Benchmarking quanTIseq immune cell profiler

To benchmark quanTIseq, we considered the expression data sets listed in Additional file 2: Table S1, using the options reported in Additional file 2: Table S3. Normalized microarray data were downloaded from the Gene Expression Omnibus (GEO) (https://www.ncbi.nlm.nih.gov/geo) with the GEOquery R package [34 (link)]. Probes were mapped to gene symbols with the biomaRt R package [35 (link)]. In case of multiple probes mapping to the same gene symbol, the probe with the highest average expression across all samples was selected. Immune cell fractions estimated with flow cytometry, Coulter Counter, or from images of stained tissue slides were used as ground truth to validate quanTIseq. Where necessary, different functional states of an immune cell type were aggregated by summing up the corresponding cell fractions (e.g., for the Newman’s data set [17 (link)], B cells were quantified summing up the fractions of naïve and memory B cells).