A framework to identify structured behavioral patterns within rodent spatial trajectories.

Area of Science:

• Neuroscience
• Computational Biology
• Animal Behavior

Background:

• Animal behavior is complex and structured, but traditional data analysis methods struggle to reveal underlying patterns.
• Existing methods like principal component analysis (PCA) have limitations in capturing the full dynamics of movement.

Purpose of the Study:

• To introduce a novel computational framework for quantitatively characterizing rodent behavior during spatial navigation.
• To test the hypothesis that animal movements are composed of a limited set of motor primitives combined sparsely over time.

Main Methods:

• Formalized the hypothesis using sparse dictionary learning to extract motor primitives from rodent position and velocity data.
• Applied the method to reconstruct past trajectories and predict novel ones, validating against PCA and single sparsity approaches.

Main Results:

• Behavioral trajectories were robustly reconstructed from incomplete data, outperforming standard dimensionality reduction techniques.
• Extracted motor primitives generalized across experimental sessions and maze variations, enabling accurate behavioral reconstruction.
• The number of identified motor primitives correlated with maze complexity, indicating sensitivity to task structure.

Conclusions:

• The proposed framework offers a powerful tool for analyzing complex animal behavior and neural data.
• Motor primitives provide a quantitative basis for characterizing behavioral complexity, habit formation, and predicting neural activity.