Cancer Analysis Agent
Omi's specialised co-pilot for cancer analysis work
I'm the agent for translational cancer single-cell work. Survival analysis, prognostic signatures, tumor vs normal comparisons, malignant cell identification, TME deconvolution, and bridging your scRNA findings to TCGA, ICGC, and bulk cohorts — so your single-cell story actually lands in a clinical journal.
What I can do for you
I run survival analyses — Kaplan-Meier curves, log-rank tests, univariate and multivariate Cox regression — on signatures derived from your single-cell data, applied to TCGA, ICGC, METABRIC, or your own bulk cohorts.
I identify prognostic gene signatures from your scRNA clusters (e.g. an exhausted-CD8 signature, a CAF-S1 signature) and test whether they stratify patient survival in independent datasets — that's how a cell type becomes a biomarker.
I separate malignant from normal cells with inferCNV/copyKAT, score tumor heterogeneity, identify subclones, and compare tumor cells to matched normal — finding the genes that are tumor-specific rather than just tissue-of-origin markers.
I deconvolute the tumor microenvironment, compute immune infiltration scores, link them to outcome, and compare your single-cell-derived insights against TCGA pan-cancer patterns — bringing the bulk world and the single-cell world into one analysis.
Examples of what you can ask me
Copy any of these straight into the demo, or adapt them to your data.
- 1"Test whether my exhausted-CD8 signature predicts survival in TCGA melanoma."
- 2"Run Kaplan-Meier on my prognostic signature, stratified by stage."
- 3"Cox regression: is my signature independent of age, stage, and grade?"
- 4"Compare tumor vs adjacent normal cells and find tumor-specific markers."
- 5"Identify malignant cells with inferCNV and quantify subclonal heterogeneity."
- 6"Score immune infiltration in TCGA using my single-cell-derived gene panels."
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
Cancer biologists, clinician-scientists, translational oncology labs, biomarker discovery teams, and anyone who wants their single-cell paper to include the survival curve that reviewers always ask for.
References
- Kaplan-Meier estimator (Kaplan & Meier, 1958) – JASA
- Cox proportional hazards (Cox, 1972) – JRSS-B
- lifelines (Davidson-Pilon, 2019) – JOSS
- TCGA (The Cancer Genome Atlas Research Network, 2013) – Nature Genetics
- inferCNV (Tickle et al., Trinity CTAT)
- copyKAT (Gao et al., 2021) – Nature Biotechnology
- CIBERSORTx (Newman et al., 2019) – Nature Biotechnology
Try Cancer Analysis now
Jump into the demo with a starter prompt already loaded. Upload your data, or play with our example dataset first.