improved hierarchical model

This release includes updates to the posterior sampling and summarization code. It also switches to a more consistent run interface, switching from number of iterations to number of epochs for SVI and MAP inference.

treats multiple plasmid transformations

• Added correction to inferred theta values accounting for the fact cells can pick up more than one plasmid
• Refactored the util submodule from flat to hierarchical
• Tweaks to increase speed of inference
• Updated command line interface to analyze_theta to interface with the model changes
• Lots of docstring clean up
• Added a bunch of unit tests
• Various bug fixes

hierarchical modeling

• Hierarchical modeling analysis now complete.
• Cleaned up analysis and demo notebooks.

v0.2.0: Merge pull request #4 from harmsm/main

Code now does:

• Simulation of experiments
• Processing raw reads (fastq through ln_cfu/condition)
• Independent modeling (extract that for each genotype independently)
• Hierarchical modeling (Bayesian analysis of theta together)
• Epistasis analysis
• Epistasis vs. [titrant] categorization analysis
• Various helpful plotting utilities.

Additionally, the API is now (roughly) stable. I antiquate no major changes to the simulation, raw-processing, or independent modeling code base. I expect to build out the hierarchical modeling codebase and clean up the utilities functions. None of these will be architectural changes, however, just minor refactors.

initial release

This is a first draft of the code that allows simulation and analysis of a transcription factor high-throughput screen. Expect large changes to the API.