Κατεύθυνση Βιοπληροφορική-Υπολογιστική Βιολογία
Library of the School of Science

2024-04-15

2024

Rammenou Elli

Βασιλική Οικονομίδου, Αναπληρώτρια Καθηγήτρια Τμήμα Βιολογίας, ΕΚΠΑ

Πρόβλεψη της σοβαρότητας των παρενεργειών των εμβολίων κατά της CoViD-19 με χρήση αλγορίθμων μηχανικής μάθησης

Greek

Prediction of severity of CoViD-19 vaccines using machine learning algorithms

The COVID-19 pandemic has presented an unprecedented global health challenge, leading to the rapid development and emergency use of vaccines against the disease. Despite their efficacy, vaccines may induce adverse reactions, necessitating robust tools for predicting and managing their severity. In this study, we developed a machine learning classification algorithm aimed at predicting the level of severity of COVID-19 vaccine reactions using data retrieved from the EudraVigilance database. Our methodology involved training various classification algorithms on subsets of data with increasing sizes to evaluate performance across different sample sizes. Among these algorithms, Random Forest and XGBoost exhibited the most promising performance, with XGBoost demonstrating a slight advantage in predictive accuracy. Subsequently, we applied SHAP (SHapley Additive exPlanations) analysis to the trained XGBoost model to elucidate feature importance. Our findings reveal age as the most critical predictor, indicating that older individuals are more prone to experiencing severe vaccine reactions. Furthermore, specific symptoms such as chest pain, hypersensitivity, vomiting, dyspnoea, and seizures, along with the Moderna vaccine, emerged as significant factors associated with heightened severity of reactions. The implications of our model are profound, offering insights into demographic and symptomatic factors influencing vaccine reactions. By identifying high-risk groups and associated symptoms, healthcare professionals can prioritize monitoring and intervention strategies, potentially mitigating adverse outcomes. Moreover, the integration of machine learning techniques with pharmacovigilance databases like EudraVigilance presents a powerful tool for real-time surveillance and risk assessment in vaccine safety monitoring efforts. In conclusion, our study underscores the importance of leveraging machine learning algorithms to enhance our understanding of vaccine safety profiles amidst the COVID-19 pandemic and contributes to the ongoing efforts to optimize vaccination strategies and ensure the well-being of populations worldwide.

Science

machine learning, classification, covid-19, vaccines, bioinformatics, databases

Yes

3

Yes

181

138

File access is restricted until 2025-04-17.

https://pergamos.lib.uoa.gr/uoa/dl/object/3396535