Perturbation Simulation Agent
Omi's specialised co-pilot for perturbation simulation work
I let you ask 'what if' questions of your single-cell data. What happens if you knock out FOXP3? What's the predicted response to a JAK inhibitor? I run real in-silico perturbation models so you can prioritise the experiment that's actually worth doing at the bench.
What I can do for you
I run scGen, CPA, GEARS, or scFoundation-based perturbation models on your dataset to predict expression changes after gene knockout, knock-in, or combinatorial perturbations — including ones never measured in your data.
I match your transcriptomic state against drug-response databases (LINCS L1000, CMap, DrugComb) and rank candidate compounds that would push your disease cells toward a healthy state.
I analyse Perturb-seq, CROP-seq, or ECCITE-seq experiments end-to-end: guide assignment, knockdown QC, perturbation-vs-control DE, and gene-program inference so you can see which sgRNAs hit which pathway.
I explain the predictions in plain biology — which TFs are driving the predicted shift, which programs are turning on or off, and which predictions you should trust vs treat as exploratory.
Examples of what you can ask me
Copy any of these straight into the demo, or adapt them to your data.
- 1"Predict what happens to my CD8 T cells if I knock out TOX."
- 2"Which FDA-approved drugs would push my fibroblasts back to a quiescent state?"
- 3"Analyse my Perturb-seq screen and rank guides by transcriptomic effect size."
- 4"Simulate combinatorial knockout of STAT1 + STAT3 in my macrophages."
- 5"Which perturbations make my tumor cells look more like normal epithelium?"
- 6"Score my dataset against the LINCS drug response signatures."
How I work
I run real Scanpy (Python) or Seurat (R) code on the secure MCP server — no hallucinations, no made-up gene lists. Every result comes with the exact code I executed and the parameters I used, so your analysis is fully reproducible and ready for the Methods section.
Best for
Drug discovery teams, CRISPR screen designers, translational labs trying to prioritise next experiments, and computational biologists who want to validate models before wet-lab spend.
References
- scGen (Lotfollahi et al., 2019) – Nature Methods
- CPA (Lotfollahi et al., 2023) – Molecular Systems Biology
- GEARS (Roohani et al., 2024) – Nature Biotechnology
- scFoundation (Hao et al., 2024) – Nature Methods
- LINCS L1000 / CMap (Subramanian et al., 2017) – Cell
- Perturb-seq (Dixit et al., 2016) – Cell
Try Perturbation Simulation now
Jump into the demo with a starter prompt already loaded. Upload your data, or play with our example dataset first.