Advanced

Artificial neural networks in models of specialization and sympatric speciation

Norrström, Niclas LU (2009)
Abstract
This thesis deals with specialization and how it is linked to sympatric speciation. The trait driving specialization is a cue recognition trait modelled with artificial neural networks that exploiters use to discriminate beneficial resources from detrimental resources based on the signals of the resources. Paper I investigates how haploid exploiters and the resources coevolve when the signals of the resources can evolve through mutations. We find that this coevolution can be a cyclic process with saltational changes between different stages and that evolution is only directional and the exploiters are only specialists in parts of this cycle. In simulations underlying Paper II the signals of the resources can not mutate but the exploiters... (More)
This thesis deals with specialization and how it is linked to sympatric speciation. The trait driving specialization is a cue recognition trait modelled with artificial neural networks that exploiters use to discriminate beneficial resources from detrimental resources based on the signals of the resources. Paper I investigates how haploid exploiters and the resources coevolve when the signals of the resources can evolve through mutations. We find that this coevolution can be a cyclic process with saltational changes between different stages and that evolution is only directional and the exploiters are only specialists in parts of this cycle. In simulations underlying Paper II the signals of the resources can not mutate but the exploiters have a diploid genome and the organisms reproduce sexually. We show that the disruptive selection stemming from exploiters specializing on different resources can overcome the homogenizing effect of sexual recombination when exploiters mate randomly and produce a functional genetic polymorphism with specialized exploiters. A functional genetic polymorphism removes the force of reinforcement but we run simulations where the exploiters have a mating gene determining if mating is random or if exploiters should mate assortatively in Paper III. We find that assortative invades the exploiter population and homozygote specialists evolve because the genetic polymorphism pays a cost by having some alleles being silenced (that is they do not contribute to the complete phenotype) in certain genotypes so a mutation in these silenced alleles is not selected against, which cause these alleles to accumulate deleterious mutations. The homozygote specialists, mating assortatively, are much more efficiently removing deleterious mutations from the population and hence can invade the population. Finally, in Paper IV we investigate the effects of resource and resource signal arrangements in the environment. We show that the environment can influence the evolution of specialization and sympatric speciation. By modelling a resource discrimination trait based on the interaction of epistatic genes we find a novel force promoting sympatric speciation over genetic polymorphisms. (Less)
Please use this url to cite or link to this publication:
author
supervisor
opponent
  • Kisdi, Eva, Department of Mathematics and Statistics, University of Helsinki, Gustaf Hällströmin katu 2b, Helsingfors, Finland
organization
publishing date
type
Thesis
publication status
published
subject
defense location
Blå Hallen, Ekologihuset
defense date
2009-02-20 13:00
language
English
LU publication?
yes
id
b5dfa7f2-1c83-4101-9f87-76cb2c12ad70 (old id 1288236)
date added to LUP
2009-01-29 12:35:37
date last changed
2016-09-19 08:45:17
@phdthesis{b5dfa7f2-1c83-4101-9f87-76cb2c12ad70,
  abstract     = {This thesis deals with specialization and how it is linked to sympatric speciation. The trait driving specialization is a cue recognition trait modelled with artificial neural networks that exploiters use to discriminate beneficial resources from detrimental resources based on the signals of the resources. Paper I investigates how haploid exploiters and the resources coevolve when the signals of the resources can evolve through mutations. We find that this coevolution can be a cyclic process with saltational changes between different stages and that evolution is only directional and the exploiters are only specialists in parts of this cycle. In simulations underlying Paper II the signals of the resources can not mutate but the exploiters have a diploid genome and the organisms reproduce sexually. We show that the disruptive selection stemming from exploiters specializing on different resources can overcome the homogenizing effect of sexual recombination when exploiters mate randomly and produce a functional genetic polymorphism with specialized exploiters. A functional genetic polymorphism removes the force of reinforcement but we run simulations where the exploiters have a mating gene determining if mating is random or if exploiters should mate assortatively in Paper III. We find that assortative invades the exploiter population and homozygote specialists evolve because the genetic polymorphism pays a cost by having some alleles being silenced (that is they do not contribute to the complete phenotype) in certain genotypes so a mutation in these silenced alleles is not selected against, which cause these alleles to accumulate deleterious mutations. The homozygote specialists, mating assortatively, are much more efficiently removing deleterious mutations from the population and hence can invade the population. Finally, in Paper IV we investigate the effects of resource and resource signal arrangements in the environment. We show that the environment can influence the evolution of specialization and sympatric speciation. By modelling a resource discrimination trait based on the interaction of epistatic genes we find a novel force promoting sympatric speciation over genetic polymorphisms.},
  author       = {Norrström, Niclas},
  language     = {eng},
  school       = {Lund University},
  title        = {Artificial neural networks in models of specialization and sympatric speciation},
  year         = {2009},
}