The BioJava modules

The BioJava library is organized into several modules for maximum flexibility. Users can choose what subset of modules to depend on in their projects.

The core module provides interfaces and routines to work with protein and nucleotide sequences. Some of the functionality includes parsing sequences from local files and remote resources, conversion between file formats and gene to protein translation. This module acts as a base module and others can depend on it. The alignment module supplies standard algorithms and data structures for pairwise and multiple sequence alignments. In version 5, the phylo module was integrated into the alignment module to support phylogenetic analyses using the Forester library [7] (https://github.com/cmzmasek/forester). The structure module provides data structures and algorithms to parse, manipulate and compare 3D structures of biological macromolecules, and the structure-gui module allows visualization of structures and structure alignments in Jmol [8].

Other smaller modules provide more specific functionality for different Bioinformatics fields. The genome module deals with genomic data supporting memory-efficient parsers for GTF, GFF2, GFF3 and FASTQ file formats. For protein analyses, the aa-prop module provides a range of physicochemical properties (e.g. molecular weight, isoelectric point, extinction coefficient, net charge), the protein-disorder module implements a parallelized version of the Regional Order Neural Network (RONN) [9] for disorder prediction, and the modfinder module implements routines to identify protein modifications observed in 3D structures [10]. Survival analyses using the Kaplan-Meier estimator [11] are possible with the survival module. The ontology module adds support for ontologies and parsing OBO files. And finally, several bioinformatics services can be accessed using REST protocols using the ws module.

More detailed information about each BioJava module can be found in the Wikipedia page (https://en.wikipedia.org/wiki/BioJava) and BioJava documentation, as well as in the previous publication [3].

New features

A number of new features have been added to BioJava in the last few years, most of which are related to structural biology data handling. Below we highlight a few of the most relevant.

Multiple structural alignments.

BioJava implements a wide range of pairwise structure alignment algorithms to perform rigid, flexible, and non-topological alignments. We introduced a custom implementation of the CE-MC procedure [14] in the org.biojava.nbio.structure.align.multiple package to generate multiple structure alignments by combining the output of pairwise alignment algorithms. Results are stored in a novel hierarchical data structure that supports rigid, flexible, and non-topological multiple structure alignments. Tools to manage and visualize alignments have also been adapted to enable multiple aligned structures, as demonstrated in Fig 1. More information can be found in the BioJava tutorial.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Multiple structure alignment of circularly permuted lectins generated and visualized with BioJava.

Implementations of CE-CP and CE-MC were used for the structural alignment, visualized using the Jmol based structure panel (left), the multiple alignment panel (top right), and a Forester based dendrogram of structural similarities (bottom right).

https://doi.org/10.1371/journal.pcbi.1006791.g001

BioJava release cycle

BioJava releases depend on the number and importance of contributions made to the library. Developer contributions happen via GitHub’s pull requests, where new code and fixes are reviewed by the community.

Since version 4, the semantic versioning philosophy has been strictly followed. Changes that break the API represent new major releases, additions to the API are minor releases, and bug fixes are regarded as bugfix releases. We have since released two major versions of the library (version 4 in January 2015 and version 5 in March 2018), and two minor releases for BioJava 4 and one for BioJava 5. In addition, we routinely released bugfix versions every few months. In December 2018, the library is at version 5.1.1.

All BioJava releases can be found on GitHub (https://github.com/biojava/biojava/releases). In addition, a manually curated changelog is kept in a separate file to facilitate understanding of the project’s history.

BioJava for method development

The extensive support of BioJava in basic operations like parsing and manipulating sequences and structures allows developers of novel algorithms to focus all their efforts on the bioinformatics problem itself. For example, BioJava has recently been part of the development of altORFev [20], a method to predict alternative open reading frames in eukaryotic mRNAs, CE-Symm [16], a detector of internal symmetry in protein structures, and EPPIC [21], a predictor of biological assemblies in crystal structures. Similarly, BioJava’s features to generate compressed files and convert between the structural formats were used for the development of the macromolecular compression format (MMTF) [12, 13].

Integration into large-scale analyses

BioJava can also be used for large-scale bioinformatics applications. Methods of the library can be efficiently run in parallel on large datasets with MMTF-Spark (https://github.com/sbl-sdsc/mmtf-spark), a project to promote scalable analysis of big data in structural bioinformatics. BioJava is also used by CloudPhylo [22], a tool written in Scala and built on Spark that is capable of processing large-scale genomic datasets for phylogeny reconstruction. As another example, BioJava methods were used to compare thousands of protein assembly models to experimental structures during the assessment of biological assemblies in CASP12 [23].

Scripting and notebooks

In recent years the JVM platform has grown beyond the Java language itself. A plethora of scripting languages that can interoperate with Java libraries have appeared, e.g. Scala, Kotlin, Clojure or Groovy. As a JVM-based library, BioJava can be seamlessly integrated into software written in any of those languages, and a few examples can already be found in the literature. For instance, some work towards Scala integration was accomplished in the BioScala project (https://github.com/bioscala/bioscala). Equally, integration with popular notebook software like Jupyter has become possible through projects like BeakerX (http://beakerx.com) that provide JVM support for Jupyter. In this context, a demo application providing a geometrical analysis of PDB data based on BioJava code is available at https://github.com/sbl-sdsc/biojava-notebooks.

Academic and commercial software

Last but not least, BioJava is a popular choice for the development of software platforms and web services that integrate several different bioinformatics applications. These include RobiNA [24], an integrated software solution for RNA-Seq-based transcriptomics, the HDX Workbench [25], an integrated desktop program for Hydrogen/Deuterium exchange mass spectrometry (HDX-MS) analysis, and G2S [26], a web-service for annotating genomic variants on 3D protein structures, to mention a few of the many examples. BioJava is also widely used by the RCSB Protein Data Bank (PDB) for their web-services [27], including protein quaternary symmetry annotation and visualization, structural comparisons and the exploration of protein modifications. Finally, the BioJava project powers a number of commercial software products such as those from Genomnia and DNASTAR.