Genotype Pattern Mining for Pairs of Interacting Variants Underlying Digenic Traits

Abstract

Some genetic diseases (“digenic traits”) are due to the interaction between two DNA variants, which presumably reflects biochemical interactions. For example, certain forms of Retinitis Pigmentosa, a type of blindness, occur in the presence of two mutant variants, one each in the ROM1 and RDS genes, while the occurrence of only one such variant results in a normal phenotype. Detecting variant pairs underlying digenic traits by standard genetic methods is difficult and is downright impossible when individual variants alone have minimal effects. Frequent pattern mining (FPM) methods are known to detect patterns of items. We make use of FPM approaches to find pairs of genotypes (from different variants) that can discriminate between cases and controls. Our method is based on genotype patterns of length two, and permutation testing allows assigning p-values to genotype patterns, where the null hypothesis refers to equal pattern frequencies in cases and controls. We compare different interaction search approaches and their properties on the basis of published datasets. Our implementation of FPM to case-control studies is freely available.

Overview of frequent pattern mining

“…However, this practice is fallacious as a strong disease association of a variant has the effect that it will show up in many genotype patterns. In other words, a genotype pattern may show a strong disease association mostly because of main effects of one or the other of the two participating variants rather than an interaction effect [ 17 ]. If the aim is to uncover interaction effects, the recommended approach is to first remove individually significant variants and only then proceed to genotype pattern analysis.…”

“…As a case in point, we apply our GPM approach [ 17 ] to a published dataset on wet age-related macular degeneration collected in Hong Kong [ 73 ] and analyze results for significance and discovery. This dataset contained 96 cases and 127 controls, each genotyped for 81,934 variants genome-wide.…”

“…Pairwise analyses may be carried out at the level of genotype, variant [ 74 ], or gene [ 75 ], with respective decreases in granularity and total numbers of pairs. Genotype pairs (patterns) offer the greatest precision as signals of a single genotype pair may be hidden among the nine genotype pairs in a pair of variants [ 17 ]. On the other hand, searching among very large numbers of genotype pairs requires good computer resources.…”

“…Various approaches to handling the latter problem have been discussed [ 18 ], but in practice, it has been almost inevitable that some observations or variables must be deleted. For example, when patterns of genotypes are mined for their association with disease [ 17 ], preprocessing the data may involve deleting either individuals or DNA variants with a proportion of missing genotypes exceeding some threshold, whichever is more tolerable. More sophisticated approaches have been proposed in the form of noise tolerant item sets called approximate frequent itemsets (AFIs) [ 19 ], which obviate the need for deleting data but are computationally more demanding than frequent itemset mining (FIM) methods in error-free data.…”

“…We previously exploited FPM methodology in genome-wide disease association testing [ 16 , 17 ], where each individual is assigned a genotype at each of a large number of DNA variants, and two classes of individuals exist; cases are those affected by a heritable disease and controls are unaffected. Classically, we would test each variant to see whether a genotype at that variant is associated with disease, that is, shows higher frequency in cases than controls.…”