A simple strategy for sample annotation error detection in cytometry datasets

The purpose of the experiment was to identify samples that had been labelled with the incorrect participant metadata by identifying samples that were positive or negative for HLA-A*02 and HLA-B*07 alleles by CyTOF analysis of whole blood stained with a 33-marker panel designed to identify major immune cell types, then comparing longitudinal samples from the same participant.
Conclusion:
3 out of 123 samples did not match the HLA type of their longitudinal pairs. This error rate is comparable to similar datasets analyzed via genomic techniques that are publicly available. We concluded this method is useful for quality control of longitudinal cytometry data, and could be useful for cross-sectional data when paired with another technique. Samples were processed within 24 hours of collection and stored at 4C to prevent sample degradation. Samples were stained with aliquots of a staining cocktail master batch to minimize batch to batch variation. Staining intensity was quantified and visually assessed to be consistent over time. Bridging of reagents was conducted when new master batches were prepared and aliquoted. Control samples including cryopreserved human PBMC’s, and VeriCells (Biolegend) were periodically stained throughout the study. EQ Beads (Fluidigm) were included in each sample and were used to normalize deviations in intensity.