search

veg / hyphy

HyPhy: Hypothesis testing using Phylogenies
http://www.hyphy.org
Other
209 stars 69 forks source link

Omega values in hyphy vs codeml calculated dN/dS? #1335

Closed gtb-togerther closed 3 years ago

gtb-togerther commented 3 years ago

Dear Sergei, thanks for developing and maintaining the great tool hyphy, I am not sure the omega value been calculated by hyphy are the same thing given by Codeml free ratio model? I wonder do you have some clues why hyphy calculated omega values have so many 0 and 1 in tip branch? Does omega value 1 stand for the neutral evolution of this sequence?

Another thing is could hyphy ABSREL model calculated Baseline MG94xREV omega value in 'branch attributes' of each node be correlated with trait data like body mass?

Thank you so much and looking forward to your answers!

WJ Chen

spond commented 3 years ago

Dear @gtb-togerther,

  1. Yes, HyPhy mean ω values and dN/dS values reported by codeml should be broadly comparable. You really do need to be careful with which models you use, because HyPhy has about 20 different ways to calculate branch level ω values.
  2. And ω value of 0 typically means that a particular branch has no non-synonymous changes. A value of exactly 1 is unusual and probably means that the branch has 0 length, i.e. ω is not identifiable (can't be inferred).
  3. I don't see why such a correlation could not be carried out. Are you taking about inferring ω values and then regressing them on a trait, or directly including trait values in ω estimation?

Best, Sergei

gtb-togerther commented 3 years ago

Dear Sergei,

Really appreciate your quick and detailed reply.

  1. I am dealing with ω values of the ABSREL model from its branch attributes parsed from JSON files. Does this a feasible model to calculate branch level ω values or is there any other better model to calculate the omega value of each node given a tree and an alignment?
  2. Yes it's quite strange to have too many '1' in branch-level omega( mentioned in 1.), Does this because of too few changes to calculate branch or node ω ? and gene-level omega will not like this?
  3. Yes, the intention is to inferring ω values and regressing them on a trait, to find out whether some gene have similar selection pattern associated with the trait, I have seen similar studies using tip-to-root dN/dS calculated using codeml free-ratio model (eg, G Muntané · 2018, https://doi.org/10.1093/molbev/msy105 and also see the figure) , but I am not sure is it ok to use hyphy omega to regression on a trait as well? image

Best,

WJ Chen

github-actions[bot] commented 3 years ago

Stale issue message