A Deep Learning Approach for Predicting Multiple Sclerosis.

Area of Science:

• Biomedical informatics
• Computational neuroscience
• Genetics

Background:

• Accurate diagnosis of multiple sclerosis (MS) is crucial for effective treatment.
• Gene expression profiling offers potential biomarkers for neurological diseases.
• Developing robust predictive models for MS diagnosis remains a challenge.

Purpose of the Study:

• To propose a deep learning model for predicting multiple sclerosis diagnosis.
• To enhance prediction accuracy and reduce model complexity using a single hidden layer with regularization.
• To identify key gene expression features for improved diagnostic performance.

Main Methods:

• Development of a single hidden layer artificial neural network with a regularization term.
• Application of a dimensionality reduction technique for feature selection from 74 gene expression profiles.
• Utilizing the analysis of variance (ANOVA) test to compare model performance.

Main Results:

• The proposed deep learning model achieved higher prediction accuracy than four conventional machine learning techniques.
• The model exhibited lower prediction loss, indicating improved performance.
• Feature selection via dimensionality reduction identified the most relevant gene expression profiles.

Conclusions:

• The proposed artificial neural network model is effective for predicting multiple sclerosis diagnosis.
• Regularization in the hidden layer successfully prevented overfitting and reduced model complexity.
• Gene expression data, when analyzed with advanced machine learning, holds significant potential for MS diagnostics.