Identifying Cortical Molecular Biomarkers Potentially Associated with Learning in Mice Using Artificial Intelligence.

Area of Science:

• Neuroscience
• Artificial Intelligence
• Molecular Biology

Background:

• Learning and memory involve complex molecular changes in the cortex.
• Identifying specific molecular biomarkers for learning is crucial for understanding cognitive processes.
• Previous research has linked some proteins to learning, but a comprehensive predictive panel was lacking.

Purpose of the Study:

• To identify cortical molecular biomarkers associated with learning in mice using AI.
• To develop predictive models for learning based on protein expression levels.
• To explore potential links between learning, cortical pruning, and apoptosis.

Main Methods:

• Applied machine learning (ML) algorithms and feature selection to a public domain dataset of mouse cortical protein expression.
• Utilized supervised learning technologies to predict learning status based on protein levels.
• Developed a novel redundancy-aware feature selection method.

Main Results:

• Six cortical molecular biomarkers were identified as predictive of learning: brain-derived neurotrophic factor (BDNF), NR2A, B-cell lymphoma 2 (BCL2), histone H3 acetylation at lysine 18 (H3AcK18), protein kinase R-like endoplasmic reticulum kinase (pERK), and superoxide dismutase 1 (SOD1).
• Five of these biomarkers (BDNF, NR2A, H3AcK18, pERK, SOD1) have prior associations with learning in scientific literature.
• BDNF, NR2A, and BCL2 were previously linked to pruning, and BCL2 to apoptosis, suggesting a connection between learning and these cellular processes.

Conclusions:

• The identified panel of six protein biomarkers (BDNF, NR2A, BCL2, H3AcK18, pERK, SOD1) can accurately predict learning in mice.
• These findings highlight the potential role of cortical pruning and apoptosis in the mechanisms of learning.
• The study demonstrates the power of AI in discovering novel biomarkers for cognitive functions.