Interplay of Fidelity and Diversity in the Evolution of the Genetic Code

TL;DR

The paper tackles how the genetic code originated and why its mapping is so robust by treating code evolution as a multi-objective optimization balancing translation fidelity and amino acid diversity. It introduces a loss function $L = E + \eta D$, where $E$ captures mutation- and translation-error costs via polar-requirement distances and mutation rates, and $D$ enforces alignment with organismal amino-acid demand through a KL-divergence term between $f_\alpha$ and $p_\alpha$, with $\eta$ tuning their relative importance. Using a calibrated codon-mutation model and simulated annealing, the authors show the standard genetic code (SGC) sits near local optima and that the landscape contains rare, highly optimal codes across species; results indicate coevolution under conflicting pressures of fidelity and diversity. The work highlights that the current code is not only error-resilient but also tuned to reflect proteome composition, while acknowledging limitations like the non-evolutionary nature of simulated annealing trajectories and potential circularity in frequency estimates, suggesting avenues for experimental validation and broader evolutionary modeling.

Abstract

The origin and organizing principles of the genetic code remain fundamental puzzles in life science. The vanishingly low probability of the natural codon-to-amino acid mapping arising by chance has spurred the hypothesis that its structure is a solution optimized for robustness against mutations and translational errors. For the construction of effective molecular machines, the dictionary of encoded amino acids must also be diverse enough in physicochemical features. Here, we examine whether the standard genetic code can be understood as a near-optimal solution balancing these two objectives: minimizing error load and aligning codon assignments with the naturally occurring amino acid composition. Using simulated annealing, we explore this trade-off across a broad range of parameters. We find that the standard genetic code lies near local optima within the multidimensional parameter space. It is a highly effective solution that balances fidelity against resource availability constraints. These results suggest that the present genetic code reflects coevolution under conflicting pressures of fidelity and diversity, offering new insight into its emergence and evolution.

Figures (7)

• Figure 1: Codon table colored according to the polar requirement of the amino acids encoded by each codon, with blue indicating low values and red indicating high values. Neighboring codons tend to specify amino acids with similar polar requirement values, illustrating the nonrandom organization of the genetic code.
• Figure 2: Randomly generated genetic code in Fig \ref{['fig:FIG1']} format (left) and the heatmap representation of the error load of the corresponding local variation (right). In the heatmap, each column represents a codon subjected to the variation, while each row corresponds to an amino acid newly assigned due to deviations from the original codon assignment. Red cells indicate reduced error load (improvement), whereas blue cells indicate increased error load (deterioration). For instance, the amino acid Glu (glutamic acid), encoded by the codon CCC, exhibits a polar requirement significantly different from those of amino acids encoded by neighboring codons, especially for third position transitions; thus, most neighboring codes involving CCC show improvements. With 61 sense codons and 19 possible amino acid substitutions per codon, the total number of possible neighboring codes is $61\times19=1159$.
• Figure 3: The error load of local variation around the SGC as the third transition weight ($w_{\mathrm{tt}}$) increases. (a) Amino acid distance defined as absolute error. (b) Amino acid distance defined as squared error. The blue line represents the improvement rate, defined as the fraction of neighboring codes exhibiting a lower error load compared to the SGC. The red line indicates the maximum improvement in error load achieved among these codes for each value of $w_{\mathrm{tt}}$.
• Figure 4: The loss landscape of local variation of the SGC as a function of changes in the third transition weight, $w_{\mathrm{tt}}$, and the balancing parameter, $\eta$. (a) Amino acid distance defined as absolute error. (b) Amino acid distance defined as squared error. The color bar represents the number of codes with a lower loss value than the SGC, presented on a logarithmic scale, out of the 1159 possible neighboring codes described in Fig \ref{['fig:FIG2']}. Note that the interval $(0, 1)$ is omitted, and a value of 0 corresponds to the points where the SGC becomes a local optimum.
• Figure 5: (a)–(h) Results of optimization via simulated annealing for eight distinct values of $\tilde{\eta}$. Here, we define $\tilde{\eta}\equiv(\sigma_D/\sigma_E)\eta\approx 4\times10^7\eta$, where $\sigma_D$ and $\sigma_E$ are the standard deviations of the amino acid KLD and the error load, respectively, obtained from the ensemble of random codes. Using $\tilde{\eta}$ ensures that when $\eta$=1, the optimization assigns equal relative importance to minimizing both the error load and amino acid KLD. All points represent standardized z-scores of the error load ($E$, vertical axis) and amino acid KLD ($D$, horizontal axis). Each optimization trajectory corresponds to the average trajectory obtained from 100 simulated annealing runs, with the initial state being a randomly generated code. As in previous local variation experiments, optimization was conducted on 61 sense codons (excluding the stop codons) and based on the natural frequency of Homo sapiens. The optimization target (position of the SGC) is indicated by a red cross. (i) Final optimal codes obtained at the last iteration of simulated annealing for each of the eight $\tilde{\eta}$ values shown in panels (a)–(h). The result corresponding to the ideal value ($\tilde{\eta}=0.56$) is highlighted in blue.