Changelog#

v0.4.0#

This release includes significant improvements to documentation, as well as a few bug fixes and behavior improvements.

Changed:

  • NetworkPlanners’ mappers attribute is now always represented as a tuple (or None) upon initialization (PR #232).

  • Reorganization and general updates to documentation (PRs #231, #228, #229, #230, #246).

Deprecated:

  • Deprecated konnektor.network_analysis.get_is_connected(). Use the ligand_network’s .is_connected() method instead, which is functionally equivalent (PR #250).

Removed:

  • konnektor.Component, konnektor.ProteinComponent, and konnektor.SmallMoleculeComponent have been removed. Please use gufe.Component, gufe.ProteinComponent, and gufe.SmallMoleculeComponent, which are identical, respectively, instead (PR #225).

Fixed:

  • Fixed bug in NetworkPlanners where passing a generator to the mappers argument would consume the generator upon initialization (PR #232).

  • Fixed API docstrings to accurately reflect behavior (PRs #232, #236, #241, #242, #248).

v0.3.0#

Changed:

  • ExplicitNetworkGenerator no longer uses multiprocessing if n_processes=1 (PR #210).

Deprecated:

  • konnektor.network_tools.concatenate_networks() has been deprecated and will be removed in v0.4.0. Concatenator class methods should be used directly instead. For example, instead of network_tools.concatenate_networks(mst_concatenator, list_of_networks), use mst_concatenator.concatenate_networks(list_of_networks).

Fixed:

  • Scores generated by konnektor.toy_data scorers are now deterministic when called in parallel, meaning that results will be identical whether operations on these datasets are run in serial or in parallel. Note that this was only a bug in these toy datasets, not in the actual konnektor parallelization.

  • Fixed bug where MST concatenator would fail when run interactively, such as in a jupyter notebook (Issue #202, PR #210)

v0.2.1#

Fixed:

  • Fixed bug where konnektor.visualization.draw_ligand_network did not work for networks without scored edges (frequently star/radial or explicit networks) (PR #191).

v0.2.0#

Changed:

  • For both ExplicitNetworkGenerator.generate_network_from_indices() and ExplicitNetworkGenerator.generate_network_from_names(), all components will be included as nodes in the network. If this results in a disconnected network (in the case where one or more of the components is not included in indices, or names, respectively), a warning will be raised (#122).

  • If multiple mappers are passed to a network planner, but no scorer, the first mapping from the first mapper provided will be used. This is a change from the previous behavior, which would use the first mapping from the last mapper. The current behavior is now consistent with openfe's network planning behavior (#84).

  • clustering.auxilliary_featurizer.ChargeTransformer parameter parallel renamed to n_jobs which allows for designating the maximum number of concurrently running jobs (#133).

  • For radial (star) network generation, if central_component is already in the list passed to components, a self-edge will no longer be created, and a warning is raised (#145).

  • konnektor now treats mapping scores of 1.0 as best, and 0.0 as worst. This is an inversion from the prior behavior (0.0 best, 1.0 worst) to match the behavior of the rest of the openfe ecosystem (#138).

  • Renamed RadialLigandNetworkPlanner to RadialNetworkGenerator to be consistent with the other network generator class names (#160).

  • For RedundantMinimalSpanningTreeNetworkGenerator, if the number of redundancies (n_redundancy) is larger than the number of redundant MSTs able to be generated for the network, a warning will be raised, and the resulting network will contain the most redundant MSTs able to be generated (#122).