Files
Abstract
In biology, species is the fundamental unit of classification of organisms that share common characteristics and are able to interbreed. Determining species boundaries (i.e., species delimitation) is crucial for studies in ecology, evolutionary biology, and other biological fields. The biological definition of species based on reproductive isolation, however, is difficult to test using empirical data. Recent development of phylogenetic species provides a conceptual framework of stochastic models for species delimitation, in which species are defined as lineages in a phylogenetic tree. Researchers focusing on complex speciation scenarios during evolutionary process have shown that phylogenetic tools can accurately identify the boundaries of cryptic species. However, these methods are often time-consuming and fail to handle a massive amount of molecular sequence data. High performance of deep learning algorithms has inspired scientists to develop computationally efficient tools for species delimitation using DNA sequences. In this project, we apply neural networks algorithms for species delimitation using multi-locus sequence data. The performance of the proposed method is compared with other machine learning and model-based approaches. The results of simulation and real data analyses show that neural network outperforms the other approaches in delimitating species boundaries.