Question

Is There A Way To Get The Single Best Model From A Group Of Models/States Generated By Chromhmm?

0

Entering edit mode

11.6 years ago

Dataminer ★ 2.8k

Hi!

Is their a way to get the single best model from a group of models/states generated by ChromHMM?

Thank you

chip-seq • 3.3k views

ADD COMMENT • link updated 2.8 years ago by Ram 44k • written 11.6 years ago by Dataminer ★ 2.8k

1

Entering edit mode

you can decide arbitrarily as long as you can make sense of the number of chromatin states that you see.

ADD REPLY • link 11.6 years ago by biorepine ★ 1.5k

1

Entering edit mode

Hi, sorry for an old thread, but did you ever figure out how to do this? I am thinking of using BIC because most papers using ChromHMM end up using BIC to score their models and pick the best one. I do have a question about the "size" of my observed data though.

ADD REPLY • link 10.2 years ago by Affan ▴ 310

Ram · Answer 1 · 2014-12-02

I'm going to point you to an answer I wrote to a similar question on CrossValidated: http://stats.stackexchange.com/a/126343/61963

The linked answer is more detailed, but the short version of the answer is: BIC/AIC are not good methods to select the best model in this case, and the only good model is the one that you can interpret biologically (by cross-referencing with secondary annotations like TSSs, enhancers, etc).