LPWC

LPWC clusters short biological time series using lag-penalized weighted correlation to detect groups of genes, phosphosites, or proteins that exhibit similar but potentially time-shifted and irregularly sampled temporal patterns.

Key Features:

• Temporal alignment and lag incorporation: Aligns time-series profiles by introducing lags to accommodate delayed responses between biological entities.
• Weighted correlation: Computes a weighted correlation similarity that down-weights pairwise similarity according to the length of introduced lags.
• Handling irregular time intervals: Supports datasets with irregular sampling intervals between time points.
• Clustering performance: Recovers true clusters in simulations based on a biologically-motivated impulse model and outperforms existing time-series clustering algorithms in case studies.
• Interpretability: Produces clusters tailored to the temporal structure of high-throughput biological data, facilitating biological interpretation.

Scientific Applications:

• Gene expression analysis: Groups genes with similar temporal expression trajectories, including time-shifted responses.
• Protein phosphorylation studies: Clusters phosphosites to reveal coordinated phosphorylation dynamics over time.
• Yeast osmotic stress response: Identifies distinct temporal patterns in yeast osmotic stress response datasets.
• Axolotl limb regeneration: Identifies distinct temporal patterns in axolotl limb regeneration time-course data.

Methodology:

Temporal alignment with explicit lag incorporation; computation of a lag-penalized weighted correlation that down-weights similarity by lag length; support for irregular time intervals; validation via simulations using a biologically-motivated impulse model and application to yeast osmotic stress and axolotl limb regeneration case studies.