Reply to Patel and Kumar: Epistasis not critical to rate distributions but key in heterotachy, overdispersion, and evolutionary Stokes shift
Reply to Patel and Kumar: Epistasis not critical to rate distributions but key in heterotachy, overdispersion, and evolutionary Stokes shift
复制标题
回复帕特尔和库马尔：上位性对速率分布并不重要，但在异态性、过度分散和进化斯托克斯位移中至关重要
DOI：
10.1073/pnas.2100107118
复制
发表时间：
2021
期刊：
影响因子：
--
通讯作者：
Morcos, Faruck
中科院分区：
文献类型：
--
作者：
de la Paz, Jose Alberto;Morcos, Faruck
为了回应Patel和Kumar(1)，我们认为这封信的观察是对我们在工作(2)中暴露的东西的补充和支持，除了对我们的主张的一个微妙的误解。这封信建议我们将不同地点的可变性及其特殊形状解释为仅来自上位性的伽马分布(1)。混淆的根源可能来自(2)伽马分布定位率和异速的发现并不是先验地强加于模型，但它们仍然作为进化序列的属性出现。然而，我们并没有暗示与上位性的直接联系来解释变异的特定形状，而只是在讨论异速、过度分散和进化的斯托克斯移位时。事实上，有可能设想一种模型，其站点频谱与站点之间的变异性兼容，而不直接影响上位性，正如我们在讨论协方差模型(2)时在引言中提到的那样。相反，这种可变性源于单个站点的特征(我们在每个步骤采样站点的身份时确定PI，而不仅仅是P)。在我们的模型中，即使我们排除了模拟上位性的耦合，基于每个站点的概率分布都是单独抽样的事实，也有可能获得速率的变化。因此，上位性的预期贡献将是根据序列的进化史在动力学意义上影响这些变异。尽管如此，光谱的基本可变性可能仅仅是因为个别站点拥有不同的概率分布。关于信(1)中讨论的模型比较，我们意识到简单伽马拟合的局限性；在我们的文章(2)的末尾，我们明确地说：
In response to Patel and Kumar (1), we consider this letter’s observations to supplement and support what we exposed in our work (2) except for a subtle misunderstanding of our claims. The letter proposes that we explain variability across sites and its particular shape as a gamma distribution solely from epistasis (1). A source of confusion may come from the sentence on (2) the findings of gamma-distributed fixation rates and heterotachy were not imposed on the model a priori, but they nevertheless emerged as properties of the evolved sequences. However, we do not imply a direct association with epistasis to explain the particular shape of variation but solely when discussing heterotachy, overdispersion, and the evolutionary Stokes shift. Indeed, it is possible to conceive a model whose site frequency spectrum is compatible with variability across sites without a direct implication on epistasis, as was mentioned in our introduction when discussing the Covarion model (2). Instead, such variability arises from individual sites’ characterization (we determine Pi and not only P when sampling a site’s identity at each step). In the context of our model, it is possible to obtain variation in rates even if we were to exclude couplings that model epistasis, based on the fact that every site has its probability distribution sampled individually. Therefore, the contribution expected from epistasis would be to influence, in a dynamical sense, those variations according to the evolutionary history of a sequence. Still, the spectrum’s base variability can originate solely because individual sites possess a distinct probability distribution. Concerning model comparison, discussed in the letter (1), we are aware of the limitations of a simple gamma fit as presented; at the end of our article (2), we explicitly say:
DOI：
10.1073/pnas.1913071117
发表时间：
2020-03-17
影响因子：
11.1
作者：
de la Paz, Jose Alberto;Nartey, Charisse M.;Morcos, Faruck
通讯作者：
Morcos, Faruck