Despite the approved use of mTOR inhibitors (rapalogs) in treating advanced kidney cancers, therapeutic benefit varies and predictive biomarkers are lacking. Furthermore, intra-tumor heterogeneity questions the feasibility of genome-based targeted therapies. Here, using an ultra-deep targeted-exome (~500x) sequencing approach, we determined genetic foundation underlying long-term response (>20 months) to rapalogs. Complete loss of TSC1 or TSC2, and an activating mutation of mTOR were discovered. Interestingly, evolutionarily conserved, clustered activating mutations of mTOR occur in ~5% of KIRC TCGA (kidney cancer the cancer genome atlas) patients. The presence of predicted mTOR activating mutations in other cancers implicates rapalog treatment benefit on select patients. Multiregional genomics discovered pathway convergence through gain-of-function of mTOR and loss-of-function of TSC1 in spatially separated regions of the same tumor. Hence, pathway convergent mutations of individual tumors may point out therapeutic vantage points for personalizing cancer care in the face of tumor heterogeneity.