Primordial Soup Simulations

• Primordial Soup Simulations are computational models that recreate prebiotic conditions to explore the emergence and evolution of self-replicating entities.
• They integrate digital evolution, particle aggregation, and kinetic Monte Carlo polymer dynamics to analyze fitness landscapes, neutral networks, and competitive selection.
• These models reveal how sparse replicators form mutationally connected clusters, demonstrating the roles of historical contingency and evolvability in early life.

Primordial soup simulations refer to computational models that investigate the emergence, evolution, and population dynamics of self-replicating entities in chemically or physically plausible prebiotic environments. These models are central to the study of the origins of life, as they enable systematic analysis of hypothetical prebiotic evolutionary processes under controlled, tunable conditions—ranging from abstract digital worlds to minimal particle and chemical systems. Synthesizing results from leading frameworks, including digital evolution, particle-based aggregation, and kinetic Monte Carlo polymer dynamics, these simulations reveal how replication-competent structures proliferate, compete, and diversify in the absence of modern cellular compartmentalization or metabolic networks.

1. Digital Microcosm Models of Self-Replicator Emergence

The Avida digital evolution platform provides a discrete, abstract representation of primordial soup scenarios in which the rules of chemical kinetics are replaced by instruction-level computation (G et al., 2017). Simulations are conducted in an $M \times M$ toroidal world with maximum population $N=10\,000$ and update-based time progression, where each genome ("avidian") is a sequence of $L$ instructions from an alphabet of 26 operations. Minimal self-replicators are systematically enumerated by exhaustively generating all $26^8 \approx 2.09 \times 10^{11}$ possible sequences of length 8 and testing their ability to execute a birth–replication–division cycle. Exactly 914 distinct minimal self-replicators exist at $L=8$; none occur at shorter length.

Mutational processes are introduced through point mutation ($\mu_{\text{point}}=7.5 \times 10^{-3}$ per copied instruction) and indel ($\mu_{\text{indel}}=5.0 \times 10^{-2}$ per division) rates. Fitness is defined as $f(g) = 1/T_{\rm rep}(g)$, where $T_{\rm rep}(g)$ is the time required for replication, with $f=0$ enforced for non-replicators.

2. Fitness Landscape, Neutral Networks, and Mutational Clusters

The fitness landscape of the digital primordial soup is represented as a genotype network $G = (V, E)$, where nodes represent the 914 minimal self-replicators and edges represent single-instruction Hamming distance neighbors. Network analysis yields:

• Cluster topology: 41 mutational clusters, 13 of which contain 14 or more genotypes; four largest clusters comprise 75% of self-replicators.
• Degree statistics: Degree distributions $P(k)$ within clusters are narrowly peaked; "fg"-type genotypes (fg/gb copy-loop motif) form denser, more connected clusters than "hc"-type replicators. "fg" and "hc" classes are mutationally disconnected.
• Clustering coefficient, mean path length, modularity: Quantified for each connected component, revealing high internal community structure.

Local evolvability $E_i$ is defined as the number of neighboring genotypes $j$ with $f(j) > f(i)$; fg-type replicators have significantly higher $E_i$, larger evolved genome sizes, and greater capacity for evolving new logic traits than hc-type, with statistical support from Wilcoxon rank-sum tests and multivariate regression ($R^2 \approx 0.6$, $p < 10^{-6}$) (G et al., 2017).

3. Competition, Takeover Probability, and Historical Contingency

Primordial-soup competition experiments initialize soups with equal numbers of all 914 self-replicators at low initial frequency and evolve them under standard mutational and environmental protocols. In 200 independent replicates, just three genotypes—vvwfgxgb, vwvfgxgb, and wvvfgxgb—emerge as progenitors in 65% of populations, with winners clustered mutationally around these peaks. The empirically measured takeover probability,

$P_i = \frac{W_i}{T}$

(where $W_i$ is the number of replicates won by genotype $i$ over $T=200$ runs), is highly non-uniform, indicating strong differentiation in competitive outcomes due to fitness peaks and network centrality.

If primordial replicators are exceedingly rare, the earliest to arise "locks in" subsequent evolutionary trajectories, demonstrating strong historical contingency. If origins are more frequent, competition reliably funnels ancestry through the most evolvable, high-fitness clusters—resulting in a quasi-deterministic selection of fittest origins (G et al., 2017).

4. Particle Models of Aggregate Emergence and Population Dynamics

Particle-based models, such as the Primordial Particle System (PPS), abstract prebiotic evolution to a system of $N$ self-propelled particles in continuous space, each with position $\mathbf{p}_i$, heading $\phi_i$, and constant speed $v$ (Schmickl et al., 2015). Interactions are governed by a deterministic motion law: upon each asynchronous update, particles count neighbors to their left/right within radius $r$, then turn by an angle $\Delta\phi$ given by

$$\frac{\Delta\phi_i}{\Delta t} = \alpha + \beta N_{t,r} \,\mathrm{sign}(R_{t,r} - L_{t,r})$$

followed by displacement at speed $v$. Above a critical particle density $\rho_c \approx 0.04$, spontaneous symmetry breaking leads to the emergence of two stable aggregate types: "spores" ($\sim$18 particles) and "cells" ($\sim$48 particles), each characterized by specific colors and local density.

The macroscopic growth of spore and cell aggregates is fit by top-down logistic ODEs:

$\dot X = a(1 - X/K)X$

with empirically determined coefficients ($a_{\text{cell}}=7.1\times10^{-4}$, $K_{\text{cell}}=50.78$; $a_{\text{spore}}=4.0 \times 10^{-4}$, $K_{\text{spore}}=18.21$), and Markov-chain models track the stochastic micro-dynamics of particle state transitions (color-coded by density) (Schmickl et al., 2015).

5. Polymerization, Environmental Cycling, and Functional Selection

Kinetic Monte Carlo models of primordial sequence pools simulate polymers constructed from "food" monomers (A, B), with periodic dehydration–rehydration cycles dictating alternations between phases of spontaneous assembly, template-directed replication, hydrolysis, diffusion, and functional catalysis (Walker et al., 2012). Each polymer $X_i$ is of fixed length $R_L = 20$ (10 A, 10 B), and a lattice of $64 \times 64$ sites hosts spatially resolved chemical kinetics:

• Dehydrated phase: No diffusion; spontaneous assembly ($A + B \xrightarrow{k_s} X_{\rm new}$) and universal sequence replication ($X_i + A + B \xrightarrow{k_r} 2X_i$).
• Hydrated phase: Monomer and polymer diffusion (rate constants $k_m$, $k_p$), reversible hydrolysis ($X_i \xrightarrow{k_h} 10A + 10B$), and, if present, catalysis by functional mutants (e.g., $$pA + \mathrm{Azyme} \xrightarrow{k_c} A + \mathrm{Azyme}$$).

Dynamic-kinetic steady states are achieved after 500–1000 cycles, with continuous sequence turnover, stable local diversity (Shannon entropy $\langle S_L \rangle \sim 2-3$), and clustering behavior regulated by diffusion rates and resource recycling. Introduction of functional sequences such as "Azyme" or "Bzyme" enables functional selection without erasing population-level diversity—functional lifetimes can be 5–60 times longer than random, and population sizes up to twice as large (Walker et al., 2012).

6. Synthesis, Structural Insights, and Limitations

Across models, several general features emerge:

• Self-replicators are rare in genotype space but form clustered, mutationally connected networks with sharply peaked connectivity.
• Competition among multiple spontaneously arising replicators leads to pronounced selection for clusters with high evolvability, fitness, and mutational robustness.
• Local resource recycling (in both particle and polymer models) and non-uniform diffusivities induce spontaneous clustering and coexistence, even in the absence of explicit compartmentalization or metabolic complexity.

Limitations are noted in each paradigm. Digital microcosms abstract away real chemistry, biochemistry, and energetics, relying on logical instruction sets. Particle systems lack explicit chemical kinetics. Polymerization models are idealized in monomer composition, environmental cycling, and neglect of side reactions or catalytic cross-talk.

Nonetheless, these primordial soup simulations quantitatively illuminate the interplay between historical contingency and selection, the formation of neutral networks, and the generic emergence of self-replicating clusters—features expected to characterize a wide array of prebiotic replicator systems (G et al., 2017, Schmickl et al., 2015, Walker et al., 2012).