This repo contains the code for data analysis for the Meta-GC project.
Meta-GC is a case-control study nested in the EPIC prospective cohort. It includes 437 GC cases matched to 437 cancer-free controls. Plasma samples from participants were analysed by untargeted metabolomics using a Thermo Q-Exactive mass spectrometer. After alignment and filtering, over 2000 spectral features were found in positive and negative ionisation modes.
This code performs multivariable conditional logistic regressions on each feature in turn, using several different models. The resulting p-values are adjusted by controlling the false discovery rate (FDR) and those significant at FDR p < 0.05 retained. The data are also visualised on smile plots of OR vs p-value.
Different subgroups of data are also tested, and H. pylori status taken into account.