Read on Twitter
My lab had a tremendous pleasure today hearing from Gautam Reddy about how global epistasis emerges in a model of a complex trait.
The paper is pretty dense, but it is worth digging into. A lot of deep insights. Fantastic work! https://elifesciences.org/articles/64740 ">https://elifesciences.org/articles/...
The paper is pretty dense, but it is worth digging into. A lot of deep insights. Fantastic work! https://elifesciences.org/articles/64740 ">https://elifesciences.org/articles/...
I am not quite ready to say that diminishing returns/increasing costs epistasis (DRE/ICE) puzzle is solved, but the models proposed here and by @EvoMedDanny et al are very intriguing. Let’s call these two models “idiosyncratic epistasis” models. https://www.nature.com/articles/s41559-020-01286-y">https://www.nature.com/articles/...
Here are a few questions that I think should be resolved next. They fall into three categories. This is not an exhaustive list and not necessarily in the order of priority.
Category 1. Further tests of the idiosyncratic epistasis models. Specifically:
Category 1. Further tests of the idiosyncratic epistasis models. Specifically:
(*) Do their predictions hold over larger and smaller genetic distances as predicted?
(*) Does the DFE vary across genotypes as predicted?
(*) Which mutations/loci fail to display DRE/ICE and why?
(*) Does the DFE vary across genotypes as predicted?
(*) Which mutations/loci fail to display DRE/ICE and why?
Category 2. Model extension/testing:
(*) What kinds of JDFE shapes does the model predict across environments?
(*) Do the predictions hold?
(*) What kinds of JDFE shapes does the model predict across environments?
(*) Do the predictions hold?
Category 3. Reconciliation with other models
Idiosyncratic epistasis models are models of genotype-to-fitness map that assume lots of independent pairwise and higher-order epistatic interaction terms (hence “idiosyncratic” epistasis). But…
Idiosyncratic epistasis models are models of genotype-to-fitness map that assume lots of independent pairwise and higher-order epistatic interaction terms (hence “idiosyncratic” epistasis). But…
… we know that there is the phenotype layer in between genotype and fitness. Organisms have a complex non-random functional organization, so epistatic terms may not be independent. If idiosyncratic epistasis models are correct, what does it tell us about biological organization?
Specifically:
(*) How do idiosyncratic epistasis models relate to the Fisher’s geometric model? Guilliaume Martin’s argument for FGM is compelling. But some predictions of these two classes of models are in direct contradiction. Is FGM simply wrong? https://academic.oup.com/genetics/article/197/1/237/5935937">https://academic.oup.com/genetics/...
(*) How do idiosyncratic epistasis models relate to the Fisher’s geometric model? Guilliaume Martin’s argument for FGM is compelling. But some predictions of these two classes of models are in direct contradiction. Is FGM simply wrong? https://academic.oup.com/genetics/article/197/1/237/5935937">https://academic.oup.com/genetics/...
(*) How do idiosyncratic epistasis models relate to other phenotypic models? For example, my model of a metabolic network shows that the topology of the network constrains the types microscopic epistasis. https://elifesciences.org/articles/60200 ">https://elifesciences.org/articles/...
It’ll be interesting to know if at least some metabolic networks produce many independent idiosyncratic epistatic terms and DRE/ICE. If not, how do we build a phenotypic model that is consistent with DRE/ICE?
