Predicting Avatrombopag response in children with immune thrombocytopenia: A multi-view learning framework for...

Area of Science:

• Computational biology
• Pediatric hematology
• Machine learning in medicine

Background:

• Chronic immune thrombocytopenia (ITP) affects children, requiring effective treatments like avatrombopag.
• Early prediction of avatrombopag treatment response is crucial for personalized care and minimizing drug exposure.
• Challenges in predicting response include missing data, small sample sizes, and high-dimensional features in clinical data.

Purpose of the Study:

• To develop an effective computational framework for early prediction of avatrombopag treatment response in pediatric chronic ITP.
• To address challenges posed by missing data, small sample size, and high-dimensional features.

Main Methods:

• A multi-view learning framework was developed, creating diverse data views using various imputation techniques.
• A feature selection encoder was used for dimensionality reduction and interpretability.
• Multi-view fusion, co-regularization, and contrastive learning were integrated to enhance prediction accuracy and robustness.

Main Results:

• The proposed multi-view learning framework demonstrated superior performance compared to baseline models in predicting treatment response.
• Experiments on real-world clinical data validated the framework's effectiveness.

Conclusions:

• The study presents a practical and interpretable multi-view learning framework for early avatrombopag treatment response prediction in pediatric chronic ITP.
• This framework supports personalized treatment strategies and efficient decision-making in pediatric hematology.