Introduction to the taxa package

Scott Chamberlain and Zachary Foster


taxa defines taxonomic classes and functions to manipulate them. The goal is to use these classes as low level fundamental taxonomic classes that other R packages can build on and supply robust manipulation functions (e.g. subsetting) that are broadly useful.

There are two distinct types of classes in taxa:

Diagram of class concepts for taxa classes:

Relationship between classes implemented in the taxa package. Diamond-tipped arrows indicate that objects of one class are used in another class. For example, a database object can stored in the taxon_rank, taxon_name, or taxon_id objects. A standard arrow indicates inheritance. For example, the taxmap class inherits the taxonomy class. * means that the object (e.g. a database object) can be replaced by a simple character vector. ? means that the data is optional (Note: being able to replace objects with characters might be going away soon).


For the latest “stable” release, use the CRAN version:


For all the latest improvements, bug fixes, and bugs, you can download the development version:


The classes

Minor component classes

There are a few optional classes used to store information in other classes. These will probably mostly be of interest to developers rather than users.


Taxonomic data usually comes from a database. A common example is the NCBI Taxonomy Database used to provide taxonomic classifications to sequences deposited in other NCBI databases. The database class stores the name of the database and associated information:

To save on memory, a selection of common databases is provided with the package (database_list) and any in this list can be used by name instead of making a new database object (e.g. "ncbi" instead of the ncbi above).


Taxa might have defined ranks (e.g. species, family, etc.), ambiguous ranks (e.g. “unranked”, “unknown”), or no rank information at all. The particular selection and format of valid ranks varies with database, so the database can be optionally defined. If no database is defined, any ranks in any order are allowed.


The taxon name can be defined in the same way as rank.


Each database has its set of unique taxon IDs. These IDs are better than using the taxon name directly because they are guaranteed to be unique, whereas there are often duplicates of taxon names (e.g. Orestias elegans is the name of both an orchid and a fish).

The “taxon” class

The taxon class combines the classes containing the name, rank, and ID for the taxon. There is also a place to define an authority of the taxon.

Instead of the name, rank, and ID classes, simple character vectors can be supplied. These will be converted to objects automatically.

The taxa class is just a list of taxon classes. It is meant to store an arbitrary list of taxon objects.

The “hierarchy” class

Taxonomic classifications are an ordered set of taxa, each at a different rank. The hierarchy class stores a list of taxon classes like taxa, but hierarchy is meant to store all of the taxa in a classification in the correct order.

Multiple hierarchy classes are stored in the hierarchies class, similar to how multiple taxon are stored in taxa.

The “taxonomy” class

The taxonomy class stores unique taxon objects in a tree structure. Usually this kind of complex information would be the output of a file parsing function, but the code below shows how to construct a taxonomy object from scratch (you would not normally do this).

# define taxa
notoryctidae <- taxon(name = "Notoryctidae", rank = "family", id = 4479)
notoryctes <- taxon(name = "Notoryctes", rank = "genus", id = 4544)
typhlops <- taxon(name = "typhlops", rank = "species", id = 93036)
mammalia <- taxon(name = "Mammalia", rank = "class", id = 9681)
felidae <- taxon(name = "Felidae", rank = "family", id = 9681)
felis <- taxon(name = "Felis", rank = "genus", id = 9682)
catus <- taxon(name = "catus", rank = "species", id = 9685)
panthera <- taxon(name = "Panthera", rank = "genus", id = 146712)
tigris <- taxon(name = "tigris", rank = "species", id = 9696)
plantae <- taxon(name = "Plantae", rank = "kingdom", id = 33090)
solanaceae <- taxon(name = "Solanaceae", rank = "family", id = 4070)
solanum <- taxon(name = "Solanum", rank = "genus", id = 4107)
lycopersicum <- taxon(name = "lycopersicum", rank = "species", id = 49274)
tuberosum <- taxon(name = "tuberosum", rank = "species", id = 4113)
homo <- taxon(name = "homo", rank = "genus", id = 9605)
sapiens <- taxon(name = "sapiens", rank = "species", id = 9606)
hominidae <- taxon(name = "Hominidae", rank = "family", id = 9604)

# define hierarchies
tiger <- hierarchy(mammalia, felidae, panthera, tigris)
cat <- hierarchy(mammalia, felidae, felis, catus)
human <- hierarchy(mammalia, hominidae, homo, sapiens)
mole <- hierarchy(mammalia, notoryctidae, notoryctes, typhlops)
tomato <- hierarchy(plantae, solanaceae, solanum, lycopersicum)
potato <- hierarchy(plantae, solanaceae, solanum, tuberosum)

# make taxonomy
(tax <- taxonomy(tiger, cat, human, tomato, potato))

Unlike the hierarchies class, each unique taxon object is only represented once in the taxonomy object. Each taxon has a corresponding entry in an edge list that encode how it is related to other taxa. This makes taxonomy more compact, but harder to manipulate using standard indexing. To make manipulation easier, there are functions like filter_taxa and subtaxa that will be covered later. In general, the taxonomy and taxmap objects (covered later) would be instantiated using a parser like parse_tax_data. This is covered in detail in the parsing vignette.


A “supertaxon” is a taxon of a coarser rank that encompasses the taxon of interest (e.g. “Homo” is a supertaxon of “sapiens”). The supertaxa function returns the supertaxa of all or a subset of the taxa in a taxonomy object.

By default, the taxon IDs for the supertaxa of all taxa are returned in the same order they appear in the edge list. Taxon IDs (character) or edge list indexes (integer) can be supplied to the subset option to only return information for some taxa.

What is returned can be modified with the value option:

You can also subset based on a logical test:

The subset and value work the same for most of the following functions as well. See all_names(tax) for what can be used with value and subset. Note how value takes a character vector ("taxon_ranks"), but subset can use the same value (taxon_ranks) as a part of an expression. taxon_ranks is actually a function that is run automatically when its name is used this way:

This is an example of Non-standard evaluation (NSE). NSE makes codes easier to read an write. The call to supertaxa could also have been written without NSE like so:


The “subtaxa” of a taxon are all those of a finer rank encompassed by that taxon. For example, sapiens is a subtaxon of Homo. The subtaxa function returns all subtaxa for each taxon in a taxonomy object.

This and the following functions behaves much like supertaxa, so we will not go into the same details here.


We call taxa that have no supertaxa “roots”. The roots function returns these taxa.


We call taxa without any subtaxa “leaves”. The leaves function returns these taxa.

other functions

There are many other functions to interact with taxonomy object, such as stems and n_subtaxa, but these will not be described here for now.

The “taxmap” class

The taxmap class is used to store any number of tables, lists, or vectors associated with taxa. It is basically the same as the taxonomy class, but with the following additions:

All the functions described above for the taxonomy class can be used with the taxmap class.

In most functions that work with taxmap objects, the names of list/vector data sets, table columns, or functions can be used as if they were separate variables on their own (i.e. NSE). In the case of functions, instead of returning the function itself, the results of the functions are returned. To see what variables can be used this way, use all_names.

For example using my_taxmap$data$info$n_legs or n_legs will have the same effect inside manipulation functions like filter_taxa described below. This is similar to how taxon_ranks was used in supertaxa in a previous section. To get the values of these variables, use get_data.

Note how “taxon_names” and “dangerous” are used below.


In addition to all of the functions like subtaxa that work with taxonomy, taxmap has a set of functions to manipulate data in a taxonomic context using functions based on dplyr. Like many operations on taxmap objects, there are a pair of functions that modify the taxa as well as the associated data, which we call “observations”. The filter_taxa and filter_obs functions are an example of such a pair that can filter taxa and observations respectively. For example, we can use filter_taxa to subset all taxa with a name starting with “t”:

There can be any number of filters that resolve to TRUE/FALSE vectors, taxon ids, or edge list indexes. For example, below is a combination of a TRUE/FALSE vectors and taxon id filter:

There are many options for filter_taxa that make it very flexible. For example, the supertaxa option can make all the supertaxa of selected taxa be preserved.

The filter_obs function works in a similar way, but subsets observations in my_taxmap$data.

You can choose to filter out taxa whose observations did not pass the filter as well:

Note how both the taxonomy and the associated data sets were filtered. The drop_obs option can be used to specify which non-target (i.e. not "info") data sets are filtered when taxa are removed.


The functions sample_n_obs and sample_n_taxa are similar to filter_obs and filter_taxa, except taxa/observations are chosen randomly. All of the options of the “filter_” functions are available to the “sample_” functions

Adding columns

Adding columns to tabular data sets is done using mutate_obs.

Note how you can use newly created columns in the same call.


Sorting the edge list and observations is done using arrage_taxa and arrange_obs.

Parsing data

The taxmap class has the ability to contain and manipulate very complex data. However, this can make it difficult to parse the data into a taxmap object. For this reason, there are three functions to help creating taxmap objects from nearly any kind of data that a taxonomy can be associated with or derived from. The figure below shows simplified versions of how to create taxmap objects from different types of data in different formats.

The parse_tax_data and lookup_tax_data have, in addition to the functionality above, the ability to include additional data sets that are somehow associated with the source data sets (e.g. share a common identifier). Elements in these data sets will be assigned the taxa defined in the source data, so functions like filter_taxa and filter_obs will work on all of the data set at once. See the parsing vignette for more information.

Parsing Hierarchy and hierarchies objects

A set of functions are available for parsing objects of class Hierarchy and hierarchies. These functions are being ported from the CRAN package binomen.

The functions below are “taxonomically aware” so that you can use for example > and < operators to filter your taxonomic names data.


pick() - Pick out specific taxa, while others are dropped


pop() - Pop out taxa, that is, drop them


span() - Select a range of taxa, either by two names, or relational operators

For more information

This vignette is meant to be just an outline of what taxa can do. In the future, we plan to release additional, in-depth vignettes for specific topics. More information for specific functions and examples can be found on their man pages by typing the name of the function prefixed by a ? in an R session. For example, ?filter_taxa will pull up the help page for filter_taxa.