Project lab

Real work, documented openly.

Each project shows its research question, data, methods, validation and limitations — with the code available to inspect and rerun.

Open-source code
Public datasets
Limitations stated up front

PublishedSoftwarePython

DNA Sequence Analysis Toolkit

A modular Python toolkit for validating DNA sequences before analysis. It separates canonical bases, IUPAC ambiguity codes, alignment gaps and unsupported characters, then reports composition and quality-control metrics for single sequences or entire FASTA files.

Scope, stated plainly: a PASS result means a sequence meets the toolkit’s character-validation rules. It does not establish biological correctness or suitability for every downstream analysis.

View on GitHub

PublishedMachine learningMSc thesis

Breast cancer multi-omics outcome classification

Four supervised classifiers compared for predicting recorded patient vital status from 1,936 molecular features across 705 TCGA-derived breast cancer samples. A 2026 re-analysis corrected two choices that biased the original evaluation: fitting the scaler before the train/test split, and an unstratified split.

What the re-analysis showed: Random Forest led on accuracy (89.4%), yet a model that always predicts the majority class already scores above 86%. Only 13% of patients carry the rarer label, and the models found 21–37% of them. The study answers a prognosis question, not a detection one — and it is not a diagnostic or prognostic device.

View on GitHub

In developmentWorkflowTranscriptomics

Reproducible RNA-seq analysis pipeline

A documented analysis of a well-described public GEO study: validated sample metadata, sample-level quality control, PCA, differential expression with false-discovery control and careful biological interpretation — all inside a reproducible software environment.

Planned deliverables: code, exact dataset accession, figures, methods, a pinned environment and an explicit limitations section. The case study will be published here when the work is complete.

Follow progress on GitHub

Our case-study standard

Five things every project must show.

A result is only as useful as the reader’s ability to check it.

01The question
Research question and its biological context.
02The data
Source, access conditions and preprocessing.
03The methods
Software, versions and the full workflow.
04The evaluation
Validation design and appropriate metrics.
05The limits
Interpretation, limitations and reproducibility notes.

Collaborate

Have a project or dataset in mind?

We welcome thoughtful collaboration on open, well-documented research.

Start a conversation