A FAIR platform for reproducing mutational signature detection on tumor sequencing data
This paper presents a portable, privacy-preserving, in-browser platform for the reproducible assessment of mutational signature detection methods from sparse sequencing data generated by targeted gene panels. The platform aims to address the reproducibility challenges in mutational signature research by adhering to the FAIR principles, making it findable, accessible, interoperable, and reusable. Our approach focuses on the detection of specific mutational signatures, such as SBS3, which have been linked to specific mutagenic processes. The platform relies on publicly available data, simulation, downsampling techniques, and machine learning algorithms to generate training data and labels and to train and evaluate models. The key achievement of our platform is its transparency, reusability, and privacy preservation, enabling researchers and clinicians to analyze mutational signatures with the guarantee that no data circulates outside the client machine.
PDF Abstract
Colab
