Contents

1 Introduction

OnassisJavaLibs is a data package containing various compiled libraries in jar format and their source code used by Onassis to

2 Contents of the package

Conceptmapper Java (1.8 version) libraries (https://github.com/UCDenver-ccp/ccp-nlp) version 3.3.2, and semantic similarity libraries (https://github.com/UCDenver-ccp/ccp-nlp) version 0.9.5 have been compiled using maven.

They are available in jar format within the extdata directory of this package and can be located through .

3 Accessing and compiling the Java source code

The source code for the java libraries is available in the java directory of package tarball. Users interested in compiling their own jar files can refer to the following information.

The conceptmapper subdirectory contains the java code to annotate text with concepts from OBO ontologies. To create a jar file including all the needed dependencies, the source code can be compiled using maven (Apache Maven 3.3.9 was used).

The slib and similarity subdirectories contain the java code to determine the semantic similarities between concepts, and can be compiled and installed with the following goals, respectively:

#From the conceptmapper directory
mvn clean compile assembly:single --Dlog4j.configuration=log4j2.properties

#From the slib directory
mvn clean install 

#From the similarity directory
mvn clean install assembly:single

4 Use of the package

The methods and classes implemented in the described Java libraries can be used through R functions and methods available within Onassis. Alternatively, the Java code can be directly executed using rJava. For example a dictionary from an OBO ontology file can be created through the following code:

require(rJava)
## Loading required package: rJava
#Initializing the JVM
.jinit()

#Adding the path to the jar file
jarfilePath <- file.path(system.file('extdata', 'java', 'conceptmapper-0.0.1-SNAPSHOT-jar-with-dependencies.jar', package='OnassisJavaLibs'))

.jaddClassPath(jarfilePath)

#Creating an instance of the OntologyUtil with the sample obo file
ontoutil <- .jnew("edu.ucdenver.ccp.datasource.fileparsers.obo.OntologyUtil", .jnew('java/io/File', file.path(system.file('extdata', 'sample.cs.obo', package='OnassisJavaLibs'))))

#Creating the output file containing the conceptmapper dictionary
outputFile = .jnew("java/io/File", "dict.xml")

#Building of the dictionary from the OBO ontology
dictionary <- J("edu.ucdenver.ccp.nlp.wrapper.conceptmapper.dictionary.obo.OboToDictionary")$buildDictionary(
  outputFile,
  ontoutil,
  .jnull(),
  J("edu.ucdenver.ccp.datasource.fileparsers.obo.OntologyUtil")$SynonymType$EXACT
)

To compute the semantic similarity between two terms of the same ontology, classes in the similarity library can be used in this way:

#Adding the similarity library containing the similarity class to compute semantic similarities
jarfilePath <- file.path(system.file('extdata', 'java', 'similarity-0.0.1-SNAPSHOT-jar-with-dependencies.jar', package='OnassisJavaLibs'))

.jaddClassPath(jarfilePath)

#Creating an instance of the class Similarity
similarity <- .jnew("iit/comp/epigen/nlp/similarity/Similarity")


#Loading the ontology in a grah structure
file_obo <- file.path(system.file('extdata', 'sample.cs.obo', package='OnassisJavaLibs'))

ontology_graph <- similarity$loadOntology(file_obo)

#Setting the semantic similarity measures 
measure_configuration <- similarity$setPairwiseConfig('resnik', 'seco')

#Terms of the ontologies need to be converted into URIs

term1 <- 'http://purl.obolibrary.org/obo/CL_0000771'
  
term2 <- 'http://purl.obolibrary.org/obo/CL_0000988'

URI1 <- .jcast(similarity$createURI(term1), new.class = "org.openrdf.model.URI", check = FALSE, convert.array = FALSE)

URI2 <- .jcast(similarity$createURI(term2), new.class = "org.openrdf.model.URI", check = FALSE, convert.array = FALSE)

# Computation of the semantic similarity score

similarity_score <- .jcall(similarity, "D", "pair_similarity", URI1, URI2, .jcast(ontology_graph, new.class = "slib.graph.model.graph.G"), measure_configuration)

similarity_score
## [1] 0.1047388

5 Acknowledgements

We would like to thank you the library providers.

The methods for the conceptmapper pipeline and defining the ccp-nlp type system, have been developed and published by the Reagents of the University of Colorado under BSD 3-clause license.

The methods for computing the semantic similarities instead have been developed and published by the the Ecole des mines d’Alès under the GPL-compatible CeCILL license.

Both licenses are provided within the package.