Fig. 6: CytofIn integration of mass cytometry dataset in the public domain.
From: CytofIn enables integrated analysis of public mass cytometry datasets using generalized anchors
A Workflow for assessing the degree of marker overlap in the FlowRepository (FR) database. A total of 44 datasets were retrieved with the PBMC tag from FR. Merging one representative panel from each dataset identified a total of 192 overlapping markers suitable for integration from a total of 808 panels. B Heatmap visualization on the degree of panel overlaps within the top 50 consensus markers based on their frequency in the retrieved panels (black: presence of the marker, red: absence of the marker). C Marker frequency ranking shows that >89% of the datasets have panel overlaps within the top 3 markers, >86% have panel overlaps within the top 5 markers while > 50% have panel overlap for all 20 markers. D, E Integration of three public mass cytometry datasets containing tumor-infiltrating leukocytes (TIL) across four different cancer histologies (red: breast, blue: glioma, green: kidney, yellow: sarcoma, gray: healthy control). Comparing pre- (D) and post-normalization (E) using CytofIn demonstrated improvements in mean expression clustering between the breast and within the glioma datasets, indicating a reduction of batch effects. See Supplementary Fig. 19 for detailed quantification of batch effect reductions between and within datasets based on the pair-wise RMSD values.
