Sample and background weighting, source target flow sets separation, support for sampling and flow calculation pair generation policies

This major release introduces sample weighting and source/target sample.

• The sample to analyze can now be weighted, reflecting confidence in its involvement in a phenotype of interest
• It is possible to now supply a second set to analyze and bioflow will try to most likely connection between the two sets. This is useful in cases where two separate sets of interest have been identified, but the connection between them is unclear. For instance one is experimentally discovered hits, the other one are known effectors. Right now, the secondary set has to be disjoint from the primary set
• It is possible to now supply a weighted background set, and the nodes will be selected preferentially according to it. Thanks to that some nodes can be excluded from sampling or a more realistic background model introduced.
• BioFlow now supports the sampling policy, that can be assigned during the sampling process (including the auto-analyze) as well as flow calculation policy that generates the pairs of nodes that will be used to build the flow.

Python 3, main loop optimization, configs sanification, io sanification, main db to laplacian logic, sampling and background logic

This is a squashed version bump for three major releases:

• v.0.3.0: Python 2 to Python 3 transition.
  • Due to change in multiprocessing and memoization behavior, it gave rise to memory leaks
  • In turn, their elimination led to a profiling pass that resulted a 20x speed-up in the main sampling loop
• v.0.4.0: Added sanity to the configurations and the IO.
  • User-facing files now have a home, living in either $BIOFLOWHOME or in $HOME/bioflow
  • It has the source database downloads, logs, and outputs
  • As well as all the files used by docker deployments
• v.0.5.0: New parsing logic to build the Annotome and Interactome Laplacians, allowing for custom weighting functions
  • A more efficient and complete entity graph traversal made possible by the database interface update
  • This lead to >30x performance improvement for the Laplacian build and a much simpler logic, removing several methods with cyclomatic complexity in the 20s and removal of a bunch of unnecessary class-wide variables.
  • Support for custom-build functions implementing different weighting strategies based on the node parameters, fully customizable
  • Significance evaluation routines are now pluggable as well
  • Background ceased being stateful and is now only used for sampling.

Added complexes to the database, rebuilt a modern Neo4j interface and significance calculation engine

• Added IntAct complex database as a data source
• Modernized Neo4j DB interface.
• Unsupported Java version, bulbs library, and neo4j versions are no more needed.
• Removed dependency on 14.04 or earlier Ubuntu builds.
• Added Gene Identifier overlap conflict resolution
• Corrected a Tax ID parsing bug in Uniprot parser
• Corrected a GeneName parsing bug in Uniprot parser
• Corrected a Cross-Reference tag parsing bug in Uniprot parser
• Updated dockerfile
• Created docker compose file
• Significance calculation now uses Extreme Limit theory (Gumbel Distribution)
• published on Dockerhub
• published on pypi

Added TFs and kinases into the parsing base

• Added Transcription import to the master database
• Added PhosphoCite import to the master database (requires manual download now though)
• Corrected a bug in the UniProt parsing leading to some genes names missing upon matches.

Locally stable and tested interactome analysis

• Added automated testing and code review
• Refactored parsing/DB insertion module
• Properly debugged the command line
• Simplified the main running environment
• Improved runtime estimation
• patched memory leakage
• created docker containers based on the Ubuntu 14.04 LTS
• debugged linear algebra current loss
• added line loss, laplacian normalization and edge/hit drop-out

v0.02: First version with proper documentation and command-line interface

In this release, a full API documentation was build, setup file has been created, enabling a pip install git+http://...., database file download and a command user interface for most frequent uses.

version 0.0-alpha.1

Initial release aimed to provide a minimal usability for the non-developers. The project is in development and is likely to change radically in the near-future.