Metabolike

What's Changed
* Route search functions by y1zhou in https://github.com/y1zhou/metabolike/pull/34
* Get relevant information from the Neo4j database
* Route search starting from a given Compound node
* Map gene expression (or differential expression) data to the graph, and aggregate to reaction-level metrics
* Support end metabolite structure similarity calculation
* A simple streamlit app for easier queries

**Full Changelog**: https://github.com/y1zhou/metabolike/compare/v0.4.1...v0.5.0

To test this version, run:

sh
pip install -U -i https://test.pypi.org/simple/ metabolike==0.4.1


This is a maintenance release that focus on testing. Turns out that's something you should have 😄. The core parsing code is now unit-tested.

What's Changed
* Test all parser code by y1zhou in https://github.com/y1zhou/metabolike/pull/33
* fix: remove quoted compound IDs in pathway links.
* refactor: put all database methods into *_to_graph() methods.

**Full Changelog**: https://github.com/y1zhou/metabolike/compare/v0.4.0...v0.4.1

Release `v0.4.0` outperforms previous releases by **a lot**. With tests on my local server, parsing HumanCyc takes ~2min (previously ~7min). Parsing the entire MetaCyc, which used to be something I run overnight, now takes ~1h. This is achieved by [batch updating nodes in the database](https://medium.com/neo4j/5-tips-tricks-for-fast-batched-updates-of-graph-structures-with-neo4j-and-cypher-73c7f693c8cc).

What's New
* perf: Submit queries with batches of nodes by y1zhou in https://github.com/y1zhou/metabolike/pull/32
* fix: `metaId` property for composite reaction nodes were `name`s, and not following the SBML specification.
* feat: Progress bars when submitting the queries.
* reafactor: the parser and database classes are mostly separated. This makes testing parser code much easier.


**Full Changelog**: https://github.com/y1zhou/metabolike/compare/v0.3.3...v0.4.0

Breaking changes
* `displayName` changed to `name`, and `mcId` changed to `metaId`.
* All relationships pointing to `RDF` nodes now have type `hasRDF`. This makes querying RDF entities more consistent. The bio-qualifier terms (which used to be the relationship types) are now under the `bioQualifier` property of the relationship.

What's New
* feat: Link reaction nodes connecting related pathways by y1zhou in https://github.com/y1zhou/metabolike/pull/28
* refactor: BRENDA-related code by y1zhou in https://github.com/y1zhou/metabolike/pull/29
* feat: FBA analysis and major refactor by y1zhou in https://github.com/y1zhou/metabolike/pull/30

**Full Changelog**: https://github.com/y1zhou/metabolike/compare/v0.3.2...v0.3.3

Breaking changes
* Support for Python 3.6 is dropped as it's EOL in December 2021. Now Python>=3.8 is required because of the use of [walrus operators](https://docs.python.org/3/whatsnew/3.8.html#assignment-expressions).
* SBML parsing and database-related methods are extracted to their own classes.
* All `hasLeft` relationships point to reactants, and all `hasRight` relationships point to products of reactions. There **might be a mismatch between reaction Gibbs 0 and direction for reversible `Reaction` nodes**.
* Name of the database is now part of the config file. It's no longer needed to be passed to the CLI function `setup()`.

Dependency changes
* `pyarrow` is no longer required as parsed BRENDA files are now serialized to a JSON file instead of a parquet file.
* `orjson` is an added dependency for faster (de)serialization.
* `pandas>=1.0.0` is required because of the use of `pd.NA`.
* `neo4j>=4.3.4` is required as recommended in [this article](https://neo4j.com/developer-blog/neo4j-driver-best-practices/).

What's Changed
* Normalize the direction of reaction nodes by y1zhou in https://github.com/y1zhou/metabolike/pull/26

BREAKING CHANGE: Removes the property from reaction nodes.

**Full Changelog**: https://github.com/y1zhou/metabolike/compare/v0.3.1...v0.3.2

This is mainly a bug fix version, also testing the auto-release to PyPI from GitHub CI.

What's Changed

- build: now uses [flit](https://flit.readthedocs.io/en/latest/) to build the package.
- fix: metacyc reaction nodes could have multiple EC numbers https://github.com/y1zhou/metabolike/commit/1fb00300fca3f17e0a49589012009a90a6b2ce00
- feat: reaction nodes are now linked to compartment nodes https://github.com/y1zhou/metabolike/commit/66ee744c4b6a39d40fcdf196a1fec6fab4683066
- fix: canonical ID in reaction nodes are now back https://github.com/y1zhou/metabolike/commit/4684a0bcdad7cbadba72733da6b3d20dfbf26cc5

**Full Changelog**: https://github.com/y1zhou/metabolike/compare/v0.3.0...v0.3.1