Charting the computational landscape of single-cell genetic perturbation

Area of Science:

• Computational Biology
• Genomics
• Systems Biology

Background:

• Single-cell genetic perturbation screens are vital for gene function studies at high resolution.
• Analysis challenges arise from diverse computational methods with differing assumptions and requirements.
• This heterogeneity complicates cross-study comparisons and consistent tool selection.

Purpose of the Study:

• To organize single-cell genetic perturbation analysis into a standardized five-step computational workflow.
• To compare current algorithms based on modeling principles, assumptions, limitations, and use cases.
• To guide method selection and improve reporting and interpretation in perturbation studies.

Main Methods:

• Categorization of algorithms by modeling principles (regression, latent-state, matrix-based, generative models).
• Comparison of methods regarding assumptions, limitations, and applicability.
• Review of multimodal data integration strategies and perturbation prediction models.

Main Results:

• A structured five-step workflow (preprocessing, refinement, QC, effect quantification, annotation) for perturbation analysis.
• Classification of computational tools based on their underlying modeling approaches.
• Summary of multimodal analysis techniques and advances in perturbation prediction models.

Conclusions:

• The proposed workflow provides a framework for understanding and selecting computational tools for single-cell genetic perturbation studies.
• Linking workflow steps with model assumptions facilitates method reporting and interpretation.
• This structured approach aims to enhance reproducibility and consistency in the field.