ROBOT Tutorial

James Overton & Becky Jackson

Overview

  1. Manual editing
  2. Repetitive tasks
  3. Automated workflows
  4. Quality assurance
  5. Modular development
  6. Modular releases
  7. Other commands

Manual Editing

Status Quo

  • development mainly done in Protégé
  • requires manual editing and review

The Solution

  • ROBOT lets us automate many of those tasks
  • … and provides better quality assurance!

ROBOT Releases

17 releases:

  • v1.0.0 released Feb 8, 2018 …
  • v1.8.1 released Jan 27, 2021

Cite ROBOT

R.C. Jackson, J.P. Balhoff, E. Douglass, N.L. Harris, C.J. Mungall, and J.A. Overton.
ROBOT: A tool for automating ontology workflows.
BMC Bioinformatics, vol. 20, July 2019.

Let’s get started!

  • download the tutorial repository
  • get the latest ROBOT with Docker:
    • sh run.sh robot --version
  • or install it following the documentation
  • navigate to examples: cd examples
    • if using Docker, start each command with sh ../run.sh

ROBOT for Repetitive Tasks

merge

reason

annotate

convert

Merge

  1. merge two separate ontologies
  2. merge imports into one ontology

Merge Separate Ontologies

robot merge --input edit.owl \
  --input foo.owl \
  --output results/merged.owl

Merge Imports

ROBOT will automatically merge imports. To not merge imports, include --collapse-import-closure false. This option is not supported in v1.0.0.

robot merge --input with-import.owl \
  --output results/merged_imports.owl

Reason

  1. logical validation
  2. automatic classification

Reasoners

  • ELK: a very fast reasoner, but not as powerful as HermiT
  • HermiT: optimized to classify complex ontologies
  • EMR: materializes anonymous expressions
  • Structural: a simple reasoner

Logical Validation

Validates that the ontology contains no unsatisfiable entities and that it is not inconsistent.

robot reason --input unsatisfiable.owl

Automatic Classification

robot reason --input non-reasoned.owl \
  --output results/reasoned.owl

Annotate

  1. add metadata to an ontology for release

Add Metadata

robot annotate --input edit.owl \
  --version-iri \
    https://github.com/ontodev/robot/releases/2019-07-31/edit.owl \
  --annotation oboInOwl:date "07:31:2019 12:00" \
  --output results/annotated.owl

Convert

  1. convert an editor’s file to verbose RDF/XML
  2. convert a released ontology to OBO

Ontology Formats

  • default format is RDF/XML but…
  • ontologies are shared in many formats
    • OWL functional
    • OBO
    • Turtle
    • and more…

Convert the Edit File

robot convert --input edit.owl \
  --format owl \
  --output results/release.owl

NOTE: OWL functional syntax is defined by the suffix .ofn - if you want to convert to that format, use --format ofn or an output ending in .ofn. The .owl suffix can be used to represent any OWL ontology. Using --format owl or excluding the --format option and specifying an output with .owl will convert to RDF/XML.

Convert to OBO

robot convert --input edit.owl \
  --output results/release.obo

NOTE: You do not always need to include the --format if the extension of the --output matches the desired format.

Automated Workflows

Chaining Commands

Output ontologies can be used as the input to subsequent commands. Only the first command uses an --input, and only the last command uses an --output.

robot merge --input edit.owl --collapse-import-closure true \
  reason --reasoner ELK \
  annotate --version-iri http://purl.obolibrary.org/obo/robot/2018-08-07/release.owl \
  convert --output results/chained_release.ttl

Makefiles

A Makefile contains a set of rules to make target objects.
Here, we use it to create the release files.

make release

NOTE: This Makefile will fail on v1.0.0 and v1.1.0 - alternatively, run make build.

Quality Assurance

diff

query (part 1)

verify

report

Diff

  1. compare the axioms in two version of an ontology
  2. generate a summary page of changes

diff supports different formats:

  • plain: default, shows removed and added axioms
  • pretty: plain, plus entity labels for IRIs
  • html: HTML summary page (easy for humans!)
  • markdown: markdown summary (great for GitHub!)

Compare Axioms (Default)

robot diff --left non-reasoned.owl \
  --right results/reasoned.owl \
  --output results/diff.txt

NOTE: If you do not include an --output, the results will be printed to the terminal.

Human-Readable Diff

robot diff --left non-reasoned.owl \
  --right results/reasoned.owl \
  --format html \
  --output results/diff.html

Query

  1. run a SPARQL SELECT query
  2. run a SPARQL ASK query

Select Query

SELECT queries can be useful for statistics on the ontology, and for sharing data. For example, you could get a list of all the terms containing just the IDs and labels.

robot query --input edit.owl \
  --query select.rq results/select.tsv

Ask Query

ASK queries can help determine if something exists in the ontology or not. For example, if you want to make sure that “trunk” is in the “vertebrate core” subset.

robot query --input edit.owl \
  --query ask.rq results/ask.txt

Verify

  1. perform a successful verification
  2. perform an unsuccessful verification

Successful Verification

Sometimes, classes can be accidentally orphaned. This verification ensures that all classes, aside from the top-level “anatomical entity”, have asserted parents. If they do not, the verification fails.

robot verify --input edit.owl \
  --queries verify.rq \
  --output-dir results

Unsuccessful Verification

This verify query checks that all classes have an equivalent class statement. In our edit ontology, only some of the classes have equivalent classes, so this will fail.

robot verify --input edit.owl \
  --queries verify_fail.rq \
  --output-dir results

Report

  1. run a series of SPARQL queries to find common violations

Violations

  • annotation, logical, or metadata violations
  • three logging levels: ERROR, WARN, and INFO

Generate a Report

robot report --input edit.owl --output results/report.tsv

Continuous Integration

  • Travis CI automatically builds and tests changes pushed to GitHub
  • Often uses make test from the Makefile
  • If any part of the test fails, the build will fail

Modular Development

extract

template

Extract

  1. create an import module with SLME
  2. create an import module with MIREOT

Import Modules

  • many bioontologies use terms from external sources
  • these sources contain more terms than needed
  • extract ensures the necessary terms and their dependencies are included in module

Extract with SLME

STAR: fixpoint-nested
BOT: bottom module
TOP: top module 

robot extract \
  --input-iri http://purl.obolibrary.org/obo/obi.owl \
  --term OBI:0000443 \
  --method BOT \
  --output results/obi_bot.owl

NOTE: You can also include a list of terms to extract in a text file with --term-file.
NOTE 2: To use a local file, use --input <file> instead of --input-iri.

Extract with MIREOT

Creates a simple hierarchy of terms.
Requires lower term(s) and optional upper term(s). 

robot extract \
  --input-iri http://purl.obolibrary.org/obo/obi.owl \
  --method MIREOT \
  --lower-terms obi_terms.txt \
  --output results/obi_mireot.owl

NOTE: Without specifiying any --upper-terms, the MIREOT method will include all ancestors up to owl:Thing.
NOTE 2: To just specify one term to extract, use --lower-term

Template

  1. create a module
  2. add a class to an ontology

Create a Module

This will create a standalone module that can be included in the edit file with an import statement. To update the module, editors only need to update the spreadsheet and run this command to remake the module.

robot template --input edit.owl \
  --template module.tsv \
  --ontology-iri http://purl.obolibrary.org/obo/robot/module.owl \
  --output results/module.owl

NOTE: template gets all the entity labels from edit.owl so we are able to use the labels in the spreadsheet, instead of always specifying the ID. If we didn’t include the --input, the labels would not resolve.

Add a Class to an Ontology

For one-time class creation (especially if many classes need to be created), a temporary template can be created and the results immediately merged into the edit ontology.

robot template --input edit.owl --merge-before \
  --template new_class.tsv \
  --output results/new_class.owl

NOTE: if your ontology includes imports, use --collapse-import-closure false with any merge option to maintain the closure.

Modular Releases

query (part 2)

remove

filter

Querying with SPARQL CONSTRUCT

  • CONSTRUCT produces RDF data in Turtle format
  • allows creation of new sets of triples

Update Annotations with CONSTRUCT

robot query --input edit.owl \
  --query construct.rq results/construct.ttl \
  merge --input results/construct.ttl \
  --output results/construct.owl

The construct.ttl file isn’t much use on its own; we need to merge it. For query, the output ontology is the unchanged input ontology, so we can chain this with the merge command to merge construct.ttl.

Querying with SPARQL UPDATE

  • UPDATE changes the RDF data and outputs an updated ontology
  • allows deletion and insertion of triples
  • no need to merge after running the UPDATE

Update Annotations with UPDATE

robot query --input edit.owl \
  --update update.ru \
  --output results/updated.owl

Here, we replace ‘definition’ (IAO:0000115) with ‘external_definition’ (UBPROP:0000001) for all UBERON terms.

Remove

  1. remove a class and its descendants
  2. create a ‘simple’ version of an ontology

Configuration

  • remove a term (or terms) from an ontology and any related terms
  • highly configurable --select option to remove related terms
  • specify types of axioms to remove with --axioms

Remove a Class + Descendants

robot remove --input edit.owl \
  --term UBERON:0000475 \
  --select "self descendants" \
  --output results/removed.owl

Create a ‘Simple’ Version

robot remove --input edit.owl \
  --axioms equivalent \
  remove --select parents --select anonymous --select imports \
  --output results/simple.owl

NOTE: --select accepts a string of options, or can be passed multiple times. For a string of options (previous example) the selected set is the union of all options (both the input term and all descendants of the term). For passing in multiple select options (as above), the options are processed in order (first the parents are selected, and then only the anonymous parents are selected).

Filter

Released with v1.2.0-alpha - previously, filter only filtered for object properties.

  1. extract a branch of an ontology
  2. create a subset based on annotations

Extract a Branch

robot filter --input edit.owl \
  --term UBERON:0000475 \
  --select "self descendants annotations" \
  --output results/branch.owl

NOTE: in order to include annotations on the filtered entities, --select annotations must be included. Otherwise, you muist include all annotation properties in the set of input terms.

Create a Subset

robot filter --input edit.owl \
  --select \
  "oboInOwl:inSubset=<http://purl.obolibrary.org/obo/uberon/core#uberon_slim>" \
  --select annotations \
  --output-iri http://purl.obolibrary.org/robot/uberon_slim.owl
  --output results/uberon_slim.owl

NOTE: selecting for annotations is highly configurable:
CURIE=CURIE
CURIE="literal"^^datatype
CURIE=<IRI>
CURIE=~"regex pattern"

Other Commands

explain     materialize

 mirror     reduce

  relax    rename    repair

unmerge     validate

Questions?