The identification of targeted agents with high therapeutic index is a major challenge for cancer drug discovery. We found that screening chemical libraries across neuroblastoma (NBL) tumor subtypes for selectively-lethal compounds revealed metabolic dependencies that defined each subtype. Bioactive compounds were screened across cell models of mesenchymal (MESN) and MYCN-amplified (MYCNA) NBL subtypes, which revealed the mevalonate and folate biosynthetic pathways as MESN-selective dependencies. Treatment with lovastatin, a mevalonate biosynthesis inhibitor, selectively inhibited protein prenylation and induced apoptosis in MESN cells, while having little effect in MYCNA lines. Statin sensitivity was driven by HMGCR expression, the rate-limiting enzyme for cholesterol synthesis, which correlated with statin sensitivity across NBL cell lines, thus providing a drug sensitivity biomarker. Comparing expression profiles from sensitive and resistant cell lines revealed a TGFBR2 signaling axis that regulates HMGCR, defining an actionable addiction in that leads to MESN-subtype-dependent apoptotic cell death.