Interacting genetic variants may contribute to the genetic architecture underlying biological features of colorectal tumours. The authors aimed to identify two-way interactions between genetic variants that associate with selected tumour features in colorectal cancer, using a large-scale genome-wide genotype resource and a screening-based interaction method.
The study's stated objectives were to explore interacting loci (pairs of variants) for association with four tumour characteristics and to illustrate the application of the BOolean Operation-based Screening and Testing (BOOST) method to a large genomic dataset.
The analysis used genotype data from colorectal cancer patients enrolled in the Newfoundland Familial Colorectal Cancer Registry. The variant dataset included 4,711,309 genotyped and imputed variants. Additional cohort-level sample counts and detailed phenotype distributions were not reported in the abstract-level source text.
The authors implemented a two-stage analytic pipeline. First, they applied the BOOST method to screen for candidate two-way variant interactions across the genome. BOOST is a Boolean operation–based screening approach designed to prioritize interaction models from large-scale genotype data.
After screening, the investigators selected the top 1,000 BOOST models and re-tested those candidate interactions using logistic regression to obtain more accurate association statistics. The paper reports logistic regression p-values for top interactions and highlights the statistical range observed among leading results.
Four tumour features were analyzed as outcomes in the interaction scans:
These phenotypes were used to evaluate whether interacting genetic loci associate specifically with tumour characteristics rather than overall disease risk.
The top 50 interactions in each tumour-feature dataset produced logistic regression p-values ranging from 1.78E-10 to 1.22E-05. Most of the variants implicated in top interactions were reported as noncoding, with some variants located within genes. Many of the genes flagged by these interactions had prior literature linking them to cancer-related processes.
The authors describe the results as preliminary and novel, and they emphasize that findings may be subject to replication in independent cohorts.
The study included functional annotation analyses and literature/database queries for selected variants and implicated genes. These follow-up analyses identified shared biological features across tumour phenotypes; the authors specifically mention links to concepts such as chemical dependency and tobacco use among annotated gene functions or disease associations. Full detailed annotation results and gene lists are provided in the study’s supplementary materials (as indicated in the source).
The authors report that the study received ethical approval from the Health Research Ethics Board (HREB) with reference numbers HREB #2018.051 and #2009.106. They confirm participant consent and adherence to reporting guidelines.
All data produced in the study are reported to be available upon reasonable request to the authors. Supplementary material and external links to data/code accompany the preprint.
Declared funders include the Marathon of Hope Cancer Centres Network and multiple national and provincial sources, with specific organizations listed in the preprint's funding statement.
The authors declared no competing interests.
The authors conclude that this is, to their knowledge, the first study to explore interacting variants associated with tumour features in colorectal cancer using large-scale genomic data and the BOOST method. They report that BOOST is feasible and useful for screening interactions in large genotype datasets and that the results provide preliminary, novel signals that may help explain genetic contributors to tumour phenotype heterogeneity.
They emphasize the need for replication in independent patient cohorts and caution that, as a preprint that has not undergone peer review, the findings should not be used to guide clinical practice. Specific replication plans, effect size estimates for interactions, and broader limitations beyond those implicit in a preliminary, preprint-stage study were not detailed in the abstract-level source text.