Machine learning and cell study link shikonin to Parkin-dependent activity in hypoxic triple-negative breast cancer

Shikonin induced necroptosis and apoptosis in hypoxic MDA-MB-231 triple-negative breast cancer cells in a preclinical study that combined machine learning with laboratory experiments. The reported mechanism involved Parkin-mediated ubiquitination and degradation of HIF-1α, a protein linked in the experiments to the hypoxic adaptation of these tumor cells. The study was published in BMC Cancer as an open-access version subject to further edits.

The analytical component used a combined dataset of necroptosis-related genes and genes associated with a hypoxia-related risk score from 1,092 breast cancer patients. A machine-learning model was interpreted using SHAP, or SHapley Additive exPlanations, to assess feature importance. PELI1, an E3 ubiquitin protein ligase, and TNFRSF10B, a member of the tumor necrosis factor receptor superfamily, emerged as leading predictive contributors. The investigators then selected shikonin for cell testing because of its previously established role as a necroptosis inducer; the model was not reported to have identified the compound itself.

Laboratory experiments centered on MDA-MB-231 cells, including assessments of cell viability, protein expression associated with necroptosis, apoptosis and transcriptomic changes. RNA-sequencing results indicated that hypoxic cells relied on adaptive pathways such as glycolysis, erythropoiesis and angiogenesis, while shikonin disrupted these adaptations as cell-death processes were activated. Co-immunoprecipitation and ubiquitination assays examined interactions involving Parkin and HIF-1α. In parallel, siRNA-mediated Parkin knockdown followed by CCK-8 assays was used to assess whether shikonin-induced cytotoxicity depended on Parkin. Together, these results support a cell-culture mechanism in which Parkin promotes HIF-1α degradation as shikonin triggers necroptosis and apoptosis.

The evidence is preclinical and limited to one breast cancer cell line; the reported work does not establish safety or effectiveness in patients. The authors described shikonin as a candidate hypoxia-selective agent and proposed that combining it with approaches targeting PELI1 and TNFRSF10B might enhance efficacy, but that combination strategy was not tested in the reported experiments.

Subject of Research: Preclinical assessment of shikonin and Parkin-mediated cell death in hypoxic triple-negative breast cancer cells

Article Title: Shikonin as a Parkin-dependent hypoxia-selective agent against triple-negative breast cancer identified by machine learning

Article References: Niu, Z., & Niu, X. (2026). Shikonin as a Parkin-dependent hypoxia-selective agent against triple-negative breast cancer identified by machine learning. BMC Cancer. https://doi.org/10.1186/s12885-026-16675-y

Image Credits: AI Generated

DOI: 10.1186/s12885-026-16675-y

Keywords: shikonin, triple-negative breast cancer, hypoxia, Parkin, HIF-1α, necroptosis, apoptosis, machine learning, PELI1, TNFRSF10B, Parkin-dependent, hypoxia-selective