Kyle Beauchamp

I agree. My vote is to use `pandas.read_csv` to have our input files be _self-documenting_ csv files.

Yes, we _can_. But this might be a situation where it make sense to use a DataFrame to keep track of data+metadata.

Another thing to add to our Project Builder wish list: currently we use globbing to process FAH style trajectories, which silently allows RUN/CLONE pairs with missing frames. IMHO we should...

PS I started a discussion of analytical test cases here:#243

Agreed. AFAIK, the pathway decomposition is not a physical observable and is therefore not uniquely determined.

Yeah, that part is confusing.

I think this isn't too hard, but it does require us to preselect a reasonable distance or RMSD cutoff heuristic. The challenge here is that discarding data without justification (or...

Yeah, good point we can do that always.

See also issue #117, which asks for single trajectory support.

Robert: I wonder if we should make the library-style usage of MSMBuilder easier. Right now, our Clustering scripts do _everything_, but it's pretty hard to replicate those functions in library...