Figure 1

Unsupervised hierarchical clustering on APM genes correlates with patient survival. (A,B) Independent clustering in the Liu melanoma anti-PD1 cohort (n = 122) (A) and Van Allen anti-CTLA4 cohort (n = 40) (B) (discovery cohorts) using the trained APM gene signature. pheatmap function in pheatmap version 1.0.12 (https://www.rdocumentation.org/packages/pheatmap/versions/1.0.12) was used to generate the heatmaps. Each row represents a patient, and each column represents a gene. Gene expression is represented as log transformed transcripts per million (TPM) reads. BOR best overall response, CRPR complete/partial response, SD stable disease, MR mixed response, PD progressive disease. (C,D) Progression-free survival (PFS) stratified by the APM clusters in the Liu cohort (C1: n = 30, C2: n = 28, C3: n = 57, C4: n = 7) (C) and Van Allen cohort (C1: n = 18, C2: n = 11, C3: n = 7, C4: n = 4) (D) represented as Kaplan–Meier curves. P-value was computed by Log-rank test. (E–F) Proportion of patients with DCB versus NCB across the APM clusters in the Liu cohort (E) and Van Allen cohort (F). Percentage indicates fraction of patients with DCB within the cluster. Significance was calculated by two-sided Fisher’s exact test. DCB durable clinical benefit, NCB no clinical benefit.