Skip to content

Latest commit

 

History

History
73 lines (66 loc) · 5.68 KB

CHANGELOG.md

File metadata and controls

73 lines (66 loc) · 5.68 KB
Raw

===========

DEVELOPMENT

===========

Once I have more time and create soothsayer v1.0 then I will make all of this formatted consistently

========

Changes:

========

  • 2023.12.22 - Fixed critical error in Agglomerative that gave incorrect leaf names. It's because the disismilarity matrix was converted to a Symmetric object but the node labels changed order.
  • 2023.8.22 - Changed condensed_to_dense to condensed_to_redundant and dense_to_condensed to redundant_to_condensed. Also changed cluster_homogeneity to edge_cluster_cooccurrence
  • 2022.4.4 - Fixed the edges - edgelist in networkx draw_networkx_edges function
  • 2022.2.14 - Added protest method to Procrustes and cleaned up plotting function. Fixed ha and horzintalalignment double call in Agglomerative
  • 2022.02.06 - Added Procrustes to ordination and replaced .as_matrix() calls with .values
  • 2022.01.07 - Added get_mixed_feature_pipeline_estimator and added axis argument to pd_prepend_level_to_index
  • 2021.07.27 - Added plot_multiindexed_heatmap and iterate_nested_dict
  • 2021.05.19 - Added intersection and union functions from soothsayer_utils. Also added 2 wrappers around scipy.stats.entropy (shannon_entropy and kullback_leibler_divergence) and a wrapper around scipy.stats.ttest_ind (welchs_ttest)
  • 2021.04.27 - Reimplemented normalize_css in Python to remove dependency on metagenomeSeq. Note, metagenomeSeq is confusing because it uses p for quantile which sounds like it's referring to percentiles. To make it more consistent with numpy and scipy I switched p to q. Also, I removed the propr dependency because it's not used at all now that I've made the compositional Python package.
  • 2021.04.12 - Changed cluster_modularity to cluster_homogeneity in EnsembleNetworkX and also added community_detection
  • 2021.03.22 - Deprecated some KEGG databases because versions couldn't be confirmed. Added a read_kegg_json in soothsayer.io. Fixed an error with the MANIFEST.in not including sub-directories in db and non-.pbz2 files such as the .flat and .json files
  • 2021.03.05 - Added edgeR's exactTest, edgeR's glmLRT, and fixed np.all incompatibilities (pandas-dev/pandas#40259)
  • 2021.01.21 - Added markerscale=2 to LEGEND_KWS
  • 2021.01.13 - FINALLY removed WGCNA dependence. I have recoded all of the necessary functions in Python. This will make installation MUCH easier. New functions include: density, heterogeneity, centralization, scalefree_topology_estimation, and an updated determine_soft_threshold. I've been trying to do this for years...but never fully committed.
  • 2021.01.06 - Added differential_abundance to statistics using ALDEx2 in r_wrappers
  • 2020.12.16 - Fixed a ddof=0 and ddof=1 incompatibility. Now all ddof=1 in compositional package.
  • 2020.09.17 - Fixed LinAlgError in plot_compositional from seaborn update to v0.11
  • 2020.07.21 - Reimplemented PrincipalCoordinatesAnalysis and PrincipalComponentsAnalysis classes (and base classes). They now have the ability to plot biplots, get loadings, and get contribution of feature classes to each PC.
  • 2020.07.21 - Added add_objects_to_globals function to add objects from accessory packages like hive_networkx, ensemble_networkx, soothsayer_utils, and compositional
  • 2020.07.01 - Created EnsembleAssociationNetwork class
  • 2020.06.15-30 Make hive_networkx its own package. Moved Symmetric and Hive (along with other functions)
  • 2020.05.04 - Adapted code to support py3.7/8 and rpy2 v3.x. I only ran a few tests so hopefullyt his doesn't break
  • 2020.05.20 - Fixed nx.add_path deprecation for Hierarchical Classifier
  • 2020.05.20 - Created "infer_tree_type", "check_polytomy", "prune_tree", and "is_leaf" in utils.
  • 2020.05.20 - Changed "name_ete_nodes to "name_tree_nodes" (made skbio compatible) and moved to utils
  • 2020.05.20 - Made dependency on compositional. Replaced transform functions in soothsayer.transmute.normalization and added functions in soothsayer.symmetry.
  • 2020.05.20 - Removed automatic maximum threads for WGCNA
  • 2020.06.01 - Removed dependencies on R packages but rpy2 is still necessary
  • 2020.06.04 - Added "component_type" to alpha_diversity

Bugs:

=====

  • 2019.12.06 For some reason adjusting something in regression broke the ability to do sy.r_wrappers.philr.normalize_philr but you can still do from soothsayer.r_wrappers import philr

===========

future:

===========

Clairvoyance

------------

  • Remove name in output files [Fixed v2022.01.06]
  • Add weights in the synopsis file [Fixed v2022.01.06]
  • Remove feature type from synopsis names [Fixed v2022.01.06]
  • Change attributes to features [Fixed v2022.01.06]
  • Write synopsis file after each model_type is finished [Fixed v2022.01.06]
  • Output only the best models [Fixed v2022.01.06]
  • For normalization, normalize based on the original data not the data for each iteration. For example, this wouldn't make sense for CLR. [Fixed v2022.01.06]
  • Remove all instances of --encoding
  • Make --save_kernel, save_model, etc. flags
  • Incorporate elasticnet for LogisticRegression

Soothsayer

----------

  • Remove signed from pairwise and clean up function
  • Make Agglomerative its own package and use skbio instead of Ete3
  • Make HierarchicalClassifier its own package and use skbio instead of Ete3
  • Change create_pca_model to something else...
  • Collapse topology so it uses Agglomerative instead
  • Clean up the coloring (map_colors, scalarmapping_from_data, etc.)
  • Move TemporalNetwork to Ensemble NetworkX
  • Redo Dataset object and make a .to_anndata method
  • Add subset to orindation plot to only some of the points. useful for different markers.
  • Autodetect multiheader for read_dataframe and clean it up