decoupleR: ensemble of computational methods to infer biological activities from omics data

Area of Science:

• Computational biology
• Bioinformatics
• Systems biology

Background:

• Omics data analysis often requires dimensionality reduction for statistical power and interpretability.
• Prior knowledge resources are crucial for extracting meaningful biological insights from complex omics datasets.
• Existing frameworks may lack flexibility in incorporating diverse prior knowledge and regulatory information.

Purpose of the Study:

• To present decoupleR, a unified computational framework for extracting biological activities from omics data.
• To offer a flexible package supporting various methods, including those leveraging mode of regulation and interaction weights.
• To integrate with OmniPath, a comprehensive meta-resource of biological prior knowledge.

Main Methods:

• decoupleR package implementation in Bioconductor (R) and Python.
• Utilizing OmniPath, a meta-resource of over 100 prior knowledge databases.
• Evaluation of different activity extraction methods on transcriptomic and phospho-proteomic perturbation datasets.

Main Results:

• decoupleR provides a unified framework for diverse biological activity extraction methods.
• The package flexibly incorporates prior knowledge resources, including interaction weights and mode of regulation.
• Simple linear models and consensus scores across methods demonstrated superior performance in predicting perturbed regulators.

Conclusions:

• decoupleR offers a versatile and powerful tool for biological activity extraction from omics data.
• The framework's flexibility and integration with OmniPath enhance the interpretability of omics experiments.
• The study identifies optimal methods for predicting perturbed regulators in perturbation studies.