Personalized medicine for severe asthma: Machine learning guides biologic treatment choices.

<bold>Introduction</bold>: While biological therapies offer hope for managing severe asthma, selecting the best option for each patient remains challenging. This abstract unveils a novel machine learning (ML)-based decision support system designed to personalize biological treatment selection. <bold>Methodology:</bold> In a retrospective analysis of 129 patients, we explored using ML to personalize medication choices. Patient profiles, including demographics and asthma severity metrics, served as input features. Supervised learning models predicted both improvement status (improved/not improved) and improvement percentage for various metrics after using different biological therapies. <bold>Results:</bold> Our model predicted improvement status with up to 92% accuracy, highlighting exacerbation frequency as the strongest predictor (explaining 58% of variation). The model showed promise in suggesting potentially more effective biological options, although real-world validation is needed. Notably, patient characteristics and medication history were crucial for predictions. <bold>Conclusion:</bold> These preliminary results demonstrate the potential of ML to predict individual responses to different biological therapies and recommend potentially more effective options, paving the way for personalized asthma treatment. However, larger datasets are necessary to solidify its effectiveness in real-world scenarios.