In cboettig/taxadb: A High-Performance Local Taxonomic Database Interface

knitr::opts_chunk$set(
  collapse = TRUE,
  warning = TRUE,
  comment = "#>"
)
taxadb:::td_disconnect()

taxadb

The goal of taxadb is to provide fast, consistent access to taxonomic data, supporting common tasks such as resolving taxonomic names to identifiers, looking up higher classification ranks of given species, or returning a list of all species below a given rank. These tasks are particularly common when synthesizing data across large species assemblies, such as combining occurrence records with trait records.

Existing approaches to these problems typically rely on web APIs, which can make them impractical for work with large numbers of species or in more complex pipelines. Queries and returned formats also differ across the different taxonomic authorities, making tasks that query multiple authorities particularly complex. taxadb creates a local database of most readily available taxonomic authorities, each of which is transformed into consistent, standard, and researcher-friendly tabular formats.

Install and initial setup

To get started, install from CRAN

install.packages("taxadb")

or install the development version directly from GitHub:

devtools::install_github("ropensci/taxadb")

library(taxadb)
library(dplyr) # Used to illustrate how a typical workflow combines nicely with `dplyr`

Create a local copy of the (current) Catalogue of Life database:

td_create("col")

Read in the species list used by the Breeding Bird Survey:

bbs_species_list <- system.file("extdata/bbs.tsv", package="taxadb")
bbs <- read.delim(bbs_species_list)

Getting names and ids

Two core functions are get_ids() and get_names(). These functions take a vector of names or ids (respectively), and return a vector of ids or names (respectively). For instance, we can use this to attempt to resolve all the bird names in the Breeding Bird Survey against the Catalogue of Life:

birds <- bbs %>% 
  select(species) %>% 
  mutate(id = get_ids(species, "col"))

head(birds, 10)

Note that some names cannot be resolved to an identifier. This can occur because of miss-spellings, non-standard formatting, or the use of a synonym not recognized by the naming provider. Names that cannot be uniquely resolved because they are known synonyms of multiple different species will also return NA. The filter_name filtering functions can help us resolve this last case (see below).

get_ids() returns the IDs of accepted names, that is dwc:AcceptedNameUsageIDs. We can resolve the IDs into accepted names:

birds %>% 
  mutate(accepted_name = get_names(id, "col")) %>% 
  head()

This illustrates that some of our names, e.g. Dendrocygna bicolor are accepted in the Catalogue of Life, while others, Anser canagicus are known synonyms of a different accepted name: Chen canagica. Resolving synonyms and accepted names to identifiers helps us avoid the possible miss-matches we could have when the same species is known by two different names.

Taxonomic Data Tables

Local access to taxonomic data tables lets us do much more than look up names and ids. A family of filter_* functions in taxadb help us work directly with subsets of the taxonomic data. As we noted above, this can be useful in resolving certain ambiguous names.

For instance, Agrostis caespitosa does not resolve to an identifier in ITIS:

get_ids("Agrostis caespitosa", "itis") 

Using filter_name(), we find this is because the name resolves not to zero matches, but is a known synonym to more than one accepted name (as indicated by the accepted name usage id)

filter_name('Agrostis caespitosa', 'itis')

We can resolve the scientific name to the acceptedNameUsage using get_names() on the accepted IDs: (These also correspond to the genus and specificEpithet column, as the classification is always given only based on acceptedNameUsageID).

filter_name("Agrostis caespitosa")  %>%
  mutate(acceptedNameUsage = get_names(acceptedNameUsageID)) %>% 
  select(scientificName, taxonomicStatus, acceptedNameUsage, acceptedNameUsageID)

Similar functions filter_id, filter_rank, and filter_common take IDs, scientific ranks, or common names, respectively. Here, we can get taxonomic data on all bird names in the Catalogue of Life:

filter_rank(name = "Aves", rank = "class", provider = "col")

Combining these with dplyr functions can make it easy to explore this data: for instance, which families have the most species?

filter_rank(name = "Aves", rank = "class", provider = "col") %>%
  filter(taxonomicStatus == "accepted", taxonRank=="species") %>% 
  group_by(family) %>%
  count(sort = TRUE) %>% 
  head()

Using the database connection directly

filter_* functions by default return in-memory data frames. Because they are filtering functions, they return a subset of the full data which matches a given query (names, ids, ranks, etc), so the returned data.frames are smaller than the full record of a naming provider. Working directly with the SQL connection to the MonetDBLite database gives us access to all the data. The taxa_tbl() function provides this connection:

taxa_tbl("col")

We can still use most familiar dplyr verbs to perform common tasks. For instance: which species has the most known synonyms?

taxa_tbl("itis") %>% 
  count(acceptedNameUsageID, sort=TRUE)

However, unlike the filter_* functions which return convenient in-memory tables, this is still a remote connection. This means that direct access using the taxa_tbl() function (or directly accessing the database connection using td_connect()) is more low-level and requires greater care. For instance, we cannot just add a %>% mutate(acceptedNameUsage = get_names(acceptedNameUsageID)) to the above, because get_names does not work on a remote collection. Instead, we would first need to use a collect() to pull the summary table into memory. Users familiar with remote databases in dplyr will find using taxa_tbl() directly to be convenient and fast, while other users may find the filter_* approach to be more intuitive.

Learn more

• See richer examples the package Tutorial.

• Learn about the underlying data sources and formats in Data Sources

• Get better performance by selecting an alternative database backend engines.

taxadb:::td_disconnect()

if(require(codemeta)) codemeta::write_codemeta()

---

Please note that this project is released with a Contributor Code of Conduct. By participating in this project you agree to abide by its terms.

cboettig/taxadb documentation built on April 17, 2024, 6:34 p.m.