Department of Biology, Carleton University, 209 Nesbitt Biology Building, 1125 Colonel By Drive, Ottawa, ON, K1S 5B6, Canada.
Institute of Biochemistry, Carleton University, Ottawa, ON, Canada.
J Mol Evol. 2022 Aug;90(3-4):239-243. doi: 10.1007/s00239-022-10058-0. Epub 2022 Jun 2.
We draw attention to an under-appreciated simulation method for generating artificial data in a phylogenetic context. The approach, which we refer to as jump-chain simulation, can invoke rich models of molecular evolution having intractable likelihood functions. As an example, we simulate data under a context-dependent model allowing for CpG hypermutability and show how such a feature can mislead common codon models used for detecting positive selection. We discuss more generally how this method can serve to elucidate the ways by which currently used models for inference are susceptible to violations of their underlying assumptions. Finally, we show how the method could serve as an inference engine in the Approximate Bayesian Computation framework.
我们提请注意一种在系统发育背景下生成人工数据的未被充分重视的模拟方法。我们称之为跳跃链模拟的方法，可以调用具有难以处理的似然函数的丰富的分子进化模型。例如，我们在允许 CpG 超突变的上下文相关模型下模拟数据，并展示这样的特征如何会误导用于检测正选择的常见密码子模型。我们更一般地讨论了这种方法如何有助于阐明目前用于推理的模型如何容易违反其基本假设。最后，我们展示了该方法如何在近似贝叶斯计算框架中充当推理引擎。
