Large-scale causal discovery using interventional data sheds light on gene network structure in k562 cells

Area of Science:

• Systems biology
• Genomics
• Network inference

Background:

• Inferring directed biological networks is challenging.
• Large-scale CRISPR perturbation data offers new opportunities for network inference.
• Understanding gene regulatory networks is crucial for elucidating complex traits.

Purpose of the Study:

• To introduce inverse sparse regression (inspre), a novel method for learning causal networks from intervention-response data.
• To apply inspre to genome-wide CRISPR screens to infer a directed biological network.
• To explore the properties of the inferred network and its relationship with gene function and heritability.

Main Methods:

• Developed inverse sparse regression (inspre) for causal network inference.
• Applied inspre to the genome-wide perturb-seq dataset (788 genes).
• Integrated inferred network with external data on gene essentiality and heritability.

Main Results:

• inspre successfully inferred a directed biological network from CRISPR perturbation data.
• The discovered network exhibits small-world and scale-free properties.
• Gene eigencentrality in the inferred network correlates with gene essentiality and expression heritability.

Conclusions:

• inspre is an effective approach for inferring causal biological networks using large-scale intervention-response data.
• The inferred network structure provides insights into gene function and regulation.
• This work elucidates network properties underlying complex traits and offers a framework for future studies.