Correction to: Comprehensive benchmarking and ensemble approaches for metagenomic classifiers

Alexa B. R. McIntyre, Rachid Ounit, Ebrahim Afshinnekoo, Robert J. Prill, Elizabeth Hénaff, Noah Alexander, Samuel S. Minot, David Danko, Jonathan Foox, Sofia Ahsanuddin, Scott Tighe, Nur A. Hasan, Poorani Subramanian, Kelly Moffat, Shawn Levy, Stefano Lonardi, Nick Greenfield, Rita R. Colwell, Gail Rosen, Christopher E. MasonGenome 2019 · DOI · OpenAlex

This document serves as a formal correction to the 2017 research article "Comprehensive benchmarking and ensemble approaches for metagenomic classifiers," originally published in Genome Biology. As a cornerstone study for researchers evaluating taxonomic labeling tools for environmental DNA, maintaining the accuracy and accessibility of its resources is vital for the scientific community. This update addresses critical issues regarding data availability and software implementation to ensure the study’s continued reproducibility.

A significant challenge in bioinformatics is the longevity of data hosting. Over time, the original repositories for the datasets and analysis scripts used in the 2017 study became unavailable. To rectify this, the authors have migrated these essential resources to more permanent academic infrastructure.

New Resource Locations:

  • Primary Datasets: The benchmarking datasets are now hosted on the IMMSA server via the NCBI FTP: ftp://ftp-private.ncbi.nlm.nih.gov/nist-immsa/IMMSA/
  • Analysis Scripts and Figures: The code required to replicate the study's results and generate its figures is now available through a Cornell University Git repository: https://scu.med.cornell.edu/git/abm237/benchmarking_metagenomic_classifiers

By moving these assets to NCBI and university-backed systems, the authors aim to provide a more stable foundation for future researchers building upon this work.

Technical Clarification: The Kraken Classifier

The original publication contained a misleading instruction regarding the build command for Kraken, one of the most widely utilized metagenomic classifiers. The choice of database is a fundamental component of metagenomic analysis, and the original command defaulted to a bacterial-only scope in the context of the manuscript's specific instructions.

To utilize the standard database—which includes archaea and viruses in addition to bacteria—the --standard flag must be explicitly included during the build process.

  • Original (Incorrect for standard scope): kraken-build --db $DBNAME
  • Corrected (Standard scope): kraken-build --standard --db $DBNAME

This distinction is critical. Without the --standard flag, researchers may inadvertently limit their search space, potentially leading to false negatives for non-bacterial organisms in their samples.

The Importance of Living Benchmarks

This correction highlights a broader reality in computational biology: benchmarking studies are not static artifacts but "living" documents. As software versions evolve and hosting services change, post-publication maintenance becomes necessary. This update ensures that the 2017 study remains a functional and accurate resource for the metagenomics community, reinforcing the principles of transparency and reproducibility in high-throughput sequencing analysis.

Summary updated

Abstract

Following publication of the original article [1], the authors would like to highlight the following two corrections.
BibTeX
@inproceedings{mcintyre2019correction,
  author = {Alexa B. R. McIntyre and Rachid Ounit and Ebrahim Afshinnekoo and Robert J. Prill and Elizabeth Hénaff and Noah Alexander and Samuel S. Minot and David Danko and Jonathan Foox and Sofia Ahsanuddin and Scott Tighe and Nur A. Hasan and Poorani Subramanian and Kelly Moffat and Shawn Levy and Stefano Lonardi and Nick Greenfield and Rita R. Colwell and Gail Rosen and Christopher E. Mason},
  title = {{Correction to: Comprehensive benchmarking and ensemble approaches for metagenomic classifiers}},
  booktitle = {Genome},
  year = {2019},
  doi = {10.1186/s13059-019-1687-2},
}
Figure