TaxonoPy

TaxonoPy (taxon-o-py) is a command-line tool for creating an internally consistent taxonomic hierarchy using the Global Names Verifier (gnverifier). See below for the structure of inputs and outputs.

Purpose

The motivation for this package is to create an internally consistent and standardized classification set for organisms in a large biodiversity dataset composed from different data providers that may use very similar and overlapping but not identical taxonomic hierarchies.

Its development has been driven by its application in the TreeOfLife-200M (TOL) dataset. This dataset contains over 200 million samples of organisms from four core data providers:

• The Global Biodiversity Information Facility (GBIF)
• BIOSCAN-5M
• FathomNet
• The Encyclopedia of Life (EOL)

The names (and classification) of taxa may be (and often are) inconsistent across these resources. This package addresses this problem by creating an internally consistent classification set for such taxa.

Input

A directory containing Parquet partitions of the seven-rank Linnaean taxonomic metadata for organisms in the dataset. Labels should include:

• uuid: a unique identifier for each sample (required).
• kingdom, phylum, class, order, family, genus, species: the taxonomic ranks of the organism (required, may have sparsity).
• scientific_name: the scientific name of the organism, to the most specific rank available (optional).
• common_name: the common (i.e. vernacular) name of the organism (optional).

See the example data in

• examples/input/sample.parquet
• examples/resolved/sample.resolved.parquet (generated with taxonopy resolve)
• examples/resolved_with_common_names/sample.resolved.parquet (generated with taxonopy common-names)

Challenges

This taxonomy information is provided by each data provider and the original sources, but the classification can be...

• Inconsistent: both between and within sources (e.g. kingdom Metazoa vs. Animalia).
• Incomplete: many samples are missing one or more ranks. Some have 'holes' where higher and lower ranks are present, but intermediate ranks are missing.
• Incorrect: some samples have incorrect classifications. This can come in the form of spelling errors, nonstandard ideosyncratic terms, or outdated classifications.
• Ambiguous: homonyms, synonyms, and other terms that can be interpreted in multiple ways unless handled systematically.

Taxonomic authorities exist to standardize classification, but ...

• There are many authorities.
• They may disagree.
• A given organism may be missing from some.

Solution

TaxonoPy uses the taxonomic hierarchies provided by the TOL core data providers to query GNVerifier and create a standardized classification for each sample in the TOL dataset. It prioritizes the GBIF Backbone Taxonomy, since this represents the largest part of the TOL dataset. Where GBIF misses, backup sources such as the Catalogue of Life and Open Tree of Life (OTOL) Reference Taxonomy are used.

Installation

TaxonoPy can be installed with pip after setting up a virtual environment.

User Installation with pip

To install the latest version of TaxonoPy, run:

pip install taxonopy

Usage

You may view the help for the command line interface by running:

taxonopy --help

This will show you the available commands and options:

usage: taxonopy [-h] [--cache-dir CACHE_DIR] [--cache-input CACHE_INPUT]
                [--show-cache-path] [--cache-stats] [--clear-cache]
                [--show-config] [--version]
                {resolve,trace,common-names} ...

TaxonoPy: Resolve taxonomic names using GNVerifier and trace data provenance.

positional arguments:
  {resolve,trace,common-names}
    resolve             Run the taxonomic resolution workflow
    trace               Trace data provenance of TaxonoPy objects
    common-names        Merge vernacular names (post-process) into resolved outputs

options:
  -h, --help            show this help message and exit
  --cache-dir CACHE_DIR
                        Directory for TaxonoPy cache (can also be set with TAXONOPY_CACHE_DIR environment variable) (default: None)
  --cache-input CACHE_INPUT
                        Input dataset path to compute cache stats for when no command is provided (default: None)
  --show-cache-path     Display the current cache directory path and exit (default: False)
  --cache-stats         Display statistics about the cache and exit (default: False)
  --clear-cache         Clear the TaxonoPy object cache. May be used in isolation. (default: False)
  --show-config         Show current configuration and exit (default: False)
  --version             Show version number and exit

Cache behavior

taxonopy resolve caches parsed entries, entry groups, and every resolution attempt chain using diskcache as a stable provenance artifact tied to the TaxonoPy version and input dataset. By default the cache root is ~/.cache/taxonopy, but you can override it by setting the environment variable TAXONOPY_CACHE_DIR or specifying --cache-dir. Its primary purpose is to support the trace command, which allows you to trace the provenance of any taxonomic entry resolved by TaxonoPy.

• Each resolve run writes into resolve_v<version>_<fingerprint> where the fingerprint is a SHA-256 hash of the input files’ metadata, so namespaces stay stable per combination of dataset and package version.
• Inspect a namespace without rerunning by invoking taxonopy --cache-dir <root> --cache-input <input> --cache-stats, which reports total size, entry counts, and key-prefix breakdowns. Passing --cache-stats after resolve or trace performs the same check and exits.
• If both the namespace and the output directory already contain data, taxonopy resolve warns and exits unless you pass --full-rerun, which clears the cache namespace and output before proceeding. Use --clear-cache to wipe only the namespace.

Command: resolve

The resolve command is used to perform taxonomic resolution on a dataset. It takes a directory of Parquet partitions as input and outputs a directory of resolved Parquet partitions.

usage: taxonopy resolve [-h] -i INPUT -o OUTPUT_DIR
                        [--output-format {csv,parquet}]
                        [--log-level {DEBUG,INFO,WARNING,ERROR,CRITICAL}]
                        [--log-file LOG_FILE] [--force-input] [--full-rerun]
                        [--batch-size BATCH_SIZE] [--all-matches]
                        [--capitalize] [--fuzzy-uninomial] [--fuzzy-relaxed]
                        [--species-group] [--refresh-cache] [--cache-stats]

options:
  -h, --help            show this help message and exit
  -i, --input INPUT     Path to input Parquet or CSV file/directory
  -o, --output-dir OUTPUT_DIR
                        Directory to save resolved and unsolved output files
  --output-format {csv,parquet}
                        Output file format
  --log-level {DEBUG,INFO,WARNING,ERROR,CRITICAL}
                        Set logging level
  --log-file LOG_FILE   Optional file to write logs to
  --force-input         Force use of input metadata without resolution
  --full-rerun          Replace existing cache/output if detected for this input

GNVerifier Settings:
  --batch-size BATCH_SIZE
                        Max number of name queries per GNVerifier API/subprocess call
  --all-matches         Return all matches instead of just the best one
  --capitalize          Capitalize the first letter of each name
  --fuzzy-uninomial     Enable fuzzy matching for uninomial names
  --fuzzy-relaxed       Relax fuzzy matching criteria
  --species-group       Enable group species matching

Cache Management:
  --refresh-cache       Force refresh of cached objects (input parsing, grouping) before running.
  --cache-stats         Display cache statistics for this input and exit.

It is recommended to keep GNVerifier settings at their defaults.

Command: trace

The trace command is used to trace the provenance of a taxonomic entry. It takes a UUID and an input path as arguments and outputs the full path of the entry through TaxonoPy.

usage: taxonopy trace [-h] {entry} ...

positional arguments:
  {entry}
    entry     Trace an individual taxonomic entry by UUID

options:
  -h, --help  show this help message and exit

usage: taxonopy trace entry [-h] --uuid UUID --from-input FROM_INPUT [--format {json,text}] [--verbose]

options:
  -h, --help            show this help message and exit
  --uuid UUID           UUID of the taxonomic entry
  --from-input FROM_INPUT
                        Path to the original input dataset
  --format {json,text}  Output format
  --verbose             Show full details including all UUIDs in group

Command: common-names

The common-names command is used to merge vernacular names into the resolved output. It takes a directory of resolved Parquet partitions as input and outputs a directory of resolved Parquet partitions with common names.

usage: taxonopy common-names [-h] --resolved-dir ANNOTATION_DIR --output-dir OUTPUT_DIR

options:
  -h, --help            show this help message and exit
  --resolved-dir ANNOTATION_DIR
                        Directory containing your *.resolved.parquet files
  --output-dir OUTPUT_DIR
                        Directory to write annotated .parquet files

Note that the common-names command is a post-processing step and should be run after the resolve command.

Example Usage

To perform taxonomic resolution on a dataset with subsequent common name annotation, run:

taxonopy resolve \
    --input /path/to/formatted/input \
    --output-dir /path/to/resolved/output

taxonopy common-names \
    --resolved-dir /path/to/resolved/output \
    --output-dir /path/to/resolved_with_common-names/output

Development

See the Wiki Development Page for development instructions.