Fig. 1

Meta-analysis-based identification of prognostic genes: Genes associated with A survival and B lymph node metastasis (LNM): i Heatmap of cox regression z-scores for all survival-associated genes (Rows) in 16 HNC gene expression study datasets (Columns). Z-scores indicate the association of genes with survival (overall survival, progression-free survival, or distant metastasis-free survival) within each study. Genes are ordered by meta-z-score (Right sidebar), summarizing their association with survival across studies. Bar plots indicate the percentages of HPV positive (HPV + ve) and negative (HPV − ve) cases (HPV status), as well as the percentages of HNCs that occurred within each subanatomic region (Site). White regions of bar plots indicating missing data. Horizontal annotation bars labeled “Includes HPV + ve OPC” indicating studies that include HPV + ve oropharyngeal cancer (OPC) (i.e., studies in which HPV status was a potential confounding factor). “Incomplete data” indicates studies in which clinical annotation was insufficient to determine if the study included HPV + ve OPC. Gene symbol labels highlight functionally significant genes that are mentioned in the main text. ii Gene set enrichment analysis (GSEA) of survival-associated genes: Bar plot showing normalized GSEA enrichment scores of gene sets with the strongest positive and negative associations (The top ten of each), out of a total of 18,993 curated gene sets accessed from the Molecular Signatures Database (MSigDB). Depth of color indicates the negative log Benjamini-Hochberg-adjusted p-value for enrichment of gene sets. B Meta-analysis of genes associated with LNM: i Heatmap of scaled mean gene expression differences between primary tumors of LNM + patients and those of LNM0 patients, of all genes that were significantly associated with LNM (rows), in 20 study datasets (columns). Genes are ordered by meta-z-score (right sidebar), which summarize their association with LNM across studies. Heatmap annotations are equivalent to those shown for survival-associated genes in A. ii Bar plot illustrating GSEA of LNM-associated genes. Enrichment scores are shown for the top ten gene sets with the strongest positive and negative associations with LNM