CIBERSORT (cell type identification by estimating relative subsets of known RNA transcripts)

UC microarray gene expression datasets were formatted into mixture files with patient identifiers and corresponding gene expression levels; these files were subsequently uploaded for CIBERSORT analysis according to formatting requirements (http://cibersort.stanford.edu)¹⁷. Findings were further validated using updated CIBERSORTx analysis¹⁶. Analysis of mixture files was performed using the core LM22 signature consisting of 547 genes that precisely differentiate mature human hematopoietic cells to determine relative abundances of 22 immune cell subsets including: B-cells (naïve, memory, plasma cells), T-cells (CD8, naïve CD4, memory CD4, follicular helper, regulatory, γδ), monocytes, macrophages (M0, M1, M2), dendritic cells, mast cells, eosinophils, and neutrophils. Duplicated genes were filtered based on those meeting an adjusted p < 0.05 before being input into analysis. For those genes with multiple probes meeting significance thresholds, the average expression value of the probe identifiers was calculated and used for analysis. Immune cell output was reported as relative fractions for all immune cell subsets and represented as stacked bar charts as a proportion of one hundred percent or as fold-change differences normalized to healthy control or therapy responders.