search

Murali-group / BoolODE

Git Repo for simulating Boolean Models
GNU General Public License v3.0
33 stars 16 forks source link

Easy way to return derivatives? #13

Open calebclayreagor opened 4 years ago

calebclayreagor commented 4 years ago

Hi, I was wondering if there might be an easy way to return the time derivatives for the model / sample, or if I will have to modify the existing code to return this info. Cheers!

amoghpj commented 4 years ago

Hi! This is an interesting question. Currently, BoolODE saves the ODE model definition in model.py (which returns a list of derivatives, the left hand side of the ODEs), and the kinetic parameters in parameters.txt. A simple way to accomplish getting the deterministic time derivatives would be to:

  1. read the parameters and the simulated datasets
  2. import model.py
  3. For each time point in the dataset (a.k.a a single cell), compute the deterministic derivative.

Apart from this, since the derivative calculation is done in boolode.simulator.eulersde(), it would be trivial to write to file the derivatives with the noise term at each step. I am curious as to what you would like to use this for, and which of the above approaches would be more immediately useful for you.

calebclayreagor commented 4 years ago

Thanks for the quick response. I'm working on implementing an inference method that relies heavily on the time derivative. I am hoping that the derivative from BoolODE (either deterministic or stochastic) will be approximately equivalent to the RNA velocity in a real single cell experiment. I'm thinking the stochastic derivative will be more like the velocity, but it may be too noisy for my purposes, in which case it's a moot point. If you have any insight, I would appreciate it!

amoghpj commented 4 years ago

Sorry for the delay, I was hoping to read up some more today: in short, I still don't have a good response to your use case for RNA velocity. The derivatives for the mRNA only give an idealized view of transcription rate, unlike the transcript maturation process modeled in RNA velocity. That said, SERGIO uses essentially the same ODE framework, but I need to spend some time looking at how they generate their RNA velocity datasets. All this is to say that I don't have an informed opinion on how best to generate a ground truth RNA velocity dataset from BoolODE in its correct form. Will read up some more.