Computational protocol: Heterotachy and long-branch attraction in phylogenetics

Similar protocols

Protocol publication

[…] We followed a similar protocol as in [], with the only difference being in the branch lengths of the model topology. Briefly, DNA sequences of 10,000 nucleotides each were simulated under the Jukes and Cantor [] model with Seq-Gen version 1.2.7 []. Modelling rate heterogeneity across sites using a Gamma distribution (α = 0.5 and 1) gave similar results (data not shown). Considering a transition/transversion ratio greater than 1 (2, 5 or 10) rendered ML more accurate than standard MP (see also []), but when a weighted MP is used the same results as with a ratio of 1 were obtained (data not shown). As described in Fig. , a single parameter, τ, allows for the adjustment of the level of heterotachy from fully homotachous (τ = 0) to extreme heterotachous (τ = 1) conditions. We varied τ from 0 to 0.9 by a step of 0.1. The two partitions were always of the same size (w = 0.5). As detailed in the main text, various values of p and q are used. The internal branch r was varied from 0 to 0.4 with a step of 0.01. One hundred simulations were performed for each combination of p, q, r and τ. Phylogenies were inferred by MP and ML (with a Jukes and Cantor model) using PAUP* version 4.0b10 []. Finally, to estimate the accuracy for both methods, BL50 (i.e. the value of r for which 50% of the simulations recover the correct tree) was computed through nonlinear regression using the R software version 2.0.0 []. When r < BL50, increasing sequence length decreases tree reconstruction method accuracy [], which corresponds to the definition of inconsistency. […]

Pipeline specifications

Software tools Seq-Gen, PAUP*
Application Phylogenetics