Computational Modelling Group

Origins of Evolvability

1st June 2013
20th September 2013
Research Team
William Hurndall
Richard Watson, Markus Brede

Mutation like transitions between micellar states

Little is known about the chemical makeup of the early world ~3.5 billion years ago when life is believed to have first appeared. Some propose that the first self-replicator capable of evolving would have been constituted by a primitive information encoding template with an integrated self-replicating function. These self-replicators might have been able to undergo a rudimentary accumulation of adaptation by mutation. Hypothesese such as this RNA World scenario necessarily assume the abundance of what are still relatively complicated genetic molecules though, also presenting the problem of explaining the transition from unlinked to linked templates and the emergence of translation. The beginning of translation represents the birth of the duality between information encoding substrates and copying machinery that underpins biological reproduction in all known life today.

A class of theories have emerged that avoid assumptions of a minimally complex biological molecule though. These use collective properties of naturally forming catalysts such as proteins (today proteins form integral parts of the gene expression functions of the translation apparatus) to encode information. Self-replication takes the form of catalytic closure of sub-sets of such catalysts. The basic idea is that if many such distinct sub-sets existed, they might act as units of heredity in some dynamic setting.

The purpose of this project was to investigate the potential evolvability of growing and fissioning micelles constituted by a diverse range of lipid species. The original model that was used [1] assumed that a micelle would catalyse the integration of an external lipid in number approximately proportional to the additive catalytic strength of lipids in the micelle and number of lipids of the particular lipid species in the environment. The stochastic integration process was guided by coupled rate equations representing a complex directed (internal-external lipid) and weighted (catalytic strengths) reaction network. In population selection tests, mutation like transitions were observed, but due to the stochastic transition between attractors of the system that were determined by the kinetics, heritibility was ultimately poor and selection could not neither favour specific attractors or accumulate adaptation away from attractors.

An attempt to modify the model to allow selection to act against unwanted lipids that would promote transitions was made. This modification was to drop the assumption that interactions between lipids were not-spatially localised, ie., to make the assumption that if a lipid in a micelle integrated an external lipid, it would integrate it at a local site in a micelle rather than at a random location. The purpose was to allow the occassional integration of unwanted lipids to occur at spatially localised sites in the micelle so that the fissioning process would not faithfully pass these invaders to both daughter micelles. The results found that the coupling between attractors was too strong for this technique to work well. However, idealised networks were built in which such coupling was parameterised revealed that the modification did infact allow selection to act on lipid species and combinations that were not attractors. The strength of selection was found to be reduced as catalytic coupling increased.

Summary: The localised interactions and structured fissioning of micelles was demonstrated to drastically increase the strength of selection in many cases where coupling was minimised and the targets of selection were individual species, but the increase in variation that allowed this also prevented selection of complex lipid targets that were distinct from kinetically determined attractors. While the mutation-selection balance is known to exist in template based replication, the role of structured fissioning has not been previously contrasted to fissioning of well-mixed micelle-like compartments previously and this work has demonstrated a set of paramaterisable artificial chemistries in which this modification has increased evovability, albeit in a very limited manner.

[1] Daniel Segr ́e, Dafna Ben-Eli, and Doron Lancet. Compositional genomes: prebiotic information transfer in mutually catalytic noncovalent assemblies. Proceedings of the National Academy of Sciences, 97(8):4112–4117, 2000.


Life sciences simulation: Evolution

Physical Systems and Engineering simulation: Catalysis

Socio-technological System simulation: Self Organized Networks, Social Networks

Algorithms and computational methods: Artificial Neural Networks, Cellular automata, Graph Theory, Molecular Dynamics, Monte Carlo, statistical analysis

Visualisation and data handling software: Gnuplot, Pylab

Software Engineering Tools: Eclipse

Programming languages and libraries: C, Matlab, Python, R

Computational platforms: Linux

Transdisciplinary tags: Complex Systems, Computer Science, Scientific Computing