Abstract This document serves as a formal correction to the article "Reproducible big data science: A case study in continuous FAIRness" (DOI: 10.1371/journal.pone.0213013). It addresses critical inaccuracies in the original Data Availability statement by providing updated persistent identifiers for six major genomic datasets, thereby restoring the study's reproducibility.
Introduction
The pursuit of "continuous FAIRness"—ensuring data remains Findable, Accessible, Interoperable, and Reusable—is a central challenge in modern big data science. In 2019, a case study was published to establish a framework for reproducible genomic research. However, the digital infrastructure supporting such research is subject to "link rot," where original URLs become obsolete over time. This correction is a necessary maintenance step to ensure the long-term viability and transparency of the study’s findings.
The Challenge of Data Persistence
Reproducibility in computational biology depends on the ability of external researchers to access the exact datasets used in the original analysis. In the original publication, the Data Availability statement contained incorrect or broken URLs. This failure compromised the "Findable" and "Accessible" requirements of the FAIR principles. To rectify this, the authors have transitioned from volatile URLs to persistent identifiers (PIDs), which provide a more robust foundation for digital scholarship.
Implementation of Minimal Identifiers (minids)
The correction utilizes Minimal Identifiers (minids), a lightweight persistent identifier system. These identifiers are resolved through the identifiers.org service, which acts as a central registry. Unlike standard URLs, minids are location-independent; they map a unique string (e.g., minid:b9dt2t) to the data's current storage location. This approach ensures that even if the data is moved to a different server or repository, the identifier remains constant and the data remains discoverable through the central registry.
Updated Dataset Access
The following six datasets, essential for reproducing the study's genomic analysis, now have functional and persistent identifiers:
- DNase-Seq Data: minid:b9dt2t
- Aligned BAM Files: Genomic alignment data, minid:b9vx04
- BED Files of Footprints: Genomic region sets, minid:b9496p
- Non-redundant Motifs Database: Curated genetic motifs, minid:b97957
- Motif Intersected Hits: Results of motif analysis, minid:b9p09p
- Transcription Factor Binding Sites: Generated binding site data, minid:b9v398
Discussion: FAIRness as an Ongoing Process
The title of the original paper emphasized "continuous" FAIRness, and this correction serves as a practical demonstration of that concept. Data curation does not end at the moment of publication; it requires ongoing monitoring and updates to account for the evolution of digital repositories. By issuing a formal correction rather than an informal update, the authors maintain a transparent record of data provenance, which is a cornerstone of scientific integrity.
Conclusion
While this correction is limited to updating data pointers, it highlights a broader systemic issue in big data science: the fragility of digital citations. The use of persistent identifiers like minids is a critical step toward mitigating link rot. Future work in this field should focus on automating the detection of broken data links and streamlining the update process within registries to ensure that "FAIRness" remains a sustainable reality for the scientific community.
Summary updated