Releases: matsengrp/gctree
Releases · matsengrp/gctree
v4.0.3
v4.0.2
v4.0.1
What's Changed
- Update installation docs by @willdumm in #80
- add potential Dockerfile + test.sh by @psathyrella in #81
- Address Duncan's suggestions by @willdumm in #85
- Fix multiple tree nodes with same name by @willdumm in #86
New Contributors
- @psathyrella made their first contribution in #81
Full Changelog: v4.0.0...v4.0.1
v4.0.0
Summary
- A major enhancement for parsimony tree search is the use of a history DAG to expand the set of trees found by dnapars, often by orders of magnitude. h/t @willdumm
- Tree ranking can now use a linear combination of branching process likelihood, isotype parsimony, and mutability parsimony (e.g. from a 5mer context model) to score trees.
Backwards-incompatible changes
- The
CollapsedForest
class no longer has atrees
attribute, as the collection of trees is now represented as a history DAG. - The gctree CLI can now accept pickled parsimony forests from the current gctree version, but not from older versions.
- Only the unique optimal tree is output by
gctree infer
, if it exists. If more than one tree is optimal, up to 10 will be output arbitrarily, and their numerical rankings will be arbitrary (rather than based on likelihood as before).
Pull requests
- Fit parameters on history DAG by @willdumm in #68
- Add ambiguous sequence test and error message by @willdumm in #72
- Transplant docs by @wsdewitt in #73
- Transplant patch by @wsdewitt in #74
- Docs fixes by @wsdewitt in #76
- Update README.md by @wsdewitt in #77
- Store isotype data in nodes by @willdumm in #75
- Mutability chain split by @willdumm in #78
- Prerelease tidying by @wsdewitt in #79
Full Changelog: v3.3.0...v4.0.0
v3.3.0
What's Changed
- Add isotype utility by @willdumm in #65. See new sections in API docs and Quickstart.
- A new docs page demoing tree visualization by @wsdewitt in #69
- LB metrics for CollapsedTree with new method, by @wsdewitt in #69
- New method for generating a colormap from any numerical node feature in a CollapsedTree, by @wsdewitt in #69
- Sequence-based fitness and seeding with more than one cell for simulation in
mutation_model.MutationModel
by @wsdewitt in #69
Full Changelog: v3.2.1...v3.3.0
v3.2.1: Merge pull request #57 from matsengrp/56-disambiguate-optimally
Patches a problem with resolution of ambiguous DNA characters in dnapars
output.
h/t @willdumm
v3.2.0: Merge pull request #55 from matsengrp/render-options
New API tree rendering options scale
, branch_margin
, and node_size
described in docs.
v3.1.1: Merge pull request #54 from matsengrp/mlepatch
Patches
- Fixed a bug in empirical Bayes estimation of branching process parameters for
CollapsedForest
. Previous version v3.1.0 may have resulted in incorrect parsimony tree ranking CollapsedTree.render()
can now be used in a Jupyter notebook for tree visualization- Parameter checking for more flexible branching process simulations
v3.1.0
New features
- The CLI program
gctree infer
writes a pickledCollapsedTree
object for each parsimony tree, which can be later loaded and manipulated via the API. - Updates to
CollapsedTree.render
to annotate amino acid substitutions, and associated new parameters in API and CLI. Note: these parameters are not supported by the scons pipelines, which are deprecated. - Classes
CollapsedTree
andCollapsedForest
are exposed at package level, so they can be imported directly rather than from thebranching_processes
module. - Many improvements in documentation: matsengrp.github.io/gctree
Developer notes
- The
branching_processes
module has been refactored to remove illegitimate inheritance - Better default
p, q
parameters for MLE avoid gradient warnings.