CompensAID: An Automated Detection Tool for Reference Errors

In flow cytometry, raw detector values do not directly represent fluorochrome abundances and must be mathematically unmixed using coefficients derived from single-stained reference controls. Inaccuracies in these controls, referred to as reference errors, can distort the estimated fluorochrome abundances, leading to skewed population distributions. However, as panel complexity and dataset sizes increase, manually inspecting all marker combinations for such artifacts becomes impractical. To address this challenge, we developed CompensAID, an open-source R-based tool that automatically flags marker combinations potentially affected by reference errors, thereby supporting quality control workflows in flow cytometry. Preprocessed data from both conventional and spectral flow cytometry were used to develop and validate the tool. CompensAID applies a density-based cutoff detection to automatically gate the negative and positive populations. The positive population is then divided into equally sized segments, after which the Secondary Stain Index (SSI) is computed for each segment. Marker combinations are flagged if the last segment yields an SSI value below -1. The tool's performance was evaluated against the consensus of five flow cytometry specialists. For the conventional dataset, 1761 out of 2240 marker combinations (79%) were unanimously classified as free of reference errors, while 24 marker combinations received a consensus score of three or higher, indicating that most flow cytometry specialists suspected reference errors. CompensAID successfully identified 23 out of these 24 marker combinations, achieving a sensitivity of 0.96. The single missed case involved a spillover spread error. In the spectral dataset, 1505 out of 1656 marker combinations (90%) were deemed free of reference errors. Of the 28 marker combinations suspected of reference errors, CompensAID flagged 21 marker combinations (sensitivity, 0.74). Five out of the seven missed marker combinations had SSI values close to the threshold. In both datasets, false positive findings were observed, often attributable to suboptimal gating or low event counts within the segments contributing to negative SSI values. CompensAID offers a robust approach for detecting marker combinations exhibiting signs of reference errors. While some false positives occur, the tool significantly reduces the burden of manual inspection. Its integration into quality control pipelines can enhance the reliability of flow cytometry data analysis. The R package is available at https://github.com/Olsman/CompensAID and will be made available through the Bioconductor platform.