Accessing biological data in R with semantic web technologies

Abstract

Background. Semantic Web technologies are increasingly used in biological database systems. The improved expressiveness show advantages in tracking provenance and allowing knowledge to be more explicitly annotated. The list of semantic web standards needs a complementary set of tools to handle data in those formats to use them in bioinformatics workflows. Methods. The approach proposed in this paper uses the Apache Jena library to create an environment where semantic web technologies can be use in the statistical environment R. The code is exposed as two R packages available from the Comprehensive R Archive Network (CRAN). The RJava library and a custom convenience class is used to bridge between R and the Jena library. Results. We here present three examples showing how the Resource Description Framework (RDF) and SPARQL query standards can be employed in R. The first example takes input on BRCA1 SNPs from a BioMart and converts this into a RDF data set. The second example runs a query on an experimental remote SPARQL end point provided by Uniprot, and searches textual annotations of proteins encoded by the BRCA1 gene. The third example shows how the package can be used to handle RDF returned by OpenTox web services. Discussion. The two provided library bring basic semantic web technologies to R. While only a subset of Apache Jena is currently exposed, it provides key methods to deal with RDF data and resources. The libraries are freely available from the CRAN under the Affero GNU Public License version 3: http://cran.r-project.org/web/ packages/rrdf/.

Analysis of volatile organic compounds in exhaled breath to diagnose ventilator-associated pneumonia

“…The RRDF package 29 was then used to find pathways in Wikipathways 30 that included that specific VOC. Additionally, because Wikipathways includes only a limited number of metabolite pathways, the KEGG database 31 was analysed for pathways containing the identified VOCs.…”

3-D QSAutogrid/R: An Alternative Procedure To Build 3-D QSAR Models. Methodologies and Applications

“…In the VS package, the GA-selection variable was implemented by the means of the R-binary genetic algorithm (genalg R package) in combination with an ad hoc fitness evaluation R script in which SDEP/RMSEP was used as the discriminator.…”

“…Worthy of note is the included ability to determine, through a combinatorial calculation, the most appropriate pretreatment values to get preoptimized 3-D QSAR models; therefore, particular effort was given to data pretreatment and variable selection. To this aim, heavy use of the cross-validation (CV) techniques such as leave-one-out (LOO), leave-some-out (LSO), k-fold (KF), and Monte Carlo (MC) based CVs were applied either in standalone or in conjunction with a genetic algorithm (GA) as implemented in the genalg R package . Guided Region Selection (GRS) using just one probe, or a compilation of different probes (Multi Probe Guided Region Selection (MPGRS)), is a further available variable-selection method as previously reported. − …”

queryMed: Semantic Web functions for linking pharmacological and medical knowledge to data

“…Because data exploration is strongly linked with data analysis, several tools (Van Hage et al, 2013;Willighagen, 2014;Kurbatova et al, 2015) have made available Linked Data knowledge bases in the R statistical programming environment (R Core Team, 2017). However, designing SPARQL queries remains a technical barrier that none of the current R packages help to overcome.…”