Skip to content

cBioPortal Study Analytics in BioMCP

Download real cancer genomics studies. Run survival analysis, co-occurrence, and cohort comparisons. Generate charts. All from the command line.

cBioPortal study analytics overview

BioMCP can download cBioPortal study datasets and run analytics locally — no browser, no R, no Jupyter notebooks. You get structured data and publication-quality charts from the same commands.

This post walks through what the study analytics surface does, how it works under the hood, and how Kuva powers the charting.

What you can do

BioMCP's study commands cover five analysis types:

Command What it does
study query Mutation frequency, CNA distribution, expression values for a gene
study survival Kaplan-Meier survival analysis stratified by mutation status
study co-occurrence Pairwise mutation co-occurrence with log odds ratios and p-values
study compare Expression or mutation rate comparison between mutation-stratified groups
study cohort Sample-level cohort construction by mutation status

Every command works on locally downloaded cBioPortal datasets. Download once, query as many times as you want:

biomcp study download msk_impact_2017
biomcp study download brca_tcga_pan_can_atlas_2018

Survival analysis

The study survival command stratifies patients by mutation status and computes Kaplan-Meier estimates with log-rank testing.

biomcp study survival --study msk_impact_2017 --gene TP53

# Study Survival: TP53 (msk_impact_2017)

Endpoint: Overall Survival (OS)

| Group | N | Events | Censored | Event Rate | KM Median | 1yr | 3yr | 5yr |
|---|---|---|---|---|---|---|---|---|
| TP53-mutant | 3323 | 1167 | 2156 | 0.351189 | 21.0 | 0.694 | 0.297 | 0.297 |
| TP53-wildtype | 4251 | 1058 | 3193 | 0.248883 | 32.1 | 0.806 | 0.461 | 0.381 |

Log-rank p-value: 9.40e-29

TP53-mutant patients have a median overall survival of 21.0 months versus 32.1 months for wildtype — a difference that's statistically overwhelming (p = 9.40e-29). This is consistent with TP53's well-established role as a prognostic marker across cancer types.

Add --chart survival to get a Kaplan-Meier curve:

biomcp study survival --study msk_impact_2017 --gene TP53 \
  --chart survival -o tp53-survival.svg

TP53 Kaplan-Meier survival curve

The survival curve shows clear separation between the two groups from the first months of follow-up, with the gap widening through the observation period.

Co-occurrence analysis

The study co-occurrence command computes pairwise mutation co-occurrence across a set of genes. It reports contingency counts, log odds ratios, and Fisher's exact test p-values.

biomcp study co-occurrence --study msk_impact_2017 \
  --genes TP53,KRAS,PIK3CA,EGFR

# Study Co-occurrence: msk_impact_2017

Genes: TP53, KRAS, PIK3CA, EGFR

Total samples: 10945

| Gene A | Gene B | Both | A only | B only | Neither | Log Odds Ratio | p-value |
|---|---|---|---|---|---|---|---|
| TP53 | KRAS | 845 | 3693 | 798 | 5609 | 0.475 | 1.091e-18 |
| TP53 | PIK3CA | 578 | 3960 | 777 | 5630 | 0.056 | 3.460e-1 |
| TP53 | EGFR | 357 | 4181 | 279 | 6128 | 0.629 | 2.153e-14 |
| KRAS | PIK3CA | 222 | 1421 | 1133 | 8169 | 0.119 | 1.328e-1 |
| KRAS | EGFR | 37 | 1606 | 599 | 8703 | -1.094 | 1.929e-13 |
| PIK3CA | EGFR | 93 | 1262 | 543 | 9047 | 0.205 | 8.221e-2 |

The standout finding: KRAS and EGFR are mutually exclusive (log odds ratio -1.094, p = 1.93e-13). Only 37 of 10,945 samples carry both mutations. This is a well-known biological pattern — both genes activate the same RAS-MAPK signaling pathway, so mutating both provides no selective advantage.

Add --chart bar for a visual:

biomcp study co-occurrence --study msk_impact_2017 \
  --genes TP53,KRAS,PIK3CA,EGFR --chart bar -o cooccurrence.svg

Co-occurrence bar chart

How the study analytics work

BioMCP downloads cBioPortal study data as tab-separated files and stores them locally. When you run a query, BioMCP reads these files directly — no API calls, no network latency, no rate limits.

The survival analysis implementation:

  1. Mutation stratification — Samples are split into mutant and wildtype groups based on the mutation data file.
  2. Clinical data join — Overall survival status and months are pulled from the clinical data file and matched to each sample.
  3. Kaplan-Meier estimation — Survival probabilities are computed at each event time using the standard product-limit estimator.
  4. Log-rank test — A chi-squared statistic compares the two survival curves. The p-value comes from a chi-squared distribution with one degree of freedom.

The co-occurrence analysis:

  1. Mutation extraction — For each gene, BioMCP identifies which samples carry at least one mutation.
  2. Contingency tables — For every pair of genes, samples are counted in four buckets: both mutated, A only, B only, neither.
  3. Log odds ratio — Computed from the 2×2 contingency table. Positive values indicate co-occurrence; negative values indicate mutual exclusivity.
  4. Fisher's exact test — P-values are computed using the exact hypergeometric distribution, not an approximation.

All of this runs locally in Rust. A survival analysis across 10,945 samples completes in milliseconds.

Charting with Kuva

BioMCP's charts are powered by Kuva, an open-source Rust charting library built for this use case. Kuva renders to both terminal (Unicode block characters) and SVG.

Why a custom charting library instead of calling out to Python or R:

  • No runtime dependencies. BioMCP is a single binary. Adding a Python or R dependency would break that property.
  • Terminal-first. Kuva renders charts directly in the terminal using Unicode block characters. You see the chart where you ran the command.
  • SVG as structured data. Kuva's SVG output includes exact numeric values in element attributes, making it parseable by AI agents without vision models.
  • Validated chart types. Each study command declares which chart types are valid for its data shape. Invalid combinations fail with a clear error message instead of producing garbage.

The nine supported chart types: bar, pie, histogram, density, box, violin, ridgeline, survival, and custom comparison charts.

Available studies

BioMCP ships with three downloadable studies:

Study Cancer Type Samples
msk_impact_2017 Pan-cancer (mixed) 10,945
brca_tcga_pan_can_atlas_2018 Breast cancer 1,084
brca_sanger Breast cancer 100 
biomcp study list
biomcp study download msk_impact_2017

Try it

Install BioMCP and download a study:

uv tool install biomcp-cli
biomcp study download msk_impact_2017

Run your first survival analysis:

biomcp study survival --study msk_impact_2017 --gene TP53
biomcp study survival --study msk_impact_2017 --gene TP53 --chart survival

Explore co-occurrence:

biomcp study co-occurrence --study msk_impact_2017 \
  --genes TP53,KRAS,PIK3CA,EGFR --chart bar

Full documentation: Chart Reference.