Uncovering causal relationships in single-cell omic studies with causarray.

Area of Science:

• Genomics
• Computational Biology
• Causal Inference

Background:

• Single-cell sequencing and CRISPR technologies offer high resolution for biological studies.
• Analyzing observational genomic data for causal relationships is difficult due to bias and unmeasured confounders.

Purpose of the Study:

• To introduce causarray, a robust causal inference framework for genomic data analysis.
• To address challenges in identifying causal effects in heterogeneous and confounded datasets at both pseudo-bulk and single-cell levels.

Main Methods:

• Developed causarray, integrating a generalized confounder adjustment method.
• Employed semiparametric inference and flexible machine learning for robust statistical estimation.
• Applied the framework to array-based genomic data, including single-cell Perturb-seq and human brain transcriptomic datasets.

Main Results:

• Causarray effectively separates treatment effects from confounders while preserving biological signals.
• Identified clustered causal effects of autism risk genes in mouse brain development.
• Discovered consistently causally affected genes across Alzheimer's disease datasets, revealing relevant pathways.

Conclusions:

• Causarray provides a robust method for causal inference in complex genomic data under unmeasured confounding.
• The framework enhances understanding of disease mechanisms by identifying key genes and pathways in neurodevelopmental and neurodegenerative disorders.