search

choderalab / bayesian-itc

Python tools for the analysis and modeling of isothermal titration calorimetry (ITC) experiments.
GNU General Public License v3.0
5 stars 10 forks source link

[WIP] Fixing the competitive binding model #46

Closed bas-rustenburg closed 9 years ago

bas-rustenburg commented 9 years ago

Work in progress, not ready to merge yet.

bas-rustenburg commented 9 years ago

13

bas-rustenburg commented 9 years ago

Various different information sources are needed to build a model. If we are working with literature data, I think we need a well-structured file that contains the necessary information. I decided to go with a yaml file as a prototype, see 09d926d

bas-rustenburg commented 9 years ago

Reading of the YAML now works for TwoComponent. Made some changes to the TwoComponentBindingModel and the ExperimentDotITC classes as well. Nothing has changed to the interface itself though.

Next step is to pass on input to the CompetitiveBindingModel. I will have to modify the command line app and parser (scripts/bitc_util.py and bitc/parser.py) to get started with this.

Right now, the app works simply by specifying a file with a .yml extension, or .itc extension and it figures it out from there (haven't put .dat support in yet). I might just go for N positional arguments, being files of different type and extension.

To make it easier to use, I will make the workdir an optional flag (default directory being ./).

bas-rustenburg commented 9 years ago

Almost there. Some of the command line options (names, heats files), are not working at the moment because I changed the utility structure. Will have to reimplement those later.

I have updated the structure of the Competitive model, but there are still a few bugs in the __init__ function that I haven't figured out yet.

bas-rustenburg commented 9 years ago

There might still be some bugs floating around, but model construction works without bugging out. The MAP fit runs to completion ( for a small number of steps). Because it is so slow, I haven't been able to verify if the MCMC works correctly, but it runs for a decent number of steps without crashing. Probably will merge soon.

bas-rustenburg commented 9 years ago

Can't guarantee that everything works, and will probably want to work on the plotting of the competitive model later. Merging so that @daveminh et al. can get started with the racemic model.

Also, incredibly slow at the moment... will optimize code in separate PR.