Internship Projects

Introduction

As a Biotechnology student at the Federal University of Pará (UFPa) (ResearchGate, Lattes CV, Orcid), I completed a voluntary internship at BioSharp OÜ during the first half of 2021. In this post I summarise the main activities I carried out during that period. The most significant was adapting the Bactopia software to Nextflow DSL-2, alongside maintenance work on other BioSharp OÜ pipelines.

Internship Context

Shortly after joining BioSharp I was introduced to the company's workflow: a git-branch-based development process with tools such as Conda, Docker, and Nextflow, and cloud pipeline execution platforms (Google Life Sciences and AWS) via Nextflow Tower. Once familiar with the working methodology, I was introduced to Bactopia.

Bactopia is a Nextflow-based tool for the automated processing of bacterial genomes (fastq reads from Whole Genome Sequencing — WGS). It performs a broad range of analyses: read quality control, reference mapping, de novo assembly, variant calling (SNPs and INDELs), antimicrobial-resistance mutation profiling, functional genome annotation, and phylogenetic analysis.

Despite all its features, Bactopia was still at version 1.6.4 and upgrading to DSL-2 was one of the targets for the upcoming 2.0.0 release. The DSL-2 migration would give Bactopia a more flexible workflow structure — replacing static processes with modular, easily-modified components — allowing sub-workflows and new modules to be added without disrupting the main workflow, and making the codebase more accessible to contributors. Under the guidance of Abhinav Sharma, I began applying DSL-2 to Bactopia.

Development

Modularisation

Bactopia originally held all of its processes in a single main.nf script. With DSL-2 the number of lines in that script could be drastically reduced by splitting processes into importable modules. To keep the main script lean and allow individual modules to be modified without breaking the overall workflow, I adopted a workflow → sub-workflow → module hierarchy and organised the modules into the following folder structure:

modules/
├── Quality Control
│   └── Fastqc
│       ├── process.nf
│       ├── bin
│       ├── nextflow.config
│       ├── README.md
│       ├── templates
│       ├── test_data
│       ├── test_params.yaml

In this new structure every process became independent — each one executable on its own with dedicated tests, parameters, and stub blocks for fast mock-data testing. bioinformatics-lab/bactopia#4

This combination of DSL-2 and stubs enables independent module development, meaning main.nf can remain untouched while new modules and configurations are added. Taking advantage of this per-module testing setup, the tool's author Robert Petit later implemented a pytest-based test suite for each module.

Beyond modularisation, every process also had its input/output declarations updated to the new Nextflow Channel format — something that was only feasible thanks to the isolated per-module testing.

The full work can be followed in Pull request #3 on BioSharp OÜ's development fork.

Impact of the Work

The goal of building a DSL-2 version of Bactopia was primarily to familiarise myself with Nextflow pipeline development. However, during the modularisation process, the tool's lead developer noticed the work and reached out via GitHub expressing interest in incorporating the changes into version 2.0.

After Robert's message, Abhinav guided me through a git rebase to align the work with the upstream codebase. Following Robert's adjustments, the changes were merged into Bactopia's main branch and eventually became part of the 2.0 release. The full history of those changes can be seen in Bactopia#228.

Other Activities

Alongside the Bactopia work, I also contributed to maintenance and code updates in several open-source pipelines — an activity encouraged as an introduction to open collaboration on GitHub. These pipelines covered analyses of Legionella sp.; isoniazid (INH) mono-resistant Mycobacterium tuberculosis (MTB); multidrug-resistant (MDR) and extensively drug-resistant (XDR) MTB; and identification of novel Non-Tuberculous Mycobacteria (NTM) species. Key contributions included:

• Writing user-friendly README files
• Adding execution profiles for AWS, Azure, and GCP, supporting both Docker containers and Conda packages
• Creating stub blocks for fast pipeline testing
• Writing task files for job execution

Conclusion

My internship at BioSharp OÜ resulted in contributions to the Bactopia 2.0 release, which led to a scientific presentation at the conference Sequencing to Function: Analysis and Applications for the Future: Bactopia v2: Highly scalable, portable and customizable bacterial genome analysis.

Beyond the technical output, I gained skills that have directly shaped my academic and professional development:

• Programming in Nextflow
• Critical and empathetic engineering mindset
• Scientific pipeline design for genome analysis
• Git and GitHub-based collaboration
• Cloud computing (AWS, GCP, Kubernetes)
• Project management with JIRA and Freedcamp

This internship was enormously valuable — it broadened my bioinformatics skills, introduced me to new collaborations, and motivated me to keep learning at the intersection of programming and life sciences.