cbioportal-database

What it does

The skill provides concrete steps and code snippets to access cBioPortal data via REST API for cancer genomics. It covers querying cancer types, studies, samples, mutations, copy number alterations, clinical data, gene expression, and gene metadata. It includes example workflows for assembling mutation and CNA data, retrieving clinical attributes, and obtaining expression values, with explicit endpoints, parameters, and data handling patterns. It also demonstrates how to map gene symbols to Entrez IDs and how to fetch per-study mutation frequencies for visualization.

How it works

• Uses base URL https://www.cbioportal.org/api for all API requests.
• Defines helper functions cbio_get and cbio_post for GET and POST requests, including headers for JSON communication and timeouts.
• Query 1 (Cancer Types and Studies): fetches cancer-types, then studies by keyword, and retrieves study metadata and molecular profiles.
• Query 2 (Somatic Mutations): obtains all samples for a study, constructs a mutation fetch body with sampleIds and entrezGeneIds, and requests mutations from the mutations fetch endpoint; summarizes mutation types.
• Query 3 (Copy Number Alterations): uses a discrete CNA profile (gistic) and fetches CNA data for specified genes; decodes numerical values to labels like AMP, GAIN, DIPLOID, LOSS, HOMDEL.
• Query 4 (Clinical Data): lists clinical attributes for a study, filters for survival-related attributes, fetches patient-level clinical data, pivots to a patient × attribute matrix, and reports OS counts.
• Query 5 (Gene Expression Data): retrieves RNA-seq expression from rna_seq_v2_mrna_median_normed_log2 and builds a gene-by-sample matrix for selected genes.
• Query 6 (Gene Details and Batch Lookup): fetches single-gene metadata, resolves multiple genes from Hugo symbols via a fetch endpoint, and displays Entrez IDs and types.
• Query 7 (Visualization — Mutation Frequency Barplot): loops over several TCGA PanCancer Atlas studies, computes TP53 mutation frequency by fetching samples and mutations, and renders a horizontal bar plot with counts and percentages; saves an image and prints a summary table.
• The skill emphasizes concrete file names, endpoints, gene identifiers, and example data flows; it requires the claude-code tool as declared.

When to use it

• When you need somatic mutation, CNA, expression, and clinical data for TCGA or other cBioPortal studies.
• When you want to identify which studies have data for a given cancer type or gene and to prepare data for downstream analyses like survival or differential expression.
• When you need gene-symbol to Entrez-ID mapping for query construction.

What it can touch

• Tools: claude-code
• Endpoints used include: https://www.cbioportal.org/api, with endpoints such as cancer-types, studies, studies/{study_id}, molecular-profiles, samples, mutations/fetch, studies/{study_id}/clinical-data, genes/TP53, genes/fetch, etc.
• Data touches genes by Entrez IDs (e.g., 7157 for TP53) and Hugo symbols (e.g., BRCA1). It handles sample IDs and molecular profile IDs constructed from study IDs.

Caveats

• License declared: AGPL-3.0 for the skill; runtime license in the code is NOASSERTION in the repository metadata.
• Requires internet access; no API key is required for cBioPortal REST API.
• Rate limits described: no strict rate limits; implements a polite sleep between batch requests (time.sleep(0.2)).
• Results depend on the availability of data in cBioPortal for each study and gene; some queries may skip with errors and print a message.