Molecular Classification and Interpretation of Amyotrophic Lateral Sclerosis Using Deep Convolution Neural Networks...

Area of Science:

• Neuroscience
• Genetics
• Computational Biology

Background:

• Amyotrophic lateral sclerosis (ALS) is a neurodegenerative disease impacting motor neurons, posing classification challenges due to genetic complexity and data limitations.
• Existing deep learning models struggle with minority class accuracy and explainability in ALS molecular classification using RNA expression data.

Purpose of the Study:

• To develop and validate a deep learning framework for accurate molecular classification of ALS.
• To enhance the explainability of deep learning models in identifying disease-associated genes for ALS.

Main Methods:

• RNA expression values were converted into images using the DeepInsight similarity technique.
• A Convolutional Neural Network (CNN) was trained on these RNA expression images for ALS classification.
• Shapley Additive Explanations (SHAP) were employed to interpret the CNN model and identify relevant genes.

Main Results:

• The proposed framework achieved high accuracy in classifying ALS samples, particularly for minority classes.
• SHAP analysis successfully identified specific genes contributing to ALS molecular classification.
• The integration of CNNs and SHAP provided interpretable insights into ALS-associated genetic factors.

Conclusions:

• Deep learning, combined with explainability techniques like SHAP, offers a powerful approach for molecular classification of ALS.
• This framework can aid in discovering novel disease-associated genes, advancing our understanding of ALS.
• The study highlights the potential of machine learning in neurodegenerative disease research.