. (2001). Synthesizing life. Nature, 409(6818), 387–390.
  • Hanczyc, M. M., & Szostak, J. W. (2004). Replicating vesicles as models of primitive cell growth and division. Curr. Opin. Chem. Biol., 8(6), 660–664. doi:10.1016/j.cbpa.2004.10.002
  • Lipowsky, R., & Sackmann, E. (1995). Structure and Dynamics of Membranes. North-Holland.

Stage 4 — Protocell selection (hypercycle dynamics)

Once chemistry is compartmentalized, selection becomes possible. Protocells with efficient internal chemistry grow; those without dissolve. Growing protocells eventually divide, and the daughter inherits the parent's internal state with stochastic mutation. This is the moment Darwinian dynamics appear — not as an imposed law, but as a consequence of bounded chemical individuality with heritable variation and differential survival.

The canonical theoretical framework is Manfred Eigen and Peter Schuster's hypercycle (1977): a closed loop of replicators where each member catalyses the formation of the next. Hypercycles are evolutionarily stable in a way isolated self-replicators are not — they sidestep Eigen's "error catastrophe" by distributing information across multiple replicating species.

This stage's implementation runs the genuine Eigen-Schuster replicator dynamics. Each protocell carries a vector "genome" of internal species concentrations that evolve by the replicator ODE dxᵢ/dt = xᵢ(kᵢ·xᵢ₋₁ − Φ); a protocell grows only while its hypercycle is complete (gated by its limiting species — the weakest member of the loop), divides at a radius threshold into a mutated daughter, and old or small protocells die. A legacy scalar "proxy" fitness remains available as an A/B reference behind dynamics="proxy". Real protocell evolution involves internal RAF dynamics and stochastic mutation rates that determine the error threshold (see Adamala & Szostak's 2013 experimental work).

We include this stage to make explicit where the project's original claim of "natural selection" actually lives: not in the soup, but in the chemistry-bearing membranes that arise after compartmentalization. Once there, the four-pillar Darwinian requirements — variation, inheritance, selection, replication — are all present.

Citations

  • Eigen, M. (1971). Selforganization of matter and the evolution of biological macromolecules. Naturwissenschaften, 58(10), 465–523.
  • Eigen, M., & Schuster, P. (1977). The hypercycle: a principle of natural self-organization. Part A: emergence of the hypercycle. Naturwissenschaften, 64(11), 541–565. doi:10.1007/BF00450633
  • Eigen, M., & Schuster, P. (1978). The hypercycle. Part B: the abstract hypercycle. Naturwissenschaften, 65(1), 7–41.
  • Adamala, K., & Szostak, J. W. (2013). Nonenzymatic template-directed RNA synthesis inside model protocells. Science, 342(6162), 1098–1100.
  • Szostak, J. W. (2017). The narrow road to the deep past: in search of the chemistry of the origin of life. Angew. Chem., 56(37), 11037–11043.

RNA world — spatial quasispecies and the error catastrophe

Selectable directly as the abiogenesis-rna-world rule (not yet woven into the auto-promoting pipeline). Walter Gilbert's 1986 RNA world hypothesis is the dominant modern picture of early life: RNA served as both genotype (a copyable template) and catalyst (a ribozyme) before the DNA/protein division of labour. The quantitative law that governs any such self-replicator is Eigen's quasispecies theory.

For a single-peak fitness landscape — a "master" sequence of length L that replicates with superiority σ, every mutant replicating at rate 1 — the master is maintained only while the per-base copy-error rate stays below the error threshold:

ε_c  =  ln(σ) / L

Below ε_c the population is a quasispecies: a mutant cloud centred on the master. Above ε_c the master is lost and the population melts into random sequences — the error catastrophe. This stage is a spatial Eigen model: each cell holds an RNA strand over a 4-letter alphabet (or is empty); empty cells are colonised by a fitness-weighted occupied neighbour (selection), and the copy is made base-by-base with per-base error ε (mutation); occupied cells die at a fixed rate. Cells are coloured by Hamming distance to the master (bright = master, dark = far mutant). Dragging the error-rate slider past ε_c ≈ ln(10)/16 ≈ 0.14 reproduces the catastrophe live.

What this captures: selection on a single-peak landscape, per-base mutation, and the Eigen error threshold as an observable phase transition. What it cuts: real base-pairing/templated copying chemistry, ribozyme folding, and sequence-dependent fitness landscapes.

Citations

  • Gilbert, W. (1986). The RNA World. Nature, 319, 618.
  • Eigen, M. (1971). Selforganization of matter… Naturwissenschaften, 58(10), 465–523.
  • Spiegelman, S. (1971). An approach to the experimental analysis of precellular evolution. Q. Rev. Biophys., 4(2–3), 213–253.
  • Joyce, G. F. (2002). The antiquity of RNA-based evolution. Nature, 418(6894), 214–221.

Origin of the genetic code — coevolution of message and code

Selectable as the abiogenesis-genetic-code rule. The deepest unsolved problem at the chemistry-to-biology boundary is the origin of the genetic code itself: why is the codon→amino-acid mapping (nearly) universal across all life, and how did it arise? Three classic answers:

  • Stereochemical (Woese 1965): codons and their amino acids are physically matched — the code reflects molecular complementarity.
  • Coevolution (Wong 1975; Vetsigian, Woese & Goldenfeld 2006): the code and the messages it interprets coevolve, and innovation-sharing among proto-organisms drives the population to converge on a single shared code.
  • Frozen accident (Crick 1968): whatever code happened first locked in.

This stage models the coevolution account. Each cell on the grid carries both an RNA-like strand of codons and its own private codon→amino-acid table. Each cell decodes its own strand through its own code to produce a peptide; fitness is how well that peptide matches a fixed target catalyst. Empty cells are colonised by fitness-weighted occupied neighbours, copying both the strand (with per-base mutation) and the code (with rare swaps). Any code that happens to make a more useful peptide spreads. The code_consensus population stat tracks how much the surviving population agrees on the code: convergence toward 1.0 is the emergence of a universal genetic code.

What this captures: the coevolutionary dynamics of message and code, and selection acting on the code itself — the mechanism behind the universality of the genetic code (Vetsigian-Woese-Goldenfeld 2006). What it cuts: actual stereochemistry, actual translation by ribosomes, and the specific historical contingencies that shaped the canonical code.

Citations

  • Crick, F. H. C. (1968). The origin of the genetic code. J. Mol. Biol., 38(3), 367–379.
  • Woese, C. R. (1965). On the evolution of the genetic code. PNAS, 54, 1546–1552.
  • Wong, J. T.-F. (1975). A co-evolution theory of the genetic code. PNAS, 72(5), 1909–1912.
  • Vetsigian, K., Woese, C., & Goldenfeld, N. (2006). Collective evolution and the genetic code. PNAS, 103(28), 10696–10701.
  • Koonin, E. V. (2017). Frozen accident pushing 50… Life, 7(2), 22.

Homochirality — spontaneous mirror-symmetry breaking

Selectable as the abiogenesis-homochirality rule. Life is homochiral: it uses only L-amino acids and D-sugars, never their mirror images. A racemic prebiotic soup has equal amounts of each enantiomer, so something broke the mirror symmetry and amplified one hand to exclusivity. F. C. Frank's 1953 model shows the mechanism: autocatalysis (each enantiomer catalyses its own formation) plus mutual antagonism (opposite hands annihilate) makes the racemic state unstable — any tiny fluctuation is amplified until one hand wins.

A + L → 2L      (autocatalysis,    rate k_a)
A + R → 2R      (autocatalysis,    rate k_a)
L + R → inert   (mutual antagonism, rate k_x)

On the 2D reaction-diffusion field, local patches spontaneously break to opposite handedness, forming chiral domains (teal = L-dominant, magenta = R-dominant) that then compete. Turning the antagonism rate k_x toward zero restores a stable racemic state — no symmetry breaking. The Soai reaction (1995) is the laboratory realisation of asymmetric autocatalysis amplifying a tiny enantiomeric excess to near-homochirality.

What this captures: autocatalysis + mutual antagonism, spontaneous symmetry breaking, and spatial chiral-domain formation/competition. What it cuts: the actual chemistry of a specific autocatalyst, and the proposed sources of the initial bias (Kondepudi & Nelson 1985 on parity violation; circularly polarised light; mineral surfaces).

Citations

  • Frank, F. C. (1953). On spontaneous asymmetric synthesis. Biochim. Biophys. Acta, 11, 459–463.
  • Soai, K., et al. (1995). Asymmetric autocatalysis and amplification of enantiomeric excess. Nature, 378, 767–768.
  • Blackmond, D. G. (2004). Asymmetric autocatalysis and its implications for the origin of homochirality. PNAS, 101(16), 5732–5736.
  • Kondepudi, D. K., & Nelson, G. W. (1985). Weak neutral currents and the origin of biomolecular chirality. Nature, 314, 438–441.

Alkaline hydrothermal vents — a proton gradient does the work

Selectable as the abiogenesis-hydrothermal-vent rule. This is the metabolism-first alternative to the lightning-powered soup. Serpentinisation of ocean crust produces warm, alkaline (pH ~9–11), H₂-rich vent fluid; the early ocean was mildly acidic (CO₂-rich, pH ~5–7). Where the two meet across the thin catalytic (FeS) walls of a vent chimney there is a natural proton gradient of ~3–4 pH units — a built-in proton-motive force, the very same kind of gradient every living cell still uses to make ATP (chemiosmosis).

Lane & Martin (2012) argue that this geochemical gradient, not a hand-set feed rate, is the free-energy source for the first carbon fixation. The stage models exactly that: the chimney interior is held alkaline and the ocean edges acidic (Dirichlet sources), a steady gradient forms between them, and organic matter is synthesised in proportion to the steepness of the local gradient — so synthesis ignites along the chimney wall (the interface), not uniformly. Flatten the gradient (vent pH = ocean pH) and synthesis stops entirely: no gradient, no free energy, no chemistry. Blue = alkaline, orange = acidic, teal-green glow = organic synthesis.

What this captures: a fixed geochemical proton gradient as the energy source, and interface-localised synthesis driven by the proton-motive force. The stage exposes real thermodynamic readouts: the abstract proton field maps to actual pH (alkaline ≈ 10, ocean ≈ 5.5 by default — the early-Earth ocean estimate of Krissansen-Totton et al. 2018); at 25 °C the Nernst factor 2.303 RT/F ≈ 59.16 mV per pH unit gives the proton-motive force directly (PMF ≈ 266 mV at default ΔpH = 4.5), and Faraday's constant converts that to the available free energy (ΔG ≈ −26 kJ/mol per proton). Those numbers sit comfortably above ATP synthase's ~150 mV threshold and in the range Lane & Martin argue can drive abiotic carbon fixation.

The chemistry the gradient drives is the Wood-Ljungdahl pathway. The stage tracks dissolved H₂ (replenished inside the alkaline chimney by serpentinisation) and dissolved CO₂ (globally fed from the CO₂-rich Hadean ocean), and the synthesis term is a proper mass-action reaction with the real net stoichiometry

2 CO₂ + 4 H₂ → CH₃COOH + 2 H₂O      (ΔG° ≈ −95 kJ/mol)

Rate ∝ PMF · [H₂] · [CO₂], capped by the 2 : 1 H₂ : CO₂ requirement (the limiting reagent is enforced). Take either feedstock away and the reaction stops even though the proton gradient is still there — exactly the Russell-Martin-Sojo argument that the alkaline-vent geochemistry sets up not just a free-energy source but the actual carbon-fixation chemistry that became central metabolism. What it cuts: the actual carbon-fixation chemistry (acetyl-CoA / Wood- Ljungdahl pathway), real FeS/FeNi mineral catalysis, and fluid flow.

Citations

  • Russell, M. J., & Hall, A. J. (1997). The emergence of life from iron monosulphide bubbles… J. Geol. Soc., 154(3), 377–402.
  • Martin, W., & Russell, M. J. (2007). On the origin of biochemistry at an alkaline hydrothermal vent. Phil. Trans. R. Soc. B, 362, 1887–1925.
  • Lane, N., & Martin, W. F. (2012). The origin of membrane bioenergetics. Cell, 151(7), 1406–1416.
  • Sojo, V., et al. (2016). The origin of life in alkaline hydrothermal vents. Astrobiology, 16(2), 181–197.

Coacervates — membraneless compartments by phase separation

Selectable as the abiogenesis-coacervate rule. This is the original protocell idea — Alexander Oparin's 1924 coacervates: dense, membraneless droplets that form when macromolecules spontaneously separate from solution. It's a different answer to "how did chemistry get a boundary?" than Stage 3's lipid vesicles: no membrane at all, just liquid-liquid phase separation. The same physics underlies modern biomolecular condensates (membraneless organelles), which has revived coacervates as a serious origin-of-life model.

The dynamics are the Cahn-Hilliard equation — phase separation with a conserved order parameter φ (local composition):

μ      =  φ³ − φ − κ ∇²φ      (double-well chemistry + interface energy)
∂φ/∂t  =  M ∇²μ               (conserved: total φ is preserved)

From a near-uniform, slightly off-critical mixture, φ separates into a coacervate-rich phase (gold droplets) and a dilute phase (dark), then coarsens: small droplets dissolve to feed larger ones (Ostwald ripening) and neighbours fuse, so the droplet count peaks and then declines — exactly as real coacervates do. κ sets the line tension (bigger → fewer, larger droplets) and the mean composition controls how much rich phase forms.

What this captures: conserved liquid-liquid phase separation, droplet nucleation, and coarsening. What it cuts: the specific macromolecules (polypeptide/polynucleotide complex coacervation), electrostatics, and selective partitioning of solutes.

Citations

  • Oparin, A. I. (1924). The Origin of Life. (Coacervate hypothesis.)
  • Bungenberg de Jong, H. G. (1932). Coacervation. (original coacervate work)
  • Cahn, J. W., & Hilliard, J. E. (1958). Free energy of a nonuniform system. J. Chem. Phys., 28(2), 258–267.
  • Banani, S. F., et al. (2017). Biomolecular condensates. Nat. Rev. Mol. Cell Biol., 18, 285–298.

Mineral-surface catalysis — the first polymers form on clay

Selectable as the abiogenesis-mineral-catalysis rule. Condensation polymerisation (joining monomers into chains, releasing water) is uphill in bulk water, so dilute monomers don't spontaneously form long polymers in the open ocean. Mineral surfaces fix this. James Ferris showed montmorillonite clay catalyses RNA-nucleotide polymerisation into 30–50-unit chains by concentrating monomers on its charged surface and templating bond formation; A. G. Cairns-Smith proposed clay crystals as the first "genetic" material.

The stage models surface catalysis directly: a static clay mask sits on the grid; monomers diffuse and are fed; polymer forms at a rate that is high on the clay and near-zero off it; polymer slowly hydrolyses everywhere. Long polymer accumulates on the clay patches (teal-green) while the bulk (dark) stays monomeric — the chemistry is localised to the mineral surface. Raise the bulk-water rate to equal the clay rate and the localisation vanishes.

What this captures: surface-localised polymerisation, the bulk-water vs mineral-surface contrast, and slow hydrolysis. What it cuts: the real templating geometry of the clay interlayer, monomer activation chemistry, and sequence selectivity.

Citations

  • Ferris, J. P., Hill, A. R., Liu, R., & Orgel, L. E. (1996). Synthesis of long prebiotic oligomers on mineral surfaces. Nature, 381, 59–61.
  • Cairns-Smith, A. G. (1982). Genetic Takeover and the Mineral Origins of Life. Cambridge University Press.
  • Hanczyc, M. M., Fujikawa, S. M., & Szostak, J. W. (2003). Experimental models of primitive cellular compartments. Science, 302, 618–622.
  • Hazen, R. M., & Sverjensky, D. A. (2010). Mineral surfaces… and the origins of life. CSH Perspect. Biol., 2, a002162.

LUCA distillation — the last universal common ancestor

Selectable as the abiogenesis-luca rule. Every organism alive today descends from a single inferred ancestor, the Last Universal Common Ancestor (LUCA). We cannot dig it up; we reconstruct its genome by comparative genomics — taking the (threshold-relaxed) intersection of gene families across all sequenced lineages. Weiss et al. (2016) ran this analysis over ~6.1 million prokaryotic protein-coding genes and recovered a core gene set (~355 protein families) consistent with LUCA being a hydrothermal, hydrogen-using chemolithoautotroph.

This stage models that distillation. Each cell carries a gene-presence bitset; some genes are essential (high fitness), some accessory (mild), some deleterious (cost), and every gene has a maintenance cost. Selection + mutation drive the population; the headline luca_size stat is the number of genes present in ≥ 70% of surviving lineages — exactly the kind of prevalence threshold real LUCA reconstruction uses. From random initial genomes, luca_size climbs and locks at roughly the essential-gene count (12 by default). That intersection IS the simulated LUCA, the genome every lineage inherited.

What this captures: comparative-genomics distillation of a core ancestral genome under selection, the trade-off between gene benefit and genome-size cost, and the gene-prevalence threshold real reconstruction relies on. What it cuts: actual sequence evolution, the specific metabolic genes of real LUCA, horizontal gene transfer (the abiogenesis-genetic-code stage covers that part), and protein structure.

Citations

  • Koonin, E. V. (2003). Comparative genomics, minimal gene-sets and the last universal common ancestor. Nat. Rev. Microbiol., 1, 127–136.
  • Theobald, D. L. (2010). A formal test of the theory of universal common ancestry. Nature, 465, 219–222.
  • Weiss, M. C., et al. (2016). The physiology and habitat of the last universal common ancestor. Nat. Microbiol., 1, 16116.
  • Mirkin, B. G., et al. (2003). Parsimonious evolutionary scenarios for genome evolution. BMC Evol. Biol., 3, 2.

How to read the pipeline

The abiogenesis-pipeline rule runs the five stages in sequence with auto-transition thresholds, so a single run walks the entire chemistry-to- life narrative end to end. The transitions are heuristic — chosen so the demo holds together visually, not to match any specific origin-of-life timescale. The real Earth took ~500 million years to walk this trajectory.

You can also run any stage in isolation with cellauto gui --rule abiogenesis-stageN-... or skip ahead via the GUI's Promote stage button.

SEM mode is a rendering choice, not new physics

v4.0 ships a depth-shaded "scanning-electron-microscope" rendering path (cellauto/renderer_sem.py) alongside the v3.6 viridis renderer. The simulation maths, constants, dynamics, RNG handling, and snapshot format are unchanged from v3.6. Toggling View ▸ SEM mode does not introduce new physics, new species, new reactions, or new free parameters. The same seed under SEM mode and under viridis produces the same step count, the same population stats, and the same internal state.

What SEM mode does is reinterpret the existing per-stage render_rgb(state) output as a height field, then apply Lambertian + ambient + specular shading, a value-noise micro-texture, a sepia or mono LUT, and a LANCZOS upscale, before overlaying the LIVE SEM FEED badge, crosshair, and scale bar. Every SEM pixel still traces back to a real engine value: the height-from-luminance step is deterministic and invertible up to the LUT.

The depth shading is therefore interpretive, not measurement. The scale bar is a visual reference for the grid extent, not a calibrated micron readout. The "LIVE SEM FEED" badge is a framing device borrowed from instrument UIs; it does not imply that the simulation is being run against electron-microscope data. Readers familiar with real SEM micrographs should treat the v4.0 rendering as a stylised presentation of the v3.6 dynamics, not as a replacement instrument output.

This framing is deliberate. The v3.x line earned scientific credibility through coupled dynamics (G1), the real Eigen-Schuster ODE (G2), Helfrich bending (G3), the Miyazawa-Jernigan landscape (G4), and the pathway-graph LUCA distillation (G5). The v4.0 line is about scientific representation on top of that engine — making the chemistry-to-life story read as a microscope view rather than a heat-map. The two are independent. If you want the raw scientific signal without the shading, View ▸ SEM mode turns it off and the v3.6 viridis path renders the same state.

Honest limitations

This is a sandbox for thinking about abiogenesis, not a quantitative model.

  • Toy time/length scales. Steps are not seconds; cells are not nm or μm.
  • No real thermodynamics. Free energies, entropy of mixing, Gibbs energies of reaction are not modelled.
  • No real reaction kinetics. Rates are mass-action with phenomenological constants, not Arrhenius / transition-state theory.
  • Toy membrane physics. Stage 3 uses threshold detection rather than fluid simulation of lipid bilayers.
  • Toy fitness function. Stage 4 uses Shannon entropy × concentration as a placeholder; real protocell fitness depends on internal RAF dynamics.

For each, the reference list above points at the rigorous treatment.