基于机器学习和结构导向发现用于癌症治疗的EP300选择性、口服生物利用的降解剂

EP300 and CBP are closely related transcriptional co-activators with histone acetyltransferase (HAT) activity that regulate gene expression, cell proliferation, and differentiation. In normal tissues, EP300 and CBP compensate for each other; however, CBP-deficient tumor cells become selectively dependent on overexpressed EP300, creating a synthetic-lethal therapeutic window. Despite this opportunity, designing EP300-selective inhibitors has been challenging because the EP300 and CBP HAT domains share ~90% sequence identity. In this regard, targeted protein degradation (TPD) platform provides potential opportunity to achieve selective degradation. Recent findings indicate that even subtle structural variations, such as a single surface residue near the binding interface or the positioning of a lysine, amplify into significant differences in ternary complex stability and ubiquitination efficiency for proteolysis. Accordingly, we designed and synthesized heterobifunctional degraders that selectively degrade EP300 over CBP. Here, we report the discovery of orally bioavailable EP300 degraders and their evaluation in vitro , in silico , and in vivo . We identified early leads that selectively degrade EP300 over CBP, as confirmed by western blotting, thereby establishing proof-of-concept that TPD can confer EP300 selectivity. To further improve cellular potency, we optimized the initial lead using structure-guided design informed by molecular dynamics (MD) simulations; this analysis highlighted functionally important EP300 residues and guided modifications that increased the stability of the EP300-CRBN ternary complex relative to the starting scaffold. We applied a bioinformatics framework to identify solid tumor indications responsive to EP300 degradation. Using DepMap EP300 CRISPR scores integrated with RNA expression profiles, we developed and optimized a machine-learning model that accurately distinguished sensitive from resistant cell lines. Model-based prioritization identified several indications with the highest predicted sensitivity. Experimental validation confirmed that the optimized EP300 degrader induced robust EP300 degradation and potent growth inhibition across these tumor cell lines. In mouse xenograft models, it achieved high antitumor efficacy with lower toxicity than an EP300/CBP dual inhibitor, consistent with synthetic lethality. In conclusion, this study demonstrates that selective EP300 degradation can be achieved by applying TPD modality, enabling synthetic-lethal strategy and oral bioavailability. Structure-guided design and bioinformatics-driven indication selection provided an efficient path to preclinical candidates, underscoring the therapeutic potential of Hanmi's EP300 degraders for solid cancers.