Multi-omics Approach for Identification of Etiopathogenesis and Early Detection Biomarkers of Stomach Cancer Precursor Lesions

The study population is part of the Swedish Twin Registry (STR), which has since its establishment in the late 1950s collected questionnaire data from all twins born after 1886. The data in this specific study is a subset of the Screening Across the Lifespan (SALT) study within the STR, in which all twins born between 1911-1958 were interviewed between 1998-2002. The subcohort named TwinGene was set-up between the years 2004 and 2008 when participants were invited to respond to a questionnaire on common diseases and provide a blood sample. Samples were collected from 12,618 twins born 1958 or earlier, of which blood sample from 12,609 were available. After excluding 10 samples which were unable to link to the available environmental data, a total of 12,599 blood samples were analyzed. Corpus-dominant CAG was characterized by a PGI/PGII ratio of less than 3. Metabolomic profiling using serum samples has been performed based on Nightingale Blood Biomarker Analysis platform. Proteomic profiling will be performed by both Scanning SWATH and OLINK® Explore 384 Oncology panel. History of H. pylori infection is examined by measuring serum IgG antibodies against H. pylori, using ELISA. Detailed lifestyle information collected by questionnaires includes education, smoking, snuff dipping, alcohol drinking, drug use, diet, and height/weight, etc. Metabolites will be log transformed prior to analyses, due to its usually skewed distribution. For each metabolite, firstly, CAG patients will be compared with all the CAG-free controls. Wilcoxon-Mann-Whitney test or Kruskal-Wallis test will be used for comparing differences of protein expressions between CAG and non-CAG groups, and multiple comparisons will be adjusted using Bonferroni correction. Generalized estimation equation (GEE) models with the robust option will be fitted to estimate the odds ratios (ORs). Second, in the comparison with MZ co-twin controls and DZ co-twin controls, only complete twin pairs with discordant CAG will be included in the study. Specifically, conditional logistic regression models will be used to control for the matching within co-twin pairs. The investigators will further combine metabolomics and proteomics data, and try to build up a CAG prediction model. Covariates will include age, sex, H. pylori seropositivity, education level, smoking, snuff dipping, and alcohol drinking. Joint effects of different metabolites and interaction with other covariates will also be examined.