RIDDEN: Data-driven inference of receptor activity from transcriptomic data.

Area of Science:

• Computational Biology
• Systems Biology
• Genomics

Background:

• Receptor signaling is crucial for physiological regulation and disease, making receptors key drug targets.
• Existing computational methods for ligand-receptor interactions often focus on ligands or gene co-expression, which may not reflect functional activity.
• There is a need for tools that directly infer receptor activity from downstream gene expression changes.

Purpose of the Study:

• To develop a computational tool, RIDDEN (Receptor actIvity Data Driven inferENce), for predicting receptor activity.
• To infer receptor activity directly from receptor-regulated gene expression profiles.
• To enable systems-level analysis of cell and disease-specific receptor activity alterations.

Main Methods:

• Trained the RIDDEN model using 14,463 perturbation gene expression profiles across 229 receptors.
• RIDDEN infers receptor activity from downstream gene expression, not ligand or receptor gene expression.
• Validated the model on independent in vitro and in vivo receptor perturbation datasets.

Main Results:

• RIDDEN effectively predicts receptor activity in bulk and single-cell transcriptomics data.
• Model weights align with known receptor-transcription factor regulatory interactions.
• Predicted receptor activities correlate with receptor and ligand expression in in vivo data.
• RIDDEN identified mechanistic biomarkers in a cancer patient cohort treated with immune checkpoint blockade.

Conclusions:

• RIDDEN is the largest transcriptomics-based receptor activity inference model to date.
• The tool can identify cell populations with altered receptor activity.
• RIDDEN facilitates the study of cell-cell communication using transcriptomics data.
• This approach advances understanding of receptor function in physiological and disease states.