Machine learning-assisted identification of potential cancer inhibitors: Multifunctional biomineralized CaCO3-polyethyleneimine nanoparticle carriers and their application in lung cancer therapy

Abstract

Cancer, as a prevalent phenomenon among malignant tumors, has a late-stage diagnosis rate approaching half, significantly limiting treatment efficacy and imposing substantial physiological and psychological burdens on patients. Therefore, developing innovative therapeutic platforms that simultaneously optimize diagnostic accuracy, enhance therapeutic efficacy, and minimize side effects is a critical scientific challenge in the field of precision cancer medicine. In this study, we employed a simple and efficient one-pot synthesis method combined with polyethyleneimine (PEI)-mediated biomineralization technology to successfully prepare calcium carbonate-polyethyleneimine (CaCO3-PEI) composite nanoparticles. These nanoparticles exhibit high pH sensitivity and can rapidly degrade under the mildly acidic conditions that mimic the tumor microenvironment, providing potential for targeted tumor therapy. Subsequently, we encapsulated compound 1, which has potential for lung cancer treatment, within the CaCO3-PEI nanoparticles, successfully constructing the CaCO3-PEI@1 composite system. Structural analysis of this composite system was conducted through characterization techniques. In vitro cell experiments demonstrated that the CaCO3-PEI@1 composite system exhibited significant anticancer effects by effectively inhibiting cancer cell proliferation through the regulation of apoptosis-related protein Bax expression. This provides new insights and experimental evidence for the development of cancer treatment strategies. Based on the experiment and molecular docking identified activity against lung cancer, the compound 1 was used as the initiator for the machine learning model, and up to 10,000 episodes were generated, after thorough examination, it was found about two-thirds of the generated episodes were entirely different from each other, which demonstrated that the machine learning model was a powerful tool for designing of potential inhibitors against lung cancer.