Fig. 4

Association of chemokine gene expression with inferred immune cell infiltration in TCGA. a Analysis workflow to identify chemokine genes associated with immune cell abundance in TCGA tumor samples. Two parallel methods were used to infer immune cell abundance: image-based estimation from pathology slides (left) and gene signature-based estimation from RNA-seq (right). Linear regression models were used to find chemokine genes whose expression levels are associated with inferred immune infiltration. b Heatmap showing chemokines (rows) significantly associated with the pathology image-based TIL scores across different TCGA cancer types (columns). The color indicates regression coefficients from the linear models (red/orange = positive association between chemokine expression and immune infiltration). c Heatmap showing chemokines (rows) significantly associated with the gene signature-based immune scores across different TCGA cancer types (columns). The panels represent associations with CD8+ T cells, dendritic cells, and NK cell scores. The chemokines (rows) are ordered descendingly by the sum of the regression coefficients across all cancer types and all 3 immune cell types (i.e., all columns)