Molecular reclassification of Crohn's disease by cluster analysis of genetic variants

Crohn's Disease (CD) has a heterogeneous presentation, and is typically classified according to extent and location of disease. The genetic susceptibility to CD is well known and genome-wide association scans (GWAS) and meta-analysis thereof have identified over 30 susceptibility loci. Except for the association between ileal CD and NOD2 mutations, efforts in trying to link CD genetics to clinical subphenotypes have not been very successful. We hypothesized that the large number of confirmed genetic variants enables (better) classification of CD patients.

To look for genetic-based subgroups, genotyping results of 46 SNPs identified from CD GWAS were analyzed by Latent Class Analysis (LCA) in CD patients and in healthy controls. Six genetic-based subgroups were identified in CD patients, which were significantly different from the five subgroups found in healthy controls. The identified CD-specific clusters are therefore likely to contribute to disease behavior. We then looked at whether we could relate the genetic-based subgroups to the currently used clinical parameters. Although modest differences in prevalence of disease location and behavior could be observed among the CD clusters, Random Forest analysis showed that patients could not be allocated to one of the 6 genetic-based subgroups based on the typically used clinical parameters alone. This points to a poor relationship between the genetic-based subgroups and the used clinical subphenotypes.

This approach serves as a first step to reclassify Crohn's disease. The used technique can be applied to other common complex diseases as well, and will help to complete patient characterization, in order to evolve towards personalized medicine.