Biomarker Research
BioFlow

Throughout my academic journey I became obsessed with biomarkers — specifically genomic ones across the entire omics spectrum. The science was clear; the tooling was not. Existing bioinformatics software assumed you could write Python, configure environments, and navigate command-line interfaces. For researchers coming from wet lab backgrounds, that assumption shuts the door entirely.

I was one of those researchers. The biology was in my head. The code was not. And I refused to accept that as a permanent limitation.

If the barrier was code, the solution was to remove it — not by simplifying the science, but by abstracting the implementation. Modern AI coding tools made it possible to build sophisticated bioinformatics logic without writing every line by hand. The bet was that a visual, canvas-based interface could expose the full power of omics analysis pipelines to people who think in experiments, not functions.

Open source from the start. If this problem existed for me, it existed for thousands of others. The community should own it.

BioFlow is built as a visual canvas where researchers drag, connect, and configure analysis nodes instead of writing scripts. Each node encapsulates a bioinformatics operation — quality control, alignment, differential expression, pathway analysis — and the canvas handles the execution logic underneath.

• Visual pipeline builder — connect analysis steps like a flowchart
• Omics-ready modules — genomics, transcriptomics, proteomics workflows
• AI-assisted configuration — parameter guidance without needing to know defaults
• Open source — fully forkable, extendable by the community

Built with AI coding tools as a core part of the development process — which is itself a proof of the concept.

BioFlow is actively in development and open source. The repository is public and the architecture is taking shape. The experience of building it has confirmed the original hypothesis: AI coding tools genuinely lower the barrier to building complex scientific software for domain experts without engineering backgrounds.

The biggest lesson so far: the hardest part isn't the bioinformatics — it's designing an interface abstract enough to be accessible but expressive enough to be scientifically rigorous. That tension is what makes the problem interesting.

What I'd improve next: better error messaging for failed pipeline steps, and pre-built templates for common omics workflows so researchers can get to results faster.

View on GitHub →