Identification of novel selective estrogen receptor degraders (SERD) via physics-based and deep-learning-based virtual screening and Bioassys.

Breast cancer is the most common malignant tumor among women, most of which are ERα(Estrogen Receptor alpha) positive. SERDs(Selective Estrogen Receptor Degraders), such as Fulvestrant(the first SERD), can induce degradation of this receptor, leading to overcome the acquired endocrine resistance. However, only two SERDs (Fulvestrant and Elacestrant) are currently clinically approved, highlighting an urgent demand for novel, oral, and more potent alternatives. In this study we developed a multi-tiered virtual screening combing physics-based docking methods (Glide) with deep-learning-based docking methods (Karmadock and Carsidock) to identify SERDs with novel scaffold. After ADMET and MM-GBSA screening, four purchasable candidate compounds were selected for biological evaluation in three cell lines. Among them, two compounds exhibited significant anti-proliferation activity aganist ER-positive cells. The fingerprint analysis also revealed their structural novelty, which are distinct from the known SERDs. Further study indicated F0840-0093 could directly bound to ERα and induced its proteasomal degradation (mimicking Fulvestrant). In summary, our work not only provided a feasible virtual screening approach in drug discovery but also identified some compounds, particularly F0840-0093, which can be a promising lead with new chemical scaffold for further optimization and development as SERDs.