Systems-level integration of machine learning, SHAP explainability, PCA, and network topology reveals multitarget protective actions of tannic acid against doxorubicin toxicity in pentose phosphate pathway.

Doxorubicin (DOX) is an effective chemotherapeutic agent; however, its clinical use is limited due to dose-dependent toxicities. This study aimed to investigate the potential protective effect of tannic acid (TA), a natural polyphenol with antioxidant properties, against DOX-induced alterations in oxidative pentose phosphate pathway (PPP) enzymes in rat lung tissue. Male rats were treated with DOX, TA, or a combination of both. The activity and gene expression levels of PPP enzymes (G6PD and 6PGD), antioxidant enzymes (GPx, SOD, and CAT), and oxidative stress markers (GSH and MDA) were evaluated. DOX administration significantly reduced G6PD and 6PGD activity and gene expression, decreased GSH levels, and increased MDA content. Co-treatment with TA reversed these biochemical alterations and improved antioxidant status. To elucidate the mechanistic drivers of group discrimination, an explainable random forest classifier was developed. SHAP (Shapley Additive Explanations) analysis identified 6PGD activity, G6PD/6PGD mRNA expression, GPX and CAT activities, IL-6 mRNA, GSH, and MDA as the most influential biomarkers determining DOX toxicity profiles. In addition to conventional biochemical and molecular assessments, multivariate statistical analyses (PCA), correlation network mapping, and explainable machine learning approaches (random forest with SHAP analysis) were employed to characterize the systemic oxidative and inflammatory alterations induced by DOX and their modulation by TA. The classifier achieved excellent discriminative performance, with a low out-of-bag Brier score (0.138). TA demonstrates a protective effect against DOX-induced oxidative stress and enzymatic impairment in the rat lung, suggesting its potential as a supportive therapeutic agent.