Analysis code for "Germline causality is not tumour prognosis: dissecting the genetic and prognostic architecture of colorectal cancer in the COADREAD cohort"

The archive contains 60 analysis scripts covering preprocessing, genome-wide association study quality control, SNP-to-gene mapping, prognostic modelling, external validation, Mendelian randomisation, colocalisation and figure generation; download helpers for public genome-wide association summary statistics, expression quantitative trait locus resources and Gene Expression Omnibus series; and query scripts for Open Targets, DGIdb, the Human Protein Atlas and the protein quantitative trait locus catalogue.

The code is written in R 4.5.3 with a small amount of Python 3.13 for download and query utilities. No third-party dataset is redistributed: all inputs are publicly available and must be obtained from their original sources, as listed in the accompanying README. The analyses used only publicly available summary-level statistics and previously published de-identified data, so ethical approval and informed consent were not required.

A single parameter must be edited before running the code: the project root constant BASE at the top of analysis_code/00_core_functions.R. All other paths are derived from it. The random seed is 20260908.