Decoding epilepsy's molecular blueprint: Machine learning unravels transcriptomic subtypes and regulatory networks

Area of Science:

• Neuroscience
• Genomics
• Computational Biology

Background:

• Drug-resistant epilepsy (DRE) affects a significant portion of epilepsy patients, yet its molecular underpinnings remain unclear.
• Current epilepsy classifications lack molecular basis, hindering targeted therapy development.
• Limited access to brain tissue and genetic diversity complicate DRE research.

Purpose of the Study:

• To create a machine learning-guided framework for molecular classification of DRE using transcriptome data.
• To identify novel molecular subtypes of DRE beyond traditional classifications.
• To uncover key molecular pathways involved in DRE pathogenesis.

Main Methods:

• Comprehensive RNA sequencing was performed on 153 surgically resected DRE samples.
• Unsupervised clustering identified two distinct transcriptomic subtypes.
• A machine learning classification model was developed using weighted correlation networks and pathway analysis.

Main Results:

• Two molecular subtypes of DRE were discovered, distinct from current pathological classifications.
• A classification model based on four key pathways (ligand-receptor interaction, cAMP signaling, GABAergic synapse, calcium signaling) was built.
• The random forest model achieved 96% accuracy and an AUC of 0.95 for classification.

Conclusions:

• Identified molecular subtypes and pathways can serve as hallmarks for epilepsy.
• This transcriptome-based classification offers a novel molecular approach to understanding epilepsy.
• Findings provide a foundation for developing targeted therapies for DRE.