以结构和生物信息学驱动开发一种选择性、口服生物利用度良好的HER2小分子抑制剂（HM100714）用于癌症治疗

HER2 alterations, encompassing overexpression, amplification, and activating or other mutations, occur across multiple solid tumors and can drive cancer cell growth and metastatic progression. To address these alterations, HER2-targeted therapeutics are under development and clinical evaluation for breast, gastric, cholangiocarcinoma, colorectal, bladder, and non-small-cell lung cancers. However, several existing HER2 inhibitors also inhibit wild-type EGFR, leading to EGFR-related toxicities such as skin rash and diarrhea. Collectively, these considerations support the development of HER2-selective inhibitors that spare wild-type EGFR. Herein, HM100714 covalently targeted HER2 Cys805 and selectively inhibited HER2 while minimizing EGFR-mediated toxicities. To elucidate the improved tolerability, we performed molecular dynamics (MD) simulations of the inhibitor-protein complex. Beyond the covalent linkage to Cys805, MD revealed interactions with HER2-specific residues that are absent in wild-type EGFR, providing a structural rationale for HER2 selectivity and reduced EGFR liability. We further explored potential clinical indications using a bioinformatics framework. We trained a machine-learning model using published in vitro datasets of HER2 inhibitors and transcriptomic information from the Cancer Cell Line Encyclopedia to predict sensitivity to the drug. Further in vitro validation was done using an in-house panel of 75 cell lines. Extrapolation of the model to patient-derived datasets allowed prioritization of cancer indications predicted to be effective. These results will inform the IND-enabling studies and documentation. HM100714 inhibited HER2-altered enzymes and cell lines and showed a DMPK profile compatible with oral dosing. Guided by bioinformatics-driven indication selection, we evaluated the antitumor activity as a single agent across multiple HER2-altered xenograft models. Oral administration of HM100714 resulted in statistically significant antitumor efficacy in the NCI-N87 xenograft model, which exhibits HER2 overexpression, and in the Ba/F3 xenograft model harboring the HER2 A775_G776insYVMA mutation. We further assessed efficacy in leptomeningeal metastasis and brain metastasis models and characterized the toxicity profile to define a safety margin in support of the IND-enabling studies. In conclusion, HM100714, an orally bioavailable small-molecule inhibitor, demonstrated robust efficacy against HER2-altered tumors with minimal EGFR-related toxicities. Moreover, structure-based analyses and bioinformatics provided key data that streamline clinical readiness. These preclinical results support HM100714 as a promising therapeutic candidate for HER2-altered cancers.