| Evidence |
We first used principal component analysis to summarize the expression heterogeneity in each case (Methods) to better understand the composition of each sample. As expected, this revealed complex relationships among clusters (such as partially overlapping expression signatures), and multiple sources of heterogeneity in all samples, including variable expression of known hematopoietic cell-type markers (e.g. CD3D (T-cells), CD79A, or CD19 (B-cells), and HBA1 (erythrocytes)), cell cycle genes (e.g. TUBA1B, TOP2A), markers of myeloid lineage (e.g. AZU1, ELANE, MPO, PRTN3), mitochondrial genes, and ribosomal genes (Fig. 2a, b; Supplementary Fig. 2_5, Supplementary Data 4). This indicated that the distribution of cell types is a major source of expression heterogeneity, and varies among samples, as expected. |
