Genome build information is an essential part of genomic track files

Chakravarthi Kanduri, Diana Domańska, Eivind Hovig, Geir Kjetil SandvebioRxiv 2017 · DOI

Abstract

Genomic locations are represented as coordinates on a specific genome build version, but the build information is frequently missing when coordinates are provided. It is essential to correctly interpret and analyse the genomic intervals contained in genomic track files. Here, we demonstrate that this crucial metadatum (or rather datum) is often isolated from the genomic track files in public repositories and journal articles, which could be a major time thief. We propose best practices to ensure that genome build version is always carried along with genomic track files. Although not a substitute to the best practices, we also provide a tool to predict the genome build version of genomic track files.

Introduction

In bioinformatics, genomic locations (start and end positions) are stored in formats such as BED, GFF, or WIG. However, these numerical values are relative to a specific reference genome version, known as a "build" (e.g., hg19 vs. hg38). Coordinates without a build version are analogous to house numbers without a street name; they are biologically meaningless and lead to "silent errors" where analysis tools produce incorrect results without warning.

The Metadata Detachment Gap

A significant issue in current data sharing practices is "metadata detachment," where build information is stored on a webpage or in a separate metadata file rather than within the data file itself. When researchers download or share these files, the critical context is often lost.

An audit of major public repositories reveals that even when metadata requirements are strict, the information rarely persists within the downloadable files. As shown in Table 1, representative files from leading repositories frequently lack build information in both the filename and the file header.

Table 1: Detachment of genome build information in major public repositories

Quantifying the Risk

The consequences of this detachment are severe. Between the human genome builds hg19 and hg38, approximately 99.4% of 1kb genomic bins are incompatible. Integrating data across these builds without proper coordinate "liftover" results in erroneous biological conclusions.

The researchers conducted a large-scale audit of NCBI GEO and ENCODE to quantify the presence of build information. The results, summarized in Table 2, highlight a systemic failure: while 46% of BED files in GEO included build info, only 0.6% of GFF files did. Most strikingly, none of the 26,503 ENCODE BED files checked contained build information in their filenames or headers.

Table 2: Frequency of genome build information within filenames and file headers

A Predictive Solution

To address the problem of "orphan" datasets, the authors developed a "Genome Build Predictor." This tool, available as a web application and an R package, analyzes the distribution of coordinates and chromosome names to statistically predict the likely genome build. In testing, the tool correctly identified the build for 98.2% of ENCODE files. While highly effective, the authors emphasize that prediction is a last resort and not a substitute for proper documentation.

Conclusion and Recommendations

Genome build information should be treated as an essential part of the data, not optional metadata. The authors propose a new standard where build information is an obligatory field within the header of every genomic track file (e.g., using a #genome=hg38 comment line). Adopting such standards across repositories and tool development is crucial for ensuring the reproducibility and accuracy of genomic research.

Summary updated

Abstract

Abstract Genomic locations are represented as coordinates on a specific genome build version, but the build information is frequently missing when coordinates are provided. It is essential to correctly interpret and analyse the genomic intervals contained in genomic track files. Here, we demonstrate that this crucial metadatum (or rather datum) is often isolated from the genomic track files in public repositories and journal articles, which could be a major time thief. We propose best practices to ensure that genome build version is always carried along with genomic track files. Although not a substitute to the best practices, we also provide a tool to predict the genome build version of genomic track files.
BibTeX
@misc{kanduri2017genome,
  author = {Chakravarthi Kanduri and Diana Domańska and Eivind Hovig and Geir Kjetil Sandve},
  title = {{Genome build information is an essential part of genomic track files}},
  year = {2017},
  doi = {10.1101/120923},
  howpublished = {bioRxiv},
}
Figure