The machine learning-based predictor to identify putative COVID-۱۹-like host jumping viruses

Zoonotic viruses, capable of crossing the species barrier from animals to humans, pose significant threats to global health, as demonstrated by outbreaks such as SARS, MERS, and COVID-۱۹. Early identification of these viruses is critical for pandemic preparedness and containment strategies. Machine learning has increasingly been utilized in healthcare and virology to enhance predictive modeling. This study presents a machine learning-based approach for assessing the zoonotic potential of viruses by analyzing key biological features, including protein stability, RNA energy, protein folding success, and codon usage patterns. A curated dataset of viral spike protein sequences was compiled, encompassing both zoonotic and non-zoonotic viruses. To address class imbalance, the Synthetic Minority Oversampling Technique (SMOTE) was applied, ensuring a balanced representation of both categories. The dataset was then normalized using z-score transformation to standardize feature distributions. A logistic regression model was trained and optimized through hyperparameter tuning to achieve an optimal balance between sensitivity, specificity, and accuracy. The model was evaluated using multiple validation strategies, including an independent testing dataset, to assess its robustness and generalizability. Results indicate that the model achieved a prediction accuracy of ۷۸.۵۷%, demonstrating its reliability in distinguishing zoonotic from non-zoonotic viruses. The high specificity ensures that the model effectively minimizes false positives, while sensitivity enables the detection of potential zoonotic threats. The interpretable nature of logistic regression makes the model transparent and applicable to real-world decision-making. By providing a systematic and data-driven approach, this study contributes to the early identification of emerging zoonotic threats, ultimately enhancing global health preparedness and response strategies.

کلیدواژه ها:

COVID ، ۱۹ Logistic regression Machine learning SMOTE Zoonotic viruses