Metadata Extraction and JSON Export Pipeline for LAGE Sequencing Data

Recent advances in sequencing technologies have led to the production of large-scale genomic datasets, significantly accelerating progress in the life sciences. However, this rapid expansion introduces substantial data management challenges, particularly within sequencing facilities such as the Laboratory of Genomics and Epigenomics (LAGE) in Area Science Park (Trieste). The diversity of file formats, the fragmentation of metadata across multiple sources, and the absence of standardised traceability mechanisms hinder efficient data handling and long-term usability. This work presents the development of a modular and automated Python pipeline for the extraction, normalisation, and FAIR-compliant packaging of sequencing metadata. The system operates on heterogeneous laboratory outputs, using content-based file identification to drive metadata extraction through eleven specialised modules and packaging datasets as RO-Crate 1.2 research objects with machine-actionable metadata serialised in JSON-LD. The registry based modular architecture enables the flexible integration of new sequencing platforms without modifying the core system. To enhance data traceability, dedicated utility modules reconstruct sample lineage across multiple experimental steps through recursive project-level data aggre gation and chronological sample history generation. The pipeline was evaluated against three real-world datasets generated by major sequencing platforms at LAGE, including the Illumina NovaSeq6000, the Oxford Nanopore PromethION, and the Illumina BeadStudio iScan. The results demonstrate that the system successfully pro cesses heterogeneous datasets, standardises metadata, reconstructs sample provenance across multiple runs, and generates RO-Crate compliant research objects validated against the RO Crate 1.2 specification. From an operational perspective, the pipeline automates metadata extraction across the major sequencing platforms at LAGE, reducing human intervention to output verification and directly addressing the scalability and reproducibility challenges identified at the facility. Be yond this specific context, its domain-agnostic design makes it applicable to any data-intensive facility producing heterogeneous instrument outputs.