Detecting patterns of species diversification in the presence of both rate shifts and mass extinctions

Abstract

BackgroundRecent methodological advances allow better examination of speciation and extinction processes and patterns. A major open question is the origin of large discrepancies in species number between groups of the same age. Existing frameworks to model this diversity either focus on changes between lineages, neglecting global effects such as mass extinctions, or focus on changes over time which would affect all lineages. Yet it seems probable that both lineages differences and mass extinctions affect the same groups.ResultsHere we used simulations to test the performance of two widely used methods under complex scenarios of diversification. We report good performances, although with a tendency to over-predict events with increasing complexity of the scenario.ConclusionOverall, we find that lineage shifts are better detected than mass extinctions. This work has significance to assess the methods currently used to estimate changes in diversification using phylogenetic trees. Our results also point toward the need to develop new models of diversification to expand our capabilities to analyse realistic and complex evolutionary scenarios.Electronic supplementary materialThe online version of this article (doi:10.1186/s12862-015-0432-z) contains supplementary material, which is available to authorized users.

What affects power to estimate speciation rate shifts?

“…Not surprisingly, power increased as the speciation rate asymmetry increased, which has also been reported in other studies (e.g., Davis, Midford & Maddison, 2013 ; Laurent, Robinson-Rechavi & Salamin, 2015 ). All methods performed poorly when the subtree speciation rate increased by only 50% (1.5X) relative to the basetree , suggesting that moderate rate increases are difficult to detect.…”

“…Previous studies have identified shortcomings specific to particular modelling approaches for estimation of rate shifts in phylogenies (e.g., Rabosky, 2010 ; Davis, Midford & Maddison, 2013 ; Laurent, Robinson-Rechavi & Salamin, 2015 ; Rabosky & Goldberg, 2015 ; Gamisch, 2016 ; Moore et al, 2016 ; Kozak & Wiens, 2016 ; May & Moore, 2016 ; Wiens, 2017 ; Burin et al, 2018 ; Simpson et al, 2018 ). We simulated large sets of trees where speciation and extinction rates remained constant throughout the tree ( basetree ), apart from an increase in speciation rate at a single node ( subtree ), and analyzed these trees using three widely used modelling approaches.…”

“…They showed that BiSSE analyses on small trees are prone to high Type II error. Laurent, Robinson-Rechavi & Salamin (2015) tested the performance of MEDUSA for both simple scenarios with a single speciation rate shift and more complex scenarios with multiple shifts and mass extinctions. Interestingly, they found that overall tree size increased power in the complex scenarios (their Figure 5a), but not for the single rate shift scenario (their Fig.…”

“…Testing hypothesis of diversification rate variation is underpinned by the ability to decouple and accurately estimate these rates. Estimates of extinction rates from phylogenies of extant taxa appear to be error prone ( Rabosky, 2010 ; Laurent, Robinson-Rechavi & Salamin, 2015 ; May & Moore, 2016 ; but see Stadler, 2013 ; Beaulieu & O’meara, 2016 ). On the other hand, speciation rate estimates are generally considered to be more robust.…”

“…30% on average), and that the estimated diversification parameters are biased. The probability of rate shift detection in MEDUSA depends on the number of terminals in the tree ( Laurent, Robinson-Rechavi & Salamin, 2015 ). Moore et al (2016) showed that the accuracy of BAMM is strongly affected by the priors specified, and that the estimates diversification rate parameters are unreliable (although see Rabosky, Mitchell & Chang, 2017 ).…”