A computational study of cardiac glycosides from Vernonia amygdalina as PI3K inhibitors for targeting HER2 positive breast cancer.

The PI3K/Akt pathway plays a crucial role in regulating a broad network of proteins involved in the proliferation of HER2-positive breast cancer. The ethyl acetate fraction of Vernonia amygdalina, which contains cardiac glycosides, has been shown to reduce the expression of PI3K and mTOR. However, the specific cardiac glycoside compounds with significant potential as PI3K inhibitors have yet to be clearly identified. This study employs machine learning to perform virtual screening of cardiac glycosides from V. amygdalina against the p110 subunit of PI3K. Initially, Lipinski's Rule of Five was used to filter the PIK3CA inhibitor database via KNIME software. Subsequently, QSAR modeling was conducted using KNIME's machine learning platform, employing six different algorithms. Cardiac glycosides from V. amygdalina were then evaluated using the best-performing QSAR model. The top three compounds identified underwent molecular docking and molecular dynamics simulations. The random forest algorithm was selected as the primary predictive model, which identified Vernoamyosides A (VG-1), Vernoniamyosides D (VG-8), and Vernoniosides A4 (VG-10) as the compounds with the highest confidence levels. Molecular docking results indicated that these three compounds exhibited stronger and more stable interactions with the PIK3CA receptor compared to alpelisib, a known PIK3CA inhibitor. Furthermore, molecular dynamics simulations revealed that VG-10 had the lowest binding free energy, as determined by MM-GBSA analysis. The findings of this study provide a foundational basis for preclinical and clinical investigations aimed at developing PI3K inhibitors derived from cardiac glycosides of V. amygdalina for the treatment of HER2+ breast cancer.