Average Overlap Frequency: A simple metric to evaluate staining quality and community identification in high dimensional mass cytometry experiments

Abstract

High dimensional cytometry now allows measurement of over 50 parameters in a single sample, and is typically visualized using sophisticated dimensionality-reducing methods and analyzed with automated clustering algorithms. While these tools facilitate the identification and presentation of key findings, it remains challenging to effectively monitor and report the staining quality of individual markers. We present the Average Overlap Frequency (AOF), a simple and efficient metric to evaluate and quantify the robustness of staining and clustering quality in high-dimensional data. We leverage the AOF to compare and determine the optimal storage conditions for stained whole blood samples prior to mass cytometry analysis. We also show that the AOF can be easily incorporated as part of automated analysis pipelines in large scale immune monitoring studies and used to flag and exclude samples with poor staining quality. We propose that the AOF may be incorporated as an essential quality control metric to better identify and report the underlying sample quality in all CyTOF and other high-dimensional cytometry experiments.

Keywords: Bioinformatics; Computational biology; CyTOF; Flow cytometry; High dimensionality; Mass cytometry; Single cell.