Rising incidence of early-onset prostate and other solid tumors underscores the need for experimental systems that model how normal tissues traverse premalignant states, acquire mutations, and become therapy-responsive malignancies under authentic immune and stromal pressures. A major barrier has been the lack of tractable in vivo platforms that enable genome-wide discovery while preserving continuous tumor evolution without catastrophic chromosomal instability. To address this, the Bose Lab developed Stochastically Emergent Tumors (SETs), an organoid-derived in vivo evolution engine that redefines discovery for early-onset and understudied patient groups. In this system, mismatch-repair deficiency is induced in non-malignant human organoids, which are passaged to accumulate stochastic point mutations and transplanted into mice to permit malignancy to emerge under physiologic selection. SETs evolve primarily through high-resolution point mutations rather than broad copy-number changes, yielding bioinformatically tractable clonal dynamics ideally suited for whole-genome driver discovery and machine learning. Compared with conventional xenografts, SETs display greater intertumoral heterogeneity and reproducible recovery of sensitizing and resistance alleles under therapeutic pressure. As proof of principle in prostate cancer, endocrine therapy applied to SET pools recovered known determinants of androgen-pathway sensitivity and uncovered new drivers. Loss of ZFHX3, typically obscured within a multigene suppressor locus in bulk cohorts, promoted luminal histology and sensitized tumors to androgen-receptor inhibition in vivo, whereas KMT2D or CIC alterations mediated resistance. Consistent with model predictions, ZFHX3 loss in patients correlated with significantly improved survival, a finding comparable in magnitude to the most favorable molecular subtypes of advanced prostate cancer. SETs also quantify evolutionary thresholds: in a Pten-null background, approximately 900 coding mutations accumulated over 208 days were sufficient for malignant transformation in half of grafts. This defines a measurable axis linking mutation burden, genotype, and tumor incidence. Because SETs generate neoantigen-rich point-mutation landscapes, they can be extended to immunocompetent hosts to study tumor–immune coevolution, early T-cell surveillance, macrophage-mediated immune exclusion, and myeloid checkpoints that enable immune escape. In summary, SETs provide a scalable, evolution-aware platform that connects mutational dynamics to therapeutic vulnerability, enabling identification of lineage- and ancestry-associated drivers, immunopreventive targets, and biomarkers of early-onset prostate cancer. Citation Format: Ruhollah Moussavi-Baygi, Matthew Ryan, Woogwang Sim, Samuel Hoelscher, Valbona Luga, Arun Chandrakumar, Lore Hoes, Junghwa Cha, Young Sun Lee, Katelyn Herm, Ben Doron, Danika Bakke, C.K. Cornelia Ding, Bradley Stohr, Peng Jin, Tejasveeta Nadkarni, Xiangyi
