Discovering root causal genes with high-throughput perturbations.

Area of Science:

• Genomics and Bioinformatics
• Systems Biology
• Computational Biology

Background:

• Root causal genes initiate disease-associated gene expression changes, making their identification crucial for early therapeutic intervention.
• Existing algorithms struggle to accurately identify root causal genes from RNA-sequencing (RNA-seq) data due to challenges like noise, high dimensionality, and non-linearity.
• Perturb-seq offers a high-throughput method combining perturbations with single-cell RNA-seq for learning gene causal order.

Purpose of the Study:

• To develop a novel computational approach for identifying root causal genes from biological data.
• To overcome the limitations of current methods in analyzing complex RNA-seq data for causal gene discovery.
• To enable the identification of patient-specific root causal genes for personalized medicine.

Main Methods:

• Utilized Perturb-seq data to establish the causal relationships and order between genes.
• Transferred the learned causal order from Perturb-seq to bulk RNA-seq data.
• Developed and applied a novel statistical method to identify patient-specific root causal genes.

Main Results:

• Demonstrated significant performance improvements compared to existing state-of-the-art approaches.
• Successfully identified root causal genes in applications to macular degeneration and multiple sclerosis.
• Revealed root causal genes on known pathogenic pathways, aiding in patient subgroup delineation and suggesting an omnigenic root causal model.

Conclusions:

• The developed method effectively identifies root causal genes, offering a new tool for understanding disease initiation.
• This approach facilitates the discovery of patient-specific causal genes, paving the way for targeted therapies.
• Findings highlight the potential of integrating Perturb-seq and RNA-seq for advancing causal inference in complex diseases.