Fig. 5: Evaluation of stable channels as generalized anchors for batch normalization.

A The application of CyTOF data integration pipeline to standardize three patient PBMC samples from patients receiving CAR-T cell therapy. Batch normalization was performed using two types of generalized anchors: healthy peripheral blood mononuclear cell (PBMC) samples or stable channels. The generalized anchors were fitted to the universal reference data distribution followed by data normalization. B Identification of stable channels using a PCA-based non-redundancy score (NRS) to rank-order marker variability in three CAR-T patient samples (n = 3). The band indicates the median, the black square indicates the mean, the box indicates the first and third quartiles and the whiskers indicate ± 1.5 × interquartile range. Channels with NRS < 2 were identified. C To determine the optimal number of stable channels, the RMSD values between batch and bead normalized signals were determined by varying the number of stable channels for each sample. Minimum RMSD values were achieved with the top 3–4 stable channels. D Correlation between mean marker expression and channel stability measured by NRS values (R²~0.26). The top three stable channels: CD274, CD137, and CD127 are labeled. E Correlation between batch and bead normalization performance for three CAR-T cell samples using healthy control samples or top three stable channels as generalized anchors. Each dot represents the mean expression of 30 consensus markers normalized by bead (x-axis) or by batch normalization (y-axis). The correlation values were computed across 30 consensus markers for each sample. Batch normalization using stable channels normalized by either MSFTB or BL normalizations achieved comparable performance to that using external anchors with R² values > 0.9. Source data are provided as a Source Data file.