A system for studying evolution of life-like virtual organisms

Alex A Neyfakh,Natalya N Baranova,Lev J Mizrokhi

doi:10.1186/1745-6150-1-23

Abstract

BackgroundFitness landscapes, the dependences of fitness on the genotype, are of critical importance for the evolution of living beings. Unfortunately, fitness landscapes that are relevant to the evolution of complex biological functions are very poorly known. As a result, the existing theory of evolution is mostly based on postulated fitness landscapes, which diminishes its usefulness. Attempts to deduce fitness landscapes from models of actual biological processes led, so far, to only limited success.ResultsWe present a model system for studying the evolution of biological function, which makes it possible to attribute fitness to genotypes in a natural way. The system mimics a very simple cell and takes into account the basic properties of gene regulation and enzyme kinetics. A virtual cell contains only two small molecules, an organic nutrient A and an energy carrier X, and proteins of five types – two transcription factors, two enzymes, and a membrane transporter. The metabolism of the cell consists of importing A from the environment and utilizing it in order to produce X and an unspecified end product. The genome may carry an arbitrary number of genes, each one encoding a protein of one of the five types. Both major mutations that affect whole genes and minor mutations that affect individual characteristics of genes are possible. Fitness is determined by the ability of the cell to maintain homeostasis when its environment changes. The system has been implemented as a computer program, and several numerical experiments have been performed on it. Evolution of the virtual cells usually involves a rapid initial increase of fitness, which eventually slows down, until a fitness plateau is reached. The origin of a wide variety of genetic networks is routinely observed in independent experiments performed under the same conditions. These networks can have different, including very high, levels of complexity and often include large numbers of non-essential genes.ConclusionThe described system displays a rich repertoire of biologically sensible behaviors and, thus, can be useful for investigating a number of unresolved issues in evolutionary biology, including evolution of complexity, modularity and redundancy, as well as for studying the general properties of genotype-to-fitness maps.ReviewersThis article was reviewed by Drs. Eugene Koonin, Shamil Sunyaev and Arcady Mushegian.

Highlights

Fitness landscapes, the dependences of fitness on the genotype, are of critical importance for the evolution of living beings
Theoretical treatments of evolution are mostly confined to the level of populations, where organisms are represented as black boxes and, the genotype-to-fitness maps must be postulated, because they cannot be inferred
Population-level theory has been very successful in studying the dynamics of genotype frequencies [1] but can never address the key issues concerned with evolution of functioning phenotypes, such as optimality, complexity, modularity, robustness, and evolve-ability

Summary

Introduction

The dependences of fitness on the genotype, are of critical importance for the evolution of living beings. Theoretical treatments of evolution are mostly confined to the level of populations, where organisms are represented as black boxes and, the genotype-to-fitness maps (fitness landscapes) must be postulated, because they cannot be inferred. Computer scientists, impressed by the beauty and apparent universality of Darwin's theory, proposed a concept of genetic algorithm, a software environment in which strings of information ("chromosomes") undergo repetitive cycles of fitness determination, selection, and mutation/recombination [2,3]. "Computer scientists, impressed by the beauty and apparent universality of Darwin's theory, proposed a concept of genetic algorithm, a software environment in which strings of information ("chromosomes") undergo repetitive cycles of fitness determination, selection, multiplication, and mutation/recombination [2,3]. Genetic algorithms proved to be efficient in diverse optimization tasks in engineering [4,5] and software design [6]

Methods

Results

Discussion

Conclusion