Machine learning improves detection of alpha thalassemia carriers compared to clinical features

Area of Science:

• Genetics
• Computational Biology
• Hematology

Background:

• Alpha-thalassemia is a prevalent genetic blood disorder.
• Distinguishing between alpha-plus (α⁺) and alpha-zero (α⁰) types is crucial for effective screening and patient management.

Purpose of the Study:

• To develop and assess machine learning models for classifying alpha-thalassemia carriers (α⁺ vs. α⁰).
• To identify key hematological parameters predictive of alpha-thalassemia types.

Main Methods:

• Analysis of a dataset comprising 956 cases with hematological parameters.
• Application of feature selection techniques to identify predictive markers.
• Training and comparison of five machine learning models, including ensemble methods.

Main Results:

• The stacking ensemble model achieved the highest performance with 94% accuracy and a high F1-score.
• Key predictors identified include red blood cell (RBC) count, mean corpuscular volume (MCV), mean corpuscular hemoglobin (MCH), and mean corpuscular hemoglobin concentration (MCHC).
• Strong interrelationships were observed among RBC indices, with moderate associations for platelet (PLT) and white blood cell (WBC) parameters.

Conclusions:

• Machine learning, especially ensemble methods, can significantly enhance the detection of alpha-thalassemia carriers.
• The developed models offer a flexible framework for screening and may support personalized approaches in future research.