search

veg / hyphy-analyses

HyPhy standalone analyses
MIT License
37 stars 17 forks source link

RELAX-scan/RELAX.bf results #28

Open francicco opened 1 year ago

francicco commented 1 year ago

Hi @spond,

I'm running the Relax scan analysis. It seems like I always have different results using the same dataset and command line, also on the example data. I wonder if it is how it is suppose to be or there's something wrong with it. Cheers F

spond commented 1 year ago

Dear @francicco,

Some of these analyses (including RELAX-scan) are quite noisy (complex models). The original README was test-run on a HyPhy release that is quite old and the new version has more stable and robust optimization routines.

If you run hyphy RELAX.bf --alignment data/Fig4H.nex with the current version you get

  1. A better log likelihood (-1614.99 vs -1616.21) for the 2-rate model
  2. The same two branches flagged as possibly relaxed/intensified and two new branches, which were marginal in the previous version
Branch Previous p Current p k (old/new)
Node 7 0.04 0.04 0 / 0.10
EU912347 ... 0.00 0.00 0.10 / 0.07
EU912347 ... 0.00 0.00 0.10 / 0.07
Node7 0.12 0.04 0.21 / 0.10
EU912349 0.09 0.01 0.22 / 0.11

Best, Sergei

francicco commented 1 year ago

Hi Sergei,

thanks a lot! What I'd to is run the regular RELAX on multiple subsequent branches and correct for multiple test afterward. It's probably slower but more reliable I guess.

Cheers F