Shape-constrained, changepoint additive models for time series omics data with cpam.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Time series omics experiments are vital for understanding dynamic biological processes like cell differentiation and responses to stimuli.
• Existing statistical methods for analyzing static omics data are insufficient for complex temporal datasets.
• A comprehensive methodology for time series omics data analysis is needed.

Purpose of the Study:

• Introduce cpam, a novel R package for temporal differential analysis of omics time series data.
• Provide a user-friendly tool with features for change-point detection and temporal trend estimation.
• Enable robust analysis of case-only and case-control designs, incorporating quantification uncertainty.

Main Methods:

• Developed cpam, an R package implementing change-point detection and shape-constrained temporal trend estimation.
• Integrated an interactive interface with customizable visualizations for biological insight.
• Evaluated performance against existing time series methods using metrics like false discovery rate and power.

Main Results:

• cpam demonstrates superior performance in controlling the false discovery rate and enhancing the power to detect temporal changes.
• The method accurately estimates changepoints in omics time series data.
• Applied cpam to human embryogenesis data for RNA isoform-level modeling and identified 910 novel light-responsive genes in Arabidopsis.

Conclusions:

• cpam provides a powerful and accurate solution for differential analysis of omics time series data.
• The package offers significant advantages over existing methods for temporal omics data interpretation.
• cpam facilitates high-resolution clustering and identification of dynamic molecular responses in biological systems.