Bayesian Network as a Decision Tool for Predicting ALS Disease.

Area of Science:

• Biomedical informatics
• Neuroscience
• Proteomics

Background:

• Early clinical diagnosis of amyotrophic lateral sclerosis (ALS) presents significant challenges.
• Blood tests offer a cost-effective and time-efficient diagnostic alternative.
• Machine learning (ML) is increasingly utilized to understand complex diseases like ALS.

Purpose of the Study:

• To develop and evaluate ML models for predicting ALS using blood plasma protein levels and personal features.
• To compare the performance of different ML methods, specifically Bayesian Networks, for ALS prediction.
• To identify key predictive factors for ALS onset and progression.

Main Methods:

• Utilized Bayesian networks and other machine learning algorithms.
• Analyzed blood plasma protein levels and independent personal characteristics of patients.
• Compared model performance using metrics such as accuracy and Area Under the Curve (AUC).

Main Results:

• Bayesian Networks achieved superior performance with an accuracy of 0.887 and an AUC of 0.970.
• Confirmed that sex and age are significant predictive variables for ALS.
• Identified a high probability of onset involvement in ALS patients and associations with other chronic/neurological diseases.
• Observed elevated Parkin protein levels in ALS patients compared to controls, contrasting with Parkinson's disease patients.

Conclusions:

• Machine learning, particularly Bayesian Networks, shows promise for improving early ALS diagnosis via blood biomarkers.
• Personal factors like age and sex, alongside comorbid conditions, play a crucial role in ALS.
• Parkin protein levels may serve as a potential biomarker differentiating ALS from other neurological conditions.