Towards a Statistical Species Concept and Speciation

Abstract

Defining species has been a central unsolved problem in biology. Modern taxonomy uses objective methods to reconstruct phylogenetic trees but there are no objective methods for delimitation of species as well as for higher taxonomic units. These algorithms do not make any null hypothesis and therefore do not show whether significant clusters exist or whether hierarchical classification exists. We propose a Statistical Species Concept (SSC) based on the Frequency Distribution plot (FDp) of the distances between individuals, which can reveal clusters by segregating the distribution of within and between cluster distances. This algorithm, based on SSC, is then tested with synthetic sequences and compared with the molecular phylogenetic analysis by Maximum Likelihood (ML) method. We find that out of the 100 cases we tested for, SSC predicted the correct hierarchy 94 times while ML gave satisfactory results only 31 times. We also tested the SSC algorithm on real genetic data to compare the predictions of SSC with existing taxonomic ranks of those individuals. We found that the species level of classification corresponded to first level of clustering in the FDp as was predicted based on SSC. We also found clustering that corresponded to the genera level of classification. Finally we develop a platform of computational models to understand speciation and use the objective criteria developed for SSC. We find interesting preliminary results; i) Neutral drift can cause speciation for asexual populations under specific circumstances ii) Patch-dynamics does not facilitate speciation and iii) Competition adds substantial stability to speciation