Knowles & Carstens 2007

Knowles, L.L. and B.C. Carstens. 2007. Delimiting species without monophyletic gene trees. Syst. Bio. 56(6): 887-895.

Most researchers who use phylogenetic analyses to delimit species would say that reciprocal monophyly is an important criterion in this endeavor. Not so! According to Knowles and Carsten, one can model the probability of the relationship between gene trees and species history and get a good answer as to what constitutes a species.

Their preliminary simulation study suggests that very recently derived species can be accurately identified long before the requisite time for reciprocal monophyly is achieved following speciation.

-species delimitation will be misled by discordance if gene lineages within a species coalesce below the species divergence (also known as the species-tree gene-tree discordance problem Maddison 1997).

*a gene tree should not be equated with a species tree*

However, gene treeS do provide information about the history of species splitting (the species tree) despite widespread incomplete lineage sorting.

Knowles and Carstens use a coalescent framework to estimate gene-tree probabilities under a particular history to evaluate the likelihood of lineage splitting (i.e. that speciation has occurred).

-focuses on the stochastic loss of gene lineages by genetic drift

Methods

-gene genealogies were simulated at different depths
-the species tree was simulated using a Yule model
-gene trees were simulated under a neutral-coalescent process without gene flow

-the product of the probabilities from the gene trees of each locus under a specific history was used to evaluate the likelihood of whether species A and B are separate species lineages
-for this the authors used the program COAL, calculated likelihoods, and likelihood-ratio tests
-used Mesquite to replicate 100 data sets

Results

-the A and B lineages were successfully delimited with the coalescent-based approach across all the different times of divergence that the authors looked at
-increased sampling of loci (genes) resulted in a decrease in false-negatives (failures to delimit the separate species)
-inferred species boundaries will only be reliable to the extent that the model used is an accurate account of the process of speciation
-the power of the test of species delimitation clearly depends on the number of sampled loci.
-the info contained in independent loci provide valuable info for delimiting species, even though the gene trees are not completely concordant -the take home message is always use more than one gene

Conclusions

-what is recognized as a species boundary is very much influenced by the method used to delimit species (Sites and Marshall 2004)
-Knowles and Carstens show that it is possible to accurately delimit species despite widespread incomplete lineage sorting and discordance among loci if one uses a probablilistic modeling approach.