Somatic retinoblastoma 1 (RB1) loss is prevalent across different cancer types and is enriched in treatment-refractory tumors, such as castration-resistant prostate cancer (CRPC) and small-cell lung cancer, but cannot be considered as a direct druggable target. In this study, we revealed that the close proximity of nudix hydrolase 15 (NUDT15) and RB1 may result in their common somatic codeletion or epigenomic cosilencing in different cancer types and subsequent significant positive correlations of their expressions at the bulk transcriptional and single-cell levels. With clinical CRPC samples, co-loss of RB1 and NUDT15 were commonly observed (14 out of 21). Due to the contribution of NUDT15 deficiency to thiopurine-induced toxicity, exploiting a vulnerability conferred by RB1-NUDT15 loss raised the possibility of repurposing thiopurine (e.g., mercaptopurine) for precise therapeutics. A positive relationship between RB1/NUDT15 ploidy score and mercaptopurine drug sensitivity was found in 543 cancer cell lines. Experimentally, knocking-down NUDT15 sensitizes the cancer cell lines to mercaptopurine treatment by inhibiting cell cycle progression and increasing apoptosis, but does not induce mercaptopurine-related leucopenia in xenograft model. Our study elucidates the molecular basis for precise mercaptopurine therapy in RB1-deficient tumors and demonstrates how leveraging collateral lethality alongside drug repurposing uncovers targetable vulnerabilities in stratified patient cohorts.© 2025 The Author(s). MedComm published by Sichuan International Medical Exchange & Promotion Association (SCIMEA) and John Wiley & Sons Australia, Ltd.