add isotype to mutation model simulation #83
Conversation
gctree/isotyping.py
Outdated
| def mutate(self, lambda0=0.01): | ||
| """Returns a new isotype object, which may be of the same class, or may be | ||
| a later class. | ||
| """ | ||
| newisotype = self.copy() | ||
| if scipy.random.poisson(lambda0) > 0: | ||
| n = len(self.order) - newisotype.isotype | ||
| if n != 1: | ||
| newisotype.isotype = self.isotype + scipy.random.randint(1, n) | ||
| return newisotype |
There was a problem hiding this comment.
I'm not sure about the Poisson rv here—you are using it to make a binary choice? Shouldn't n be based on the Poisson draw, rather than drawn uniformly (but then truncated based on the terminal isotype)?
I would also suggest documenting lambda0 in the docstring.
It's also implicit here that this function simulates for one unit of simulation time, so I wonder if it would be better to put this code in the mutation_model.MutationModel.simulate function?
There was a problem hiding this comment.
Hmm yes this doesn't make sense... I wrote this deferring thought and then forgot to come back and think about it.
I'm not sure how this should be done, though. I think we talked about fixing a probability of switching occurring and then choosing uniformly from the available states. I'm not sure if we want to choose from the available states according to a poisson process, because it's not clear that one isotype transition is more likely than another.
So I propose changing lambda0 to p, which is the probability that class switching will occur in one time step, and replacing line 168 with a comparison of p to a uniform draw from the unit interval.
|
What I've done instead is add a parameter |
Without otherwise changing the simulation in
mutation_model.MutationModel.simulate, adds isotype switching according to a new rate parameterisotype_lambda.