BioJava

Core Module

This module provides Java classes to model amino acid or nucleotide sequences. The classes were designed so that the names are familiar and make sense to biologists and also provide a concrete representation of the steps in going from a gene sequence to a protein sequence for computer scientists and programmers.

A major change between the legacy BioJava project and BioJava3 lies in the way framework has been designed to exploit then-new innovations in Java. A sequence is defined as a generic interface allowing the rest of the modules to create any utility that operates on all sequences. Specific classes for common sequences such as DNA and proteins have been defined in order to improve usability for biologists. The translation engine really leverages this work by allowing conversions between DNA, RNA and amino acid sequences. This engine can handle details such as choosing the codon table, converting start codons to methionine, trimming stop codons, specifying the reading frame and handing ambiguous sequences.

Special attention has been paid to designing the storage of sequences to minimize space needs. Special design patterns such as the Proxy pattern allowed the developers to create the framework such that sequences can be stored in memory, fetched on demand from a web service such as UniProt, or read from a FASTA file as needed. The latter two approaches save memory by not loading sequence data until it is referenced in the application. This concept can be extended to handle very large genomic datasets, such as NCBI GenBank or a proprietary database.

Protein structure modules

The protein structure modules provide tools to represent and manipulate 3D biomolecular structures. They focus on protein structure comparison.

The following algorithms have been implemented and included in BioJava.

• FATCAT algorithm for flexible and rigid body alignment.^[17]
• The standard Combinatorial Extension (CE) algorithm.^[18]
• A new version of CE that can detect circular permutations in proteins.^[19]

These algorithms are used to provide the RCSB Protein Data Bank (PDB)^[20] Protein Comparison Tool as well as systematic comparisons of all proteins in the PDB on a weekly basis.^[21]

Parsers for PDB^[22] and mmCIF^[23] file formats allow the loading of structure data into a reusable data model. This feature is used by the SIFTS project to map between UniProt sequences and PDB structures.^[24] Information from the RCSB PDB can be dynamically fetched without the need to manually download data. For visualization, an interface to the 3D viewer Jmol is provided.^[4]

Genome and Sequencing modules

This module is focused on the creation of gene sequence objects from the core module. This is realized by supporting the parsing of the following popular standard file formats generated by open source gene prediction applications:

• GTF files generated by GeneMark^[25]
• GFF2 files generated by GeneID^[26]
• GFF3 files generated by Glimmer^[27]

Then the gene sequence objects are written out as a GFF3 format and is imported into GMOD.^[28] These file formats are well defined but what gets written in the file is very flexible.

For providing input-output support for several common variants of the FASTQ file format from the next generation sequencers,^[29] a separate sequencing module is provided. For samples on how to use this module please go to this link.

Alignment module

This module contains several classes and methods that allow users to perform pairwise and multiple sequence alignment. Sequences can be aligned in both a single and multi-threaded fashion. BioJava implements the Needleman-Wunsch^[30] algorithm for optimal global alignments and the Smith and Waterman's^[31] algorithm for local alignments. The outputs of both local and global alignments are available in standard formats. In addition to these two algorithms, there is an implementation of Guan–Uberbacher algorithm^[32] which performs global sequence alignment very efficiently since it only uses linear memory.

For Multiple Sequence Alignment, any of the methods discussed above can be used to progressively perform a multiple sequence alignment.

ModFinder module

The ModFinder module provides new methods to identify and classify protein modifications in protein 3D structures. Over 400 different types of protein modifications such as phosphorylation, glycosylation, disulfide bonds metal chelation etc. were collected and curated based on annotations in PSI-MOD,^[34] RESID^[35] and RCSB PDB.^[36] The module also provides an API for detecting pre-, co-, and post-translational protein modifications within protein structures. This module can also identify phosphorylation and print all pre-loaded modifications from a structure.

Amino acid properties module

This module attempts to provide accurate physio-chemical properties of proteins. The properties that can calculated using this module are as follows:

• Molecular mass
• Extinction coefficient
• Instability index
• Aliphatic index
• Grand average of hydropathy
• Isoelectric point
• Amino acid composition

The precise molecular weights for common isotopically labelled amino acids are included in this module. There also exists flexibility to define new amino acid molecules with their molecular weights using simple XML configuration files. This can be useful where the precise mass is of high importance such as mass spectrometry experiments.

Protein disorder module

The goal of this module is to provide users ways to find disorders in protein molecules. BioJava includes a Java implementation of the RONN predictor. The BioJava 3.0.5 makes use of Java's support for multithreading to improve performance by up to 3.2 times,^[37] on a modern quad-core machine, as compared to the legacy C implementation.

There are two ways to use this module:

• Using library function calls
• Using command line

Some features of this module include:

• Calculating the probability of disorder for every residue in a sequence
• Calculating the probability of disorder for every residue in the sequence for all proteins from a FASTA input file
• Get the disordered regions of the protein for a single protein sequence or for all the proteins from a FASTA input file

Web service access module

As per the current trends in bioinformatics, web based tools are gaining popularity. The web service module allows bioinformatics services to be accessed using REST protocols. Currently, two services are implemented: NCBI Blast through the Blast URLAPI (previously known as QBlast) and the HMMER web service.^[38]