Fig. 3: Batch normalization functions.

A Batch normalization using bead-like (BL) normalization. Markers of healthy control were used to generate plate-specific slopes to fit the universal healthy reference data distribution. To minimize batch effects, mean expressions of protein markers from each sample on the plate were scaled according to each plate-specific slope as in the bead standardization procedure. B The effect of five normalization functions: meanshift (MSFT), meanshift bulk (MSFTB), variance (VAR), z-score (Z), and bead-like (BL) on the data distribution of one representative healthy control sample. The changes in mean expression and variance values of 36 consensus markers were visualized using heatmaps where the colors were correlated to signal intensity. C Density plot of 8 consensus protein markers from healthy samples (n = 3) pre- and post-batch normalization using each five normalization functions (See Supplementary Fig. 3 for the distribution of all 36 consensus markers). D Correlation analysis between bead and batch normalized signals using each of the five normalization functions assessed by mean expression, variance values, and peak intensity: MSFT, MSFTB, VAR, Z, and BL normalizations. E Visualization of batch effects in 50 healthy anchors normalized by each five normalization functions using UMAP. Note that points represent single cells and were colored according to respective cohorts. Source data are provided as a Source Data file.