Non-negligible occurrence of errors in gender description in public data sets

Abstract

Due to advances in omics technologies, numerous genome-wide studies on human samples have been published, and most of the omics data with the associated clinical information are available in public repositories, such as Gene Expression Omnibus and ArrayExpress. While analyzing several public datasets, we observed that errors in gender information occur quite often in public datasets. When we analyzed the gender description and the methylation patterns of gender-specific probes (glucose-6-phosphate dehydrogenase [G6PD], ephrin-B1 [EFNB1], and testis specific protein, Y-linked 2 [TSPY2]) in 5,611 samples produced using Infinium 450K HumanMethylation arrays, we found that 19 samples from 7 datasets were erroneously described. We also analyzed 1,819 samples produced using the Affymetrix U133Plus2 array using several gender-specific genes (X (inactive)-specific transcript [XIST], eukaryotic translation initiation factor 1A, Y-linked [EIF1AY], and DEAD [Asp-Glu-Ala-Asp] box polypeptide 3, Y-linked [DDDX3Y]) and found that 40 samples from 3 datasets were erroneously described. We suggest that the users of public datasets should not expect that the data are error-free and, whenever possible, that they should check the consistency of the data.