Visualization of multi-neuron activity by simultaneous optimization of clustering and dimension reduction

Area of Science:

• Neuroscience
• Computational Biology
• Data Science

Background:

• Simultaneous recordings of over 100 neurons are now possible using microelectrode arrays.
• Visualizing and interpreting activity patterns across numerous neurons remains a significant challenge.
• Understanding the relationship between neural activity patterns and cognitive functions, like perception, requires advanced analytical tools.

Purpose of the Study:

• To develop a novel computational method for simultaneously clustering and reducing the dimensionality of high-dimensional neural activity data.
• To provide a more intuitive way to visualize and understand complex neural firing patterns.
• To assess the effectiveness of the proposed algorithm in analyzing both simulated and real neural data.

Main Methods:

• A variational Bayes algorithm was developed and applied to perform simultaneous clustering and dimension reduction.
• The algorithm's performance was evaluated using both artificial (simulated) neural data and real experimental neural recordings.
• Comparative analysis was conducted against principal component analysis (PCA) in a reduced subspace.

Main Results:

• The variational Bayes algorithm successfully performed simultaneous clustering and dimension reduction on neural activity data.
• The algorithm demonstrated superior clustering performance compared to PCA when applied to a reduced data subspace.
• Visualizations derived from the algorithm offered greater intuition into neural activity patterns.

Conclusions:

• The developed variational Bayes algorithm offers an effective approach for analyzing large-scale neural recordings.
• This method enhances the ability to visualize and interpret complex neural activity patterns, aiding in the understanding of neural coding.
• The algorithm provides a valuable tool for neuroscience research, potentially facilitating the link between neural activity and cognitive functions.