Fig. 2

3D cluster analysis reveals numerous potentially functional rare mutations. a 3D cluster analysis identified a large number of statistically significant, yet rarely mutated residues (mutated one to three times in our dataset). The residues were binned by the number of mutations in each residue. The mutation counts for the single-residue hotspots also contain a small fraction of silent, nonsense, and splice-site mutations identified by Chang et al. 2016 [6]. b Genes with the highest number of residues in 3D clusters. c Genes with the highest frequency of tumor samples with mutations clustered in 3D structures across all cancer types. d Per-residue comparison of significance as in single-residue hotspot (vertical axis) and 3D cluster (horizontal axis). Many residues were hotspots as well as parts of 3D clusters (upper right quadrant), but some were detected only as part of 3D clusters (bottom right quadrant). e Number of residues (upper panel) and percentage of samples (bottom panel) with hotspots and 3D clusters per cancer type (see full cancer type names in the Abbreviations section). The category of a sample was assigned based on the lowest category if it had mutations that belonged to different categories