WormTensor

WormTensor clusters time-series whole-brain neural activity data from Caenorhabditis elegans to identify functional modules and common circuits across multiple animals.

Key Features:

• Time-series clustering methodology: Processes dynamic whole-brain neural activity to capture temporal patterns in C. elegans neural data.
• Handling missing data: Integrates data from multiple animals to mitigate the impact of missing observations on cluster inference.
• Modified shape-based distance measure: Employs a shape-based distance that accounts for temporal lags and mutual inhibition between cells.
• Tensor decomposition algorithm (MC-MI-HOOI): Uses matrix integration via higher-order orthogonal iteration (MC-MI-HOOI) for multi-view clustering.
• Robustness to noisy data: Shows robustness to data contamination in simulation studies.
• Performance metrics: Yields higher silhouette coefficients compared with widely used consensus clustering methods, indicating improved cluster separation.

Scientific Applications:

• Functional module identification: Identification of functional neural modules and circuits from whole-brain time-series activity in C. elegans.
• Cross-specimen integration: Integration of recordings across multiple specimens for comparative and large-scale studies of neural dynamics.
• Analysis of temporal and inhibitory interactions: Investigation of temporal lags and mutual inhibition in cell-cell interactions within neural networks.

Methodology:

Combine whole-brain activity data from multiple C. elegans individuals; compute a modified shape-based distance incorporating temporal lags and inhibitory interactions; apply the MC-MI-HOOI tensor decomposition algorithm for multi-view clustering to estimate reliability weights and common clusters.