Multi-Locus Nonparametric Linkage Analysis of Complex Trait Loci with Neural Networks

Abstract

Complex traits are generally taken to be under the influence of multiple genes, which may interact with each other to confer susceptibility to disease. Statistical methods in current use for localizing such genes essentially work under single-gene models, either implicitly or explicitly. In genomic screens for complex disease genes, some of the marker loci must be in tight linkage with disease susceptibility genes. We developed a general multi-locus approach to identify sets of such marker loci. Our approach focuses on affected sib pair data and employs a nonparametric pattern recognition technique using artificial neural networks. This technique analyzes all markers simultaneously in order to detect patterns of locus interactions. When applied to previously published sib pair data on type I diabetes, our approach finds the same genes as in the published report in addition to some new loci. For a specific two-locus model of inheritance, the power of our approach is higher than that of the currently used analysis standard.

Genome-wide scans of three independent sets of 90 Irish multiplex schizophrenia families and follow-up of selected regions in all families provides evidence for multiple susceptibility genes

“…Approaches that are more synthetic and powerful, for example haplotype analysis that incorporates multivariate and quantitative approaches, and that allows for epistatic interactions, may serve to produce more definitive results. It is anticipated also that the application of newer methods, eg neural networks 112 or classification and regression trees 113 and related algorithms, when applied to data from much larger samples probed with many more markers will also be productive. Even if such analytic advances are made, however, the use of intermediate phenotypes 114 as well as a battery of assays of gene function (eg differential expression 115 ) will be required to elucidate the genetic component of schizophrenia.…”

Selecting SNPs in two‐stage analysis of disease association data: a model‐free approach

“…Once a suitable set of markers has been selected, model building can begin in which main and interaction effects are estimated, for example, in multiple regression. More advanced curve fitting methods such as loess (nonparametric regression, computer-based curve fitting by a series of local regression curves; Cleveland, 1979) and artificial neural networks (Lucek et al 1998) are also available for investigating relationships between disease and a sufficiently small number of markers.…”

Genetics of type 1 diabetes mellitus

“…These include variance components methods for quantitative trait loci (QTL); 230,231 conditional tests for QTL by multiple regression analysis; 232,233 conditional and simultaneous search methods from high-resolution maps of identity-by-descent, 234 methods for searching for an additional locus given an established susceptibility locus, 235,236 interaction analysis, 237 and multi-locus nonparametric linkage analysis with neural networks. 238 Most of these newer methods focus on two-loci traits and their statistical properties (applications) for genome-wide searches of susceptibility genes are not well known. In current practice, the majority of whole genome scans for complex trait loci are conducted by multipoint methods [16][17][18][19][20][21]239,240 that work under the implicit assumption of a single trait locus or multiple trait loci, unrelated to each other.…”