search

ajkluber / model_builder

Represents coarse-grain models for easy simulation and parameter fitting.
4 stars 4 forks source link

Combination rule #1

Closed ajkluber closed 9 years ago

ajkluber commented 9 years ago

So should we change the default combination rule to "2" to handle beads with different radii?

https://github.com/ajkluber/model_builder/blob/master/models/CoarseGrainedModel.py#L419-L422

TensorDuck commented 9 years ago

I agree that would be best.

ajkluber commented 9 years ago

Hold up. I hadn't checked the gromacs manual thoroughly enough. Rule 2 doesn't work because it automatically sets the attraction term as well. We have to use rule 1 then.

See section 4.1.1 Gromacs_Manual.pdf

TensorDuck commented 9 years ago

Oh, my bad. I confused the rules in my head. I think we should keep rule 1 and just double all the lengths. That would work reasonably well considering most of the beads are almost the same size anyways.

TensorDuck commented 9 years ago

I've been going over the Cheung and Onuchic 2003 paper again. I noticed that they also have a factor of 0.7 in their repulsive interactions. So for a Ca-Ca interaction, they used a radii of 1.9A for each, and that lead to the excluded volume term being 2.66A, not 0.4A like we've been using. I found that if we use a 0.4A interaction term for repulsion, it actually leads to the proteins "blowing up" very easily.

Consequently, us using half radius inadvertently actually improved matters.

They also used the flavored radii in order to account for sterics better. I don't think it would make a big difference with additive versus multiplicative. But, I was thinking if you wanted to implement the additive term, we can always input a paired interaction into the .top file as a r12 interaction under the pairtypes. I think we can input using func 1 (r12-r6), and just set r6 =0, and compute the r12 manually each time. it will be extra overhead in writing the files but likely won't affect the simulation time as much. I can see about implementing some of this and I'll let you know.

TensorDuck commented 9 years ago

for reference, fixed atomic sizes in dd9ba39ed0bdcae1b7e2aa6b93054c5b098802fc

ajkluber commented 9 years ago

Thanks for this Justin. Following Cheung, et.al. 2003 sounds like the best way to go for the time being, so I will close this issue. I will leave the discussion of robustly testing the CACB model to a future date. Thanks for figuring this out!