Exploring Machine Learning Techniques to Predict the Response to Omalizumab in Chronic Spontaneous Urticaria

Area of Science:

• Immunology
• Computational Medicine
• Dermatology

Background:

• Omalizumab is a primary treatment for chronic spontaneous urticaria (CSU).
• Machine learning (ML) offers potential for predicting treatment response.
• No prior studies have applied ML to predict Omalizumab response in CSU patients.

Purpose of the Study:

• To develop and validate ML models for predicting Omalizumab response in CSU patients.
• To identify clinical and demographic predictors of treatment response.
• To classify patients into early, late, or non-responders.

Main Methods:

• Analysis of data from 132 CSU outpatients.
• Training and validation of ML models using clinical and demographic characteristics.
• Two labeling methodologies based on Urticaria Activity Score over 7 days (UAS7) at 1, 3, and 5 months, including classification into early, late, and non-responders.

Main Results:

• Early responders were often hypertensive; late responders frequently had asthma and hypothyroidism.
• A positive correlation (R²=0.21) was observed between total IgE and UAS7 at 1 month.
• D-dimer and C-reactive proteins were identified as key predictors by Variable Importance Analysis (VIA). Support Vector Machine (SVM) and k-Nearest Neighbors (k-NN) showed high specificity and sensitivity for predicting late responders.

Conclusions:

• k-NN and SVM models can effectively identify treatment response.
• D-dimer and C-reactive proteins demonstrate significant predictive power for training ML models.
• ML approaches can personalize Omalizumab treatment strategies for CSU.