Decoding Heterogeneous Single-Cell Perturbation Responses

Bioengineering Cell Biology Biochemistry Genetics Life Sciences
single-cell perturbation gene expression cellular response transcriptomics CRISPR

Background Introduction

In cell biology, understanding how cells respond differently to perturbations is crucial. Perturbation refers to altering cell states through gene editing, chemical substances, environmental changes, or mechanical forces to study their functions. However, existing methods have limitations in quantifying heterogeneous responses at the single-cell level, particularly in analyzing partial gene perturbations and dosage effects. To address this issue, researchers developed a new computational method—the Perturbation-Response Score (PS)—aimed at more accurately quantifying the heterogeneity of single-cell perturbation responses and revealing the influence of intrinsic and extrinsic factors on perturbation outcomes.

Source of the Paper

This paper was co-authored by Bicna Song, Dingyu Liu, and other researchers from multiple renowned institutions, including Children’s National Hospital, Sloan Kettering Institute, and Johns Hopkins University. The paper was published in March 2025 in the journal Nature Cell Biology, titled Decoding Heterogeneous Single-Cell Perturbation Responses.

Research Process and Results

1. Construction of the PS Framework

The core of the PS framework quantifies single-cell perturbation responses through a three-step process:
- Step 1: Target Gene Identification
By comparing transcriptomic data between perturbed and unperturbed cells, differentially expressed genes (DEGs) are identified. These genes serve as “signature genes” for the perturbation and are used in subsequent analyses.
- Step 2: Average Perturbation Effect Estimation
The scMAGeCK model is used to calculate the average perturbation effect of signature genes, generating a β score, similar to log fold change.
- Step 3: Constrained Optimization for PS Estimation
A constrained optimization algorithm is used to compute the perturbation-response score (PS) for each cell, ranging from 0 (no perturbation effect) to 1 (maximum perturbation effect).

2. Benchmarking and Validation

The researchers validated the performance of PS across multiple datasets:
- Synthetic Datasets: Using scDesign3, single-cell transcriptomic data under different perturbation efficiencies (25%-100%) were simulated. PS significantly outperformed the existing method Mixscape in quantifying partial perturbations.
- CRISPRi Perturbation Datasets: In K562 cell CROP-seq data, PS accurately estimated CRISPR interference (CRISPRi) efficiency, while Mixscape performed poorly.
- Genome-Scale Perturbation Datasets: In Jurkat T cell perturbation experiments, PS successfully identified known regulators of T cell activation and demonstrated higher accuracy in ROC curve analysis.

3. Dosage Effect Analysis

PS can analyze the dosage effects of gene perturbations without the need for titration. For example, in perturbation experiments on PD-L1 regulatory genes, PS revealed the dosage response patterns of positive regulators (e.g., STAT1) and negative regulators (e.g., CUL3). Additionally, PS identified “buffered genes” and “sensitive genes,” where the former require high-efficiency perturbations to trigger strong transcriptional responses, while the latter show significant effects even under moderate perturbations.

4. Application Cases

  • HIV Latency Expression: In Jurkat cell perturbation experiments, PS revealed the cell-state-dependent functions of BRD4 and CCNT1 in HIV latency expression.
  • Pancreatic Differentiation: In human embryonic stem cell (hESC) pancreatic differentiation experiments, PS uncovered a novel role for CCDC6 in regulating liver and pancreatic cell fate decisions.

Research Conclusions and Significance

The PS framework provides a powerful tool for analyzing single-cell perturbation data, enabling the quantification of partial perturbations, revealing dosage effects, and identifying biological factors influencing perturbation responses. The innovation of this method lies in:
1. Quantifying Partial Perturbations: Accurately analyzing effects of incomplete gene knockouts or RNA interference.
2. Dosage Effect Analysis: Inferring dose-response relationships from single-cell data without experimental titration.
3. Cell-State Dependency: Revealing cell-state-specific perturbation responses, offering new perspectives for functional genomics research.

Research Highlights

  1. High Precision Quantification: PS demonstrated outstanding performance in multiple benchmarks, particularly in partial perturbation and dosage effect analysis.
  2. Broad Applicability: Suitable for various single-cell perturbation data, including CRISPR screens and drug treatments.
  3. Biological Discoveries: Uncovered the novel role of CCDC6 in pancreatic differentiation, providing important insights for developmental biology research.

Other Valuable Information

The open-source code for the PS framework is publicly available on GitHub, allowing researchers to apply it to their own data analysis. Additionally, the paper provides detailed experimental methods and data analysis workflows, serving as a reference for future research.

This study not only developed a powerful computational tool but also provided new perspectives for understanding the heterogeneous responses of cells to perturbations, holding significant scientific and application value.