Excess mutual catalysis is required for effective evolvability.

It is widely accepted that autocatalysis constitutes a crucial facet of effective replication and evolution (e.g., in Eigen's hypercycle model). Other models for early evolution (e.g., by Dyson, Gánti, Varela, and Kauffman) invoke catalytic networks, where cross-catalysis is more apparent. A key question is how the balance between auto- (self-) and cross- (mutual) catalysis shapes the behavior of model evolving systems. This is investigated using the graded autocatalysis replication domain (GARD) model, previously shown to capture essential features of reproduction, mutation, and evolution in compositional molecular assemblies. We have performed numerical simulations of an ensemble of GARD networks, each with a different set of lognormally distributed catalytic values. We asked what is the influence of the catalytic content of such networks on beneficial evolution. Importantly, a clear trend was observed, wherein only networks with high mutual catalysis propensity (p(mc)) allowed for an augmented diversity of composomes, quasi-stationary compositions that exhibit high replication fidelity. We have reexamined a recent analysis that showed meager selection in a single GARD instance and for a few nonstationary target compositions. In contrast, when we focused here on compotypes (clusters of composomes) as targets for selection in populations of compositional assemblies, appreciable selection response was observed for a large portion of the networks simulated. Further, stronger selection response was seen for high p(mc) values. Our simulations thus demonstrate that GARD can help analyze important facets of evolving systems, and indicate that excess mutual catalysis over self-catalysis is likely to be important for the emergence of molecular systems capable of evolutionlike behavior.