Fluctuation structure predicts genome-wide perturbation outcomes.

Area of Science:

• Functional genomics
• Systems biology
• Statistical physics

Background:

• Interpreting pooled single-cell perturbation screens is challenging.
• Current methods are either opaque deep learning models or oversimplified frameworks.
• Gene co-fluctuations in unperturbed cells can inform perturbation response modeling.

Purpose of the Study:

• To present CIPHER (Covariance Inference for Perturbation and High-dimensional Expression Response), a novel framework for predicting transcriptome-wide perturbation outcomes.
• To leverage linear response theory and gene co-fluctuations for enhanced biological interpretation of perturbation screens.

Main Methods:

• Developed CIPHER, a framework using linear response theory and gene co-fluctuations.
• Validated on synthetic networks and 11 large-scale single-cell perturbation datasets (4,234 perturbations, >1.36M cells).
• Employed Bayesian inference for uncertainty-aware effect size estimation.

Main Results:

• CIPHER accurately recapitulated genome-wide responses to single and double perturbations by utilizing baseline gene covariance.
• Removing gene-gene covariances reduced model performance 11-fold, highlighting the importance of fluctuation structures.
• Gene-gene correlations were transferable across independent studies, indicating conserved fluctuation patterns.
• CIPHER outperformed differential expression metrics in identifying perturbations and provided uncertainty-aware estimates.
• Genome-wide responses propagated through the covariance matrix along ~3 global gene modules.

Conclusions:

• CIPHER demonstrates the power of theoretically-grounded models in understanding complex biological responses.
• Cellular fluctuation patterns encode fundamental design principles crucial for predicting perturbation outcomes.
• Harnessing gene co-fluctuations offers a more robust approach to functional genomics analysis.