Species diversification rates
This protocol is extracted from research article:
Multiple macroevolutionary routes to becoming a biodiversity hotspot
Sci Adv, Feb 6, 2019; DOI: 10.1126/sciadv.aau8067

The DR metric (26) was calculated as the number of nodes that separated each species from the root of the tree weighted by the distance of each node to the present. Thus, DR represents the relative phylogenetic isolation of species and has been used to determine regions that are enriched in actively diversifying lineages and older lineages (27, 44). We divided the species into four quartiles, with older lineages (here termed “ancient”) in Q1 and younger, actively diversifying lineages (“recent”) in Q4. Using simple linear regression models, we predicted Q1 and Q4 SR from the total SR in each cell. Positive residuals of the corresponding linear model indicated that a particular cell contained an excess of ancient or recent lineages, respectively, while negative residuals indicated a deficit (27, 44). For each biogeographic realm, we then assessed whether an average hotspot cell had different values for the Q1 and Q4 residuals when compared to an average non-hotspot cell. To do this, we ran two spatial simultaneous autoregressive (SAR) error models, each with the Q1 and Q4 residuals as a response and hotspot category as a predictor using the R package spdep (45). Last, we compared the mean values of Q1 and Q4 SAR residuals in hotspots and non-hotspot cells of each biogeographic realm using a Wilcoxon rank sum test.

We also analyzed the effect of phylogenetic uncertainty in our estimates of DRs. We obtained DR estimates for 100 random trees from the pseudoposterior distribution of the mammal and bird phylogenies (see above). We then verified that the MCC-based DR estimates were positively correlated with the median DR values across the 100 trees (fig. S2A) and with each individual tree (fig. S2B). We also verified that the residuals of the DR quartile linear regressions with the MCC tree were correlated with the residual values across the 100 trees (fig. S2, C and D).

DR has been shown to correlate to recent speciation events. However, it does not account for extinction and temporal variation in diversification rates within lineages (26, 27). To check whether these assumptions could influence our analyses, we estimated species-specific rates of speciation and extinction (i.e., “tip rates”) in the mammal and bird phylogenies using Bayesian analysis of macroevolutionary mixtures (BAMM) (46). BAMM models diversification rate heterogeneity across lineages and through time. Although the reliability of BAMM has recently been questioned, a focus of criticism was the influence of unobserved rate shifts within extinct lineages on diversification estimates (47). This issue affects most methods, but the contribution of shifts in extinct lineages to the overall probability of extinction is marginal under biologically plausible scenarios (48). A related concern is that the extinction rate estimates obtained from molecular phylogenies of extant species may sometimes be unreliable (49, 50). Accurate speciation rates and relative diversification rates can, however, be routinely inferred using a variety of methods (51), including BAMM. Simulation studies have shown that BAMM infers speciation rates with high accuracy, particularly for large, well-sampled trees such as ours (48), and can robustly estimate speciation rates in the absence of paleontological data (52). Despite having different assumptions, BAMM and DR estimates of diversification are generally positively correlated (53), including in our dataset (fig. S3, E and F; Pearson’s rmammals = 0.63, rbirds = 0.58; for both, P < 0.0001). This result suggests that both measure capture similar information about diversification. We also found that residuals of DR- and BAMM-based quartile regressions were positively correlated (fig. S3, E and F; Pearson’s rmammals.Q1 = 0.27, rmammals.Q4 = 0.68, rbirds.Q1 = 0.65, rbirds.Q4 = 0.77; for all, P < 0.0001). Similarly, comparisons between hotspots and non-hotspot regions across realms using DR residuals (Fig. 1) largely matched those obtained using BAMM residuals (table S1).

Note: The content above has been extracted from a research article, so it may not display correctly.

Please log in to submit your questions online.
Your question will be posted on the Bio-101 website. We will send your questions to the authors of this protocol and Bio-protocol community members who are experienced with this method. you will be informed using the email address associated with your Bio-protocol account.

We use cookies on this site to enhance your user experience. By using our website, you are agreeing to allow the storage of cookies on your computer.