This is a Rust library for reading, writing, and editing biological taxonomies. There are associated Python bindings for accessing most of the functionality from Python.
This library was developed initially as a component in One Codex's metagenomic classification pipeline before being refactored out, expanded, and open-sourced. It is designed such that it can be used as is with a number of taxonomic formats or the Taxonomy trait it provides can be used to add last common ancestor, traversal, etc. methods to a downstream package's taxonomy implementation.
The library ships with a number of features: - [X] Common support for taxonomy handling across Rust and Python - [X] Fast and low(er) memory usage - [X] NCBI taxonomy, JSON ("tree" and "nodelinkdata" formats), Newick, and PhyloXML support - [X] Easily extensible (in Rust) to support other formats and operations
This library can be added to an existing Cargo.toml file and installed straight from crates.io.
You can install the Python bindings directly from PyPI (binaries are only built for select architectures) with:
bash
pip install taxonomy
The Python taxonomy API can open and manipulate all of the formats from the Rust library. Note that Taxonomy IDs in NCBI format are integers, but they're converted to strings on import. We find working with "string taxonomy IDs" greatly simplifies inter-operation between different taxonomy systems.
Taxonomy can be loaded from a variety of sources.
Taxonomy.from_newick(value: str): loads a Taxonomy from a Newick-encoded string.
Taxonomy.from_ncbi(nodes_path: str, names_path: str): loads a Taxonomy from a pair of NCBI dump files. The paths specified are
to the individual files in the NCBI taxonomy directory (e.g. nodes.dmp and names.dmp).
Taxonomy.from_json(value: str, /, path: List[str]): loads a Taxonomy from a JSON-encoded string. The format can either be
of the tree or nodelinkdata types and will be automatically detected. If path is specified, the JSON will be traversed to that sub-object before being parsed as a taxonomy.
Taxonomy.from_phyloxml(value: &str): loads a Taxonomy from a PhyloXML-encoded string. Experimental
Assuming that the taxonomy has been instantiated as a variable named tax.
tax.to_newick(): exports a Taxonomy as a Newick-encoded byte string.
tax.to_json(/, as_node_link_data: bool): exports a Taxonomy as a JSON-encoded byte string. By default, the JSON format
is a tree format unless the as_node_link_data parameter is set to True.
Assuming that the taxonomy has been instantiated as a variable named tax. Note that TaxonomyNode is a class with
the following schema:
python
class TaxonomyNode:
id: str
name: str
parent: Optional[str]
rank: str
Note that tax_id in parameters passed in functions described below are string but for example in the case of NCBI need
to be essentially quoting integers: 562 -> "562". In that case, passing something that can't be converted to a number
will raise an exception even if the documentation below does not mention it.
tax.root -> TaxonomyNodePoints to the root of the taxonomy
tax.parent(tax_id: str, /, at_rank: str) -> Optional[TaxonomyNode]Return the immediate parent TaxonomyNode of the node id.
If at_rank is provided, scan all the nodes in the node's lineage and return
the parent id at that rank.
Examples:
```py parent = tax.parent("612") parent = tax.parent("612", at_rank="species") parent = tax.parent("612")
None if we can't find the parentparent = tax.parent("unknown") ```
tax.parent_with_distance(tax_id: str, /, at_rank: str) -> (Optional[TaxonomyNode], Optional[float])Same as parent but return the distance in addition, as a (TaxonomyNode, float) tuple.
tax.node(tax_id: str) -> Optional[TaxonomyNode]Returns the node at that id. Returns None if not found.
You can also use indexing to accomplish that: tax["some_id"] but this will raise an exception if the node
is not found.
tax.find_by_name(name: str) -> Optional[TaxonomyNode]Returns the node with that name. Returns None if not found.
In NCBI, it only accounts for scientific names and not synonyms.
tax.children(tax_id: str) -> List[TaxonomyNode]Returns all nodes below the given tax id.
tax.lineage(tax_id: str) -> List[TaxonomyNode]Returns all nodes above the given tax id, including itself.
tax.parents(tax_id: str) -> List[TaxonomyNode]Returns all nodes above the given tax id.
tax.lca(id1: str, id2: str) -> Optional[TaxonomyNode]Returns the lowest common ancestor for the 2 given nodes.
tax.prune(keep: List[str], remove: List[str])-> TaxonomyReturn a copy of the taxonomy containing:
keep and their parents if providedtax.remove_node(tax_id: str)Remove the node from the tree, re-attaching parents as needed: only a single node is removed.
tax.add_node(parent_tax_id: str, new_tax_id: str)Add a new node to the tree at the parent provided.
edit_node(tax_id: str, /, name: str, rank: str, parent_id: str, parent_dist: float)Edit properties on a taxonomy node.
Only one exception is raised intentionally by the library: TaxonomyError.
If you get a pyo3_runtime.PanicException (or anything with pyo3 in its name), this is a bug in the underlying Rust library, please open an issue.
There is a test suite runable with cargo test. To test the Python-bindings you need to use the additional python_test feature: cargo test --features python_test.
To work on the Python library on a Mac OS X/Unix system (requires Python 3): ```bash
curl https://sh.rustup.rs -sSf | sh rustup default nightly
maturin develop --cargo-extra-args="--features=python"
pip install . ```
```
maturin build --cargo-extra-args="--features=python" --release --strip
docker run --rm -v $(pwd):/io konstin2/maturin:master build --cargo-extra-args="--features=python" --release --strip twine upload target/wheels/* ```
There are taxonomic toolkits for other programming languages that offer different features and provided some inspiration for this library:
ETE Toolkit (http://etetoolkit.org/) A Python taxonomy library
Taxize (https://ropensci.github.io/taxize-book/) An R toolkit for working with taxonomic data