Large Sample Size, Wide Variant Spectrum, and Advanced Machine-Learning Technique Boost Risk Prediction for Inflammatory Bowel Disease

Abstract

We performed risk assessment for Crohn's disease (CD) and ulcerative colitis (UC), the two common forms of inflammatory bowel disease (IBD), by using data from the International IBD Genetics Consortium's Immunochip project. This data set contains ~17,000 CD cases, ~13,000 UC cases, and ~22,000 controls from 15 European countries typed on the Immunochip. This custom chip provides a more comprehensive catalog of the most promising candidate variants by picking up the remaining common variants and certain rare variants that were missed in the first generation of GWAS. Given this unprecedented large sample size and wide variant spectrum, we employed the most recent machine-learning techniques to build optimal predictive models. Our final predictive models achieved areas under the curve (AUCs) of 0.86 and 0.83 for CD and UC, respectively, in an independent evaluation. To our knowledge, this is the best prediction performance ever reported for CD and UC to date.

Genetic architecture of colorectal cancer

“…[215] In fact, an extensive GWAS of inflammatory bowel disease recently showed that including hundreds of variants with suggestive, but not genome-wide significant associations, improved substantially risk prediction compared to models limited to GWAS findings. [216] Accordingly, it can be expected that incorporating a large number of genetic variants from genome-wide scans can further improve risk-prediction models, particularly if based on large sample sizes. [216–220]…”

Systems biology in inflammatory bowel diseases

“…Recently an assay, called ImmunoChip, has been developed including 200,000 single nucleotide polymorphisms (SNPs) relevant to IBD and other immune-mediated diseases 22 . Studies of SNPs and insertion-deletion polymorphisms identified a total of 163 loci associated with IBD and revealed important pathways involved in IBD pathogenesis such as host-microbe interactions and autophagy 5,21 .…”

Genomic risk prediction of complex human disease and its clinical application

“…Only recently have similar approaches begun being applied to human disease, with notable success in CD (discussed below), T1D, and inflammatory bowel disease [13,22,23 ,24-26], and to a lesser extent in other diseases such as multiple sclerosis [27]. Commonly used models include L1-penalized logistic regression [24,26], kernel support-vector machines (SVM) [13], L1-penalized SVM [11 ,12 ,22], and variants of Bayesian models employing a genetic relatedness matrix (GRM), including linear [23 ,28] and probit [15] regression. Regardless of the model used, the same principle applies to all: each method represents a mathematical mapping from the SNP data (either a large subset of SNPs or all SNPs) to the phenotype, which is then applied to new SNP data to produce predicted phenotypes, typically in terms of a continuous genomic risk score (GRS).…”