Thumbnail Image
Long, Lijiang
Associated Organization(s)
Organizational Unit
Organizational Unit
Supplementary to
Ever since Darwin’s qualitative theory of the origin of species, there is growing demand for quantitative methods to study mechanisms underlying the speciation process. One key component towards new species formation is reproductive isolation. Reproductive isolation can be either pre-mating (e.g. mating behavior difference) or post-mating (e.g. Dobzhansky-Muller sites). In this thesis, I will present novel quantitative methods developed to study two aspects of reproductive isolation: mating rituals difference in Lake Malawi cichlids and genetic incompatibilities by selfish genetic elements in C. elegans. In chapter 2, I will talk about cichlids bower behaviors where male fishes construct bowers to attract female mates by manipulating sand with their mouths thousands of times over the course of many days. Variations in bower type (‘pits’ and ‘castles’) is one mechanism to create nonrandom mating and maintain a large number of cichlids species in Lake Malawi. To enable quantitative comparisons of these behaviors in different species, an automatic behavior quantification pipeline was built. Specifically, pixel-based Hidden Markov Modeling was combined with density-based spatiotemporal clustering for action detection. Each action video clip was then classified into ten categories using a 3D Residual Network (3D ResNet). These ten categories distinguish spitting, scooping, fin swipes and spawning. I showed that this approach is accurate (> 76% accuracy) in distinguishing fish behaviors and animal intent can be determined from these clips, as spits and scoops performed during bower construction are classified independently from spits and scoops performed during feeding. I applied this approach to >700 hours of video recordings taken from seven independent trials encompassing multiple species and hybrid crosses, collectively containing hundreds of thousands of independent behavioral events. In chapter 3, I use quantitative methods to measure fitness combined with population modeling to study the evolutionary origin of selfish genetic elements and their ability to spread in populations. Previous research found that a toxin-antidote element called peel-zeel is under balancing selection. Here, I explore different models that could cause balancing selection on this locus, which make different predictions on the fitness effect of the peel-zeel locus in hermaphrodites. However, pair-wise competition assays showed the loss of the toxin gene peel-1 decreased fitness of hermaphrodites, contradicting my expectation that peel-1 will decrease animal fitness due to its toxicity. This fitness advantage is independent of the antidote gene zeel-1. This work showed that toxin-antidote systems can spread through populations independent of their selfish effects and suggests linked variants for dauer pheromone response could be responsible for the balancing selection. Finally, in chapter 4, I use simulation methods to study the effect of toxin-antidote elements on linked and unlinked genetic variation in the case of admixture. While both simulation and calculation showed toxin-antidote elements are able to quickly spread in a population without toxin-antidote element, the evolution trajectories of the rest of the genome depends on the initial frequency of the toxin-antidote haplotype in the admixed population. Using calculations and simulations, I showed that unlinked neutral genetic variants will increase their frequency when the initial frequency of peel-zeel is higher than 1/3 and decrease when the initial frequency of peel-zeel is lower than 1/3. My doctoral thesis with many quantitative methods will advance our understanding of the genetic basis of species evolution and evolutional dynamics of selfish genetic elements.
Date Issued
Resource Type
Resource Subtype
Rights Statement
Rights URI