Machine learning-based prediction of drug response in ischemia reperfusion animal model.

Myocardial ischemia is a major global contributor to mortality. While reperfusion therapy remains the most effective treatment, it paradoxically leads to myocardial ischemia-reperfusion (MI/R) injury, resulting in irreversible cardiac damage for which no effective interventions currently exist. This underscores the pressing need to unravel the pathogenesis of MI/R injury and devise new therapeutic strategies. In this study, supervised machine learning models, including logistic regression (LR), support vector machines (SVM), random forests (RF), neural networks (NN), and k-nearest neighbors (kNN), were utilized to predict treatment response. The models incorporated molecular and biochemical features to evaluate three drugs: trans-Anethole (TNA), pentoxifylline (PTX), and cyanidin-3-O-glucoside (Cy3G). The sequential forward selection (SFS) method was employed to select the most relevant features for prediction. To assess model performance, metrics such as precision, accuracy, recall (sensitivity), specificity, and the Matthews Correlation Coefficient (MCC) were analyzed for both reduced and complete models. Among the classifiers, kNN demonstrated notable performance, achieving an accuracy of 0.9156 ± 0.0242 and an average area under the ROC curve (AUC) of 0.90 across three cross-validation iterations surpassing all other classifiers. This observed performance is in line with recent literature that employs advanced computational methods in similar domains. A key advantage of our study is the use of a two-layer framework-integrating molecular signatures with biochemical markers-which can provide improved robustness and biological relevance. This multi-layer integration enhances interpretability and better reflects the multifactorial nature of MI/R injury, while supporting model generalization. Feature selection identified one molecular marker (SOX5) and two biochemical markers (dP/dtmax and cTnT) as significant predictors of drug response. This integrative approach has the potential to enhance personalized therapy for myocardial ischemia by enabling precise drug response predictions and guiding the development of targeted treatment strategies.