Fig. 3: Application of ADTnorm in analyzing human hematopoietic progenitor study.

A. Hematopoietic pilot study to determine the optimal antibody concentration in the titration cocktail. ADTnorm assigns stain quality scores to each protein marker per sample, allowing the identification of markers that exhibit successful staining and positive population signals. B. Among protein markers with detectable positive populations, stain qualities vary with antibody concentration, depending on cell sorting groups. The sample size for each box plot is the total number of protein markers with a staining quality score larger than two. C. Using stain quality scores, ADTnorm identified 82 protein markers with strong evidence of successful staining, 70 of which overlapped with the 132-plex protein panel in the original study. D. Density distribution of protein markers recommended by ADTnorm but not included in the original publication. The remaining 11 protein markers are presented in Supplementary Fig. 24. E. Comparison of stain quality scores for the 132-plex proteins panel in the original study. The sample size for the box plots is 70 for ADTnorm Recommended and 62 for Others in Original Paper. Box boundaries of (B) and (E) represent the 25th and 75th percentiles. The center line indicates the median. Whiskers extend to the largest and smallest points within 1.5x the interquartile range from the 25th or 75th percentile. Data points beyond the whiskers are outliers. F. UMAP visualization comparing protein data processed with Arcsinh transformation and ADTnorm normalization regarding batch correction and cell type separation. G. Heatmap depicting the percentage of positive populations across the hematopoietic cell type trajectory. The complete heatmap for all 132 proteins is in Supplementary Fig. 25B. H. Heatmap showing log₂-fold-change of totalVI-normalized counts relative to the mean value per protein marker. G, H use the same hematopoietic cell type order as in the original study. The protein markers are ordered by hierarchical clustering based on ADTnorm positive population percentages. The complete heatmap ordered by hierarchical clustering of the totalVI is provided in Supplementary Fig. 25D. Red squares highlight the differences between the protein marker positive proportion obtained from ADTnorm and log₂-fold-changes of the totalVI normalized counts used in the original study.