Software testing in microbial bioinformatics: a call to action

Abstract

Computational algorithms have become an essential component of research, with great efforts by the scientific community to raise standards on development and distribution of code. Despite these efforts, sustainability and reproducibility are major issues since continued validation through software testing is still not a widely adopted practice. Here, we report seven recommendations that help researchers implement software testing in microbial bioinformatics. We have developed these recommendations based on our experience from a collaborative hackathon organised prior to the American Society for Microbiology Next Generation Sequencing (ASM NGS) 2020 conference. We also present a repository hosting examples and guidelines for testing, available from https://github.com/microbinfie-hackathon2020/CSIS.

Keywords: computational biology; continuous integration; software testing.

Fig. 1.

Testing strategies. (a) White-box vs. black-box testing. In white-box testing, the tester knows the underlying code and structure of the software, where the tester does not know this in black-box testing. Note that this distinction is not strictly dichotomous and is considered less useful nowadays (b) Unit vs. integration vs. system testing. When software comprises several modules, it is possible to test each single module (unit testing), groups of related modules (integration testing) or all modules (system testing). Note that the terms white-box testing and unit testing are sometimes used interchangeably but relate to different concepts.