Journal: Molecular Biology and Evolution

Article Title: Molecular clock dating using complex mixture models: applied to ancient symbionts

doi: 10.1093/molbev/msag039

Figure Lengend Snippet: Comparison of the accuracy of divergence time and branch-specific substitution rate estimation under substitution models LG + G and LG + C60 + G by simulation. a and b) Each plot contains 30 dots representing the mean relative difference (compared to the true values used in simulation) based on MCMCtree molecular clock dating with the corresponding calibration strategy on 30 simulated datasets. For each simulation, 30 timetrees each containing 20 tips are simulated using a birth-death process with a mean substitution rate 0.25 substitutions/site/Gyr across four different root ages ranging from 1.0 to 4.0 Ga. Sequence alignments are generated with LG + C60 + G4{1.0} (+G4{1.0} indicates α = 1 in a four-category discrete Gamma-distributed across-site relative rates) under AR (left-hand side) and IR (right-hand side) clock models, respectively. P -values are obtained from a paired Wilcoxon signed-rank test. The box plots (median, interquartile range, 1.5×IQR whiskers) illustrate the relative differences in time a) and rate b) estimates compared to true values. c) A schematic figure showing the five calibration schemes used for molecular clock analyses on simulated data. root_only : a single bounded calibration (both upper and lower bounds); single_min : bounded calibration at the root and one minimum-bounded internal node; single_interval : bounded calibration at both the root and an internal node; two_min : bounded calibration at the root and two minimum-bounded internal nodes; two_intervals : bounded calibration at both the root and two internal nodes.

Article Snippet: Under complete taxon sampling, the difference in time estimates between the LG + G and LG + C60 + G models became more obvious (scheme complete_taxon_sampling in ).

Techniques: Comparison, Sequencing, Generated

Journal: Molecular Biology and Evolution

Article Title: Molecular clock dating using complex mixture models: applied to ancient symbionts

doi: 10.1093/molbev/msag039

Figure Lengend Snippet: Evaluation of the impacts of calibration and model misspecification on molecular clock dating. Accuracy of divergence time estimation (relative difference) is shown as the y -axis. a and b) The lower and upper time bounds of all calibrations are shifted forward a) and backward b) in time by 20% (f0.2, b0.2) or 50% (f0.5, b0.5) of their true values, respectively. Three calibration schemes are considered: (i) a root-only calibration ( root_only ), (ii) the root plus one fully calibrated internal node ( single_interval ), and (iii) the root plus two fully calibrated internal nodes ( two_intervals ), each with time ranges of the true value ±20% or ±50%. c) Assessment of the accuracy of divergence time estimation under alternative substitution models used in simulation (LG + C60 + G used in most other analyses): LG + G, LG + G + I, LG + R, LG + C20 + G, and LG + C60 + G + I. Boxplots compare the accuracy obtained by the true model (purple) and by LG + C60 + G (blue). Each plot contains 30 dots representing the mean relative difference based on MCMCtree molecular clock dating with the corresponding calibration strategy on 30 simulated datasets. For each simulation, 30 timetrees each containing 20 tips are simulated using a birth-death process with a mean substitution rate 0.25 substitutions/site/Gyr across four different root ages ranging from 1.0 to 4.0 Ga. The plots on the left and right correspond to analyses using the AR and IR clock models, respectively. Calibration strategies are the same as those used in .

Article Snippet: Under complete taxon sampling, the difference in time estimates between the LG + G and LG + C60 + G models became more obvious (scheme complete_taxon_sampling in ).

Techniques:

Journal: Molecular Biology and Evolution

Article Title: Molecular clock dating using complex mixture models: applied to ancient symbionts

doi: 10.1093/molbev/msag039

Figure Lengend Snippet: Evaluation of the approximation of PMSF and bootstrapped Hessian on molecular clock analysis by simulation. Phylogenies are simulated under the same strategy as the focal dataset. The sequence alignment is simulated under LG + C60 + G{1.0}. The bootstrapped Hessian matrix is calculated using the software bs_inBV. The y axis indicates the relative accuracy of the posterior mean of the divergence times a) or substitution rates b) compared to that obtained by the first MCMC run under LG + C60 + G (note that this is different from other figures where the baseline in comparison is the true simulated times and rates). Each plot contains 30 dots representing the mean relative difference based on MCMCtree molecular clock dating with the corresponding calibration strategy on 30 simulated datasets. For each simulation, 30 timetrees each containing 20 tips are simulated using a birth-death process with a mean substitution rate 0.25 substitutions/site/Gyr across four different root ages ranging from 1.0 to 4.0 Ga. LG + C60 + G_run2 : the second independent MCMC run of MCMCtree under LG + C60 + G. LG + C60 + G + PMSF : PMSF approximation with LG + C60 + G. LG + C60 + G + bs_inBV : bootstrapped Hessian under LG + C60 + G. LG + C60 + G + PMSF + bs_inBV: bootstrapped Hessian under LG + C60 + G + PMSF. Calibration strategies are the same as those used in .

Article Snippet: Under complete taxon sampling, the difference in time estimates between the LG + G and LG + C60 + G models became more obvious (scheme complete_taxon_sampling in ).

Techniques: Sequencing, Software, Comparison