The Doppelgänger Effect: Hidden Duplicates in Databases of Transcriptome Profiles

Area of Science:

• Genomics
• Bioinformatics
• Cancer Research

Background:

• Whole-genome analysis of cancer specimens is increasingly common.
• Re-using and sharing specimens can lead to duplicate expression profiles in public databases.
• Undetected duplicates, termed "doppelgängers," can negatively impact data re-analysis and research reproducibility.

Purpose of the Study:

• To propose and validate a method for accurately matching duplicate cancer transcriptomes.
• To address the challenge of identifying duplicates when nucleotide-level sequence data are unavailable.
• To establish a standard procedure for screening transcriptome databases for potential duplication.

Main Methods:

• Developed a method to match duplicate cancer transcriptomes using expression profiles.
• Applied the method to diverse datasets including ovarian, breast, bladder, and colorectal cancer microarrays.
• Validated the method by matching microarray and RNA sequencing profiles from The Cancer Genome Atlas (TCGA).

Main Results:

• Identified probable duplicate profiles in over 50% of the analyzed studies.
• Duplicates were found across different continents, technologies, publication years, and within TCGA.
• The method effectively matched samples profiled by different microarray technologies and by microarray versus RNA sequencing.

Conclusions:

• A robust method for detecting duplicate cancer transcriptomes is essential for accurate genomic data analysis.
• Doppelgänger-checking should be integrated into standard procedures for combining multiple genomic datasets.
• The doppelgangR Bioconductor package is provided to facilitate the screening of transcriptome databases for duplicates.