• Add RMSF calculation along a single trajectory to INDUSAnalysis package
• Average RMSF across replicas and color PDB by bfactor for each value of phi
• See differences in regional fluctuations at equilibrium and at low values of phi, then at high values of phi (indication of changes in flexibility at different bias potentials, better indicator than individual atom RMSD!!!!)

• Migrate to using pandas for handling timeseries data instead of just numpy arrays
• Use HDF5 format for data storage and retrieval with pandas

• Add feature
• Change data structures and data access patterns for performance speedup
• Maintain backwards compatibility, by saving sum of contacts with same name, but saving individual contacts with different name.

Avoid naming conflicts with pip.

Not urgent - only before first release version.

Use cdef and cpdef to speed up function performance.

• Perform PCA
• Project all frames after burn-in time onto top-k principal components, compute centroid
• Use either (a) coordinates reconstructed from centroid PCs, or (b) coordinates from frame closest to centroid PC as representative dynamic image

• Fix loadTimeSeriesFromDAT()
• Evaluate labeling options

Add a module for performing phi-ensemble reweighting using free energy profiles obtained from WHAM.

Support directly interfacing with WHAM.

Change "type H" to "name H*"

• Review protein_order_params in particular

Clean and upload private scripts after relevant papers are published.

Implement error estimation using block bootstrapping with sample size = N/g, given g as input/option to calculate using statsmodels. Write tests to compare with pymbar statistical inefficiency.

Cythonize code for performance improvements:

• Basic cythonize
  -- Unit testing, integration testing
  -- Update git workflow
  -- Code profiling
• Add type definitions
  -- Unit testing, integration testing