Using machine learning to predict Hemophilia A severity.

Area of Science:

• Genetics
• Biochemistry
• Computational Biology

Background:

• Hemophilia A is a rare genetic bleeding disorder caused by Factor VIII (FVIII) deficiency, primarily affecting males.
• Accurate classification of Hemophilia A severity (mild, moderate, severe) is crucial for patient management.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting Hemophilia A severity based on FVIII protein point mutations.
• To identify key features within FVIII mutations that correlate with disease severity.

Main Methods:

• Utilized machine learning algorithms including RandomForest, XGBoost, and LightGBM for classification.
• Performed feature selection analysis to identify the most predictive mutation data.
• Compared model performance against a Gaussian Naive Bayes baseline.

Main Results:

• The developed classification model achieved an accuracy of 65.5%.
• This performance significantly outperformed the baseline Gaussian Naive Bayes model (51.1% accuracy).
• The study identified relevant features contributing to severity prediction.

Conclusions:

• Machine learning models show promise in classifying Hemophilia A severity using FVIII mutation data.
• While not a replacement for diagnostic tests, this approach offers valuable insights into FVIII protein characteristics.
• Further development could enhance predictive capabilities for Hemophilia A severity.