Data from: Testing the molecular clock using mechanistic models of fossil preservation and molecular evolution

Molecular sequence data provide information about relative times only, and fossil-based age constraints are the ultimate source of information about absolute times in molecular clock dating analyses. Thus, fossil calibrations are critical to molecular clock dating, but competing methods are difficult to evaluate empirically because the true evolutionary time scale is never known. Here, we combine mechanistic models of fossil preservation and sequence evolution in simulations to evaluate different approaches to constructing fossil calibrations and their impact on Bayesian molecular clock dating, and the relative impact of fossil versus molecular sampling. We show that divergence time estimation is impacted by the model of fossil preservation, sampling intensity and tree shape. The addition of sequence data may improve molecular clock estimates, but accuracy and precision is dominated by the quality of the fossil calibrations. Posterior means and medians are poor representatives of true divergence times; posterior intervals provide a much more accurate estimate of divergence times, though they may be wide and often do not have high coverage probability. Our results highlight the importance of increased fossil sampling and improved statistical approaches to generating calibrations, which should incorporate the non-uniform nature of ecological and temporal fossil species distributions.

在分子钟定年（molecular clock dating）分析中，分子序列数据仅能提供相对时间信息，而基于化石的年代约束才是绝对时间信息的终极来源。因此，化石校准（fossil calibrations）对于分子钟定年至关重要，但由于真实演化时间尺度始终未知，各类竞争性定年方法难以通过实证方式开展评估。本研究通过模拟实验整合化石保存与序列演化的机制模型，用以评估构建化石校准的各类方法，及其对贝叶斯分子钟定年（Bayesian molecular clock dating）的影响，同时对比化石采样与分子采样的相对影响。研究结果表明，分化时间估计（divergence time estimation）会受到化石保存模型、采样强度以及演化树型的影响。新增分子序列数据或可提升分子钟定年的估计结果，但定年的准确度与精密度仍主要取决于化石校准的质量。后验均值（posterior means）与后验中位数（posterior medians）难以准确反映真实分化时间；而后验区间（posterior intervals）虽往往较宽且覆盖概率（coverage probability）通常不高，却能更为准确地估计分化时间。本研究结果凸显了加大化石采样力度、优化校准生成统计方法的重要性，此类方法应纳入化石物种在生态与时间维度上的非均匀分布特征。