Virtual Knockout Tool – a computational method that simulates gene deletions using single‑cell RNA‑seq data. By constructing a gene regulatory network from wild‑type samples and mathematically removing the target gene, it predicts functional consequences without physical experiments.
A curated collection of computational tools for in-silico gene knockout experiments using single-cell RNA sequencing data.
Virtual knockout (KO) refers to computational methods that simulate the functional deletion of a gene without physical CRISPR or RNAi experiments. These tools predict downstream transcriptional changes and regulatory effects by computationally perturbing gene regulatory networks (GRNs) constructed from wild-type (WT) scRNA-seq data alone.
- Cost-effective: No cell culture, no animal models, no gene editing reagents needed
- Scalable: Systematically knockout hundreds or thousands of genes simultaneously
- Cell-type specific: Leverages single-cell resolution for precise predictions
- Accelerates research: Predicts outcomes before committing to expensive animal KO experiments
Most virtual knockout tools follow a similar pipeline:
- Construct GRN: Build a gene regulatory network from WT scRNA-seq data
- Simulate KO: Remove all outgoing edges of the target gene (or use other perturbation strategies)
- Compare networks: Align the perturbed network to the original to identify differentially regulated genes
- Interpret results: Enrichment analysis reveals affected pathways and biological processes
A machine learning workflow that performs virtual knockout experiments by constructing gene regulatory networks from scRNA-seq data, then removing outgoing edges of the target gene. Manifold alignment compares original and perturbed networks to identify differentially regulated genes.
-
Paper: Osorio, D. et al. Patterns, 2022, 3(3), 100434. DOI: [10.1016/j.patter.2022.100434
](https://doi.org/10.1016/j.patter.2022.100434 ) -
Preprint: bioRxiv, DOI: [10.1101/2021.03.22.436484
](https://doi.org/10.1101/2021.03.22.436484 ) -
GitHub: cailab-tamu/scTenifoldKnk (R/MATLAB/Python)
-
CRAN: scTenifoldKnk
-
Documentation: ReadTheDocs
-
Languages: R, MATLAB, Python
Uses a variational graph autoencoder (VGAE) to learn latent representations of genes from WT scRNA-seq data. Virtual KO is simulated by removing all edges of the target gene from the GRN, and differences are discerned through latent parameters of the trained VGAE.
-
Paper: Yang, Y. et al. Nucleic Acids Research, 2023, 51(13), 6578-6592. DOI: [10.1093/nar/gkad450
](https://doi.org/10.1093/nar/gkad450 ) -
GitHub: yjgeno/GenKI
-
Language: R/Python
Adapts a perturbation-response framework originally developed for protein structural dynamics to identify gene modules most susceptible to perturbation. Introduces scPII, a data-driven metric derived from GRNs that quantifies system-level responses to gene perturbations.
-
Paper: Gupta, S., Romero, S., Cai, J.J. bioRxiv, 2025. DOI: [10.64898/2025.12.15.694358
](https://doi.org/10.64898/2025.12.15.694358 ) -
Status: Preprint (December 2025)
-
Code: Not yet released as dedicated package (from Cai Lab)
Predicts how transcriptional factor (TF) knockouts alter cell differentiation trajectories and cell fate decisions. Constructs base GRNs using scATAC-seq data, pre-built promoter networks, or user-defined TF-target lists. Uniquely capable of simulating cell fate transitions.
-
Paper: Kamimoto, K. et al. Cell Systems, 2020, 11(4), 343-354. DOI: [10.1016/j.cels.2020.08.013
](https://doi.org/10.1016/j.cels.2020.08.013 ) -
GitHub: morris-lab/Celloracle
-
Language: Python
Note: Celloracle is specifically designed for TF knockout analysis and requires scATAC-seq or pre-defined GRN data. Not suitable for all gene types.
Model-based integrated pipeline for single-cell CRISPR screening data analysis, capable of prioritizing gene perturbation effects at the cellular heterogeneity level.
-
Paper: Duan, B. et al. Nature Communications, 2019, 10, 2233. DOI: [10.1038/s41467-019-10284-3
](https://doi.org/10.1038/s41467-019-10284-3 ) -
Code: R package (See paper for access details)
-
Language: R
Note: MUSIC requires real KO sample data from single-cell CRISPR screens, making it distinct from purely WT-based tools.
A large-scale GRN simulation and benchmarking tool that can generate realistic single-cell gene expression data with perturbations (including knockdowns). While primarily a simulation/benchmarking tool, it is the only tool that allows simulations of knockdown perturbations using a GRN as the basis for data generation.
-
Paper: Garbulowski, M. et al. Bioinformatics Advances, 2024, 6(3), lqae121. DOI: [10.1093/bioadv/lqae121
](https://doi.org/10.1093/bioadv/lqae121 ) -
GitHub: Not specified (MATLAB toolbox)
A platform for simulating cell state transitions using stochastic differential equations governed by a regulatory network of core TFs. Can be used to predict effects of gene perturbations on developmental trajectories.
-
Paper: Alanis-Lobato, G. et al. Bioinformatics, 2024. DOI: [10.1093/bioinformatics/btae654
](https://doi.org/10.1093/bioinformatics/btae654 ) -
Code: Available through Oxford Bioinformatics publication
| Tool | Description | Link |
|---|---|---|
| scTenifoldNet | Base network construction workflow (underpins scTenifoldKnk) | GitHub |
| PertFlow | Cloud-based workflow for perturbational modeling on scRNA-seq | Zenodo |
| IQCELL | Predicts effects of gene perturbations on developmental trajectories | Publication |
| Celcomen | Spatial causal disentanglement for perturbation modeling | GitHub |
-
Overview of Virtual Knockout Methods (Bilibili article, 2026): "不用养细胞、不用CRISPR！单细胞虚拟敲除，零实验搞定基因功能预测" — A comprehensive Chinese review covering scTenifoldKnk, Celloracle, and related tools Link
-
Robustness and applicability of functional genomics tools on scRNA-seq data (Mendeley): Discusses in-silico perturbation simulation strategies Link
If you use these tools in your research, please cite the original papers for the respective tools:
@article{osorio2022scTenifoldKnk,
title={scTenifoldKnk: An efficient virtual knockout tool for gene function predictions via single-cell gene regulatory network perturbation},
author={Osorio, D. and Zhong, Y. and Li, G. et al.},
journal={Patterns},
volume={3},
number={3},
pages={100434},
year={2022},
doi={10.1016/j.patter.2022.100434}
}
@article{yang2023genki,
title={Gene knockout inference with variational graph autoencoder learning single-cell gene regulatory networks},
author={Yang, Y. and Li, G. and Zhong, Y. et al.},
journal={Nucleic Acids Research},
volume={51},
number={13},
pages={6578-6592},
year={2023},
doi={10.1093/nar/gkad450}
}
@article{kamimoto2020celloracle,
title={Celloracle: Dissecting cell fate decisions through gene regulatory network inference},
author={Kamimoto, K. and Hoffmann, C. M. and Morris, S. A.},
journal={Cell Systems},
volume={11},
number={4},
pages={343-354},
year={2020},
doi={10.1016/j.cels.2020.08.013}
}