engleski

Statistical bioinformatics and genetics: challenges and lessons from thyroid autoimmunity

Autoimmune thyroid disorders (AITD) represent the 2nd most frequent organ-specific autoimmune disease caused by an interaction between well established genetic component and environmental factors. A large variety of statistical methodologies at various levels of complexity, from single-marker tests to multimarker data mining techniques for gene-gene interactions, centered around three main approaches: case-control association studies, linkage analysis and intrafamilial linkage disequilibrium-have been put forth to analyze genetic data and detect causally relevant variations. Yet, having its own merits and limitations, different methods, when applied in phenotypically and ethnically heterogeneous populations, have provided astonishingly discordant, often irreproducible data sets that failed to converge until recently. Hence, genetic counterparts of AITD have remained surprisingly elusive, since only a paucity of general autoimmune susceptibility loci, as well as thyroid-specific genes, has been identified thus far. As a results, the growing realization that no simple, low-dimensional solution, dominant gene, cosmopolitan variant or effect exist, has gained an almost universal acceptance, leading to the model in which predisposition to AITD depends on higher-order genetic networks, rather than unique, obligatory susceptibility loci. In response, tools have evolved over the recent years, thus allowing less expensive, high-dimensional, genome-scale, model-free analyzes motivated by the failure of the prioritization concepts based on candidate gene approach to substantially enrich AITD gene pool. Since then, tremendous quantity of sequence data generated by genome projects plus whole-genome epidemiological and gene expression data, has been far outpacing our ability to decipher its implications for the pathogenesis of common multifactorial diseases, facing genetic epidemiologists and computational biologists with numerous non-trivial predictional, computational and statistical challenges. Consequently, methods for disease gene identification are rapidly evolving, and the use of flexible class of mathematical modeling methods, as recently shown in AITD, for combinatorial incorporation of positional and functional informations, data from candidate-gene, genome wide association studies and from basic bioinformatics, with careful modeling of intermediate phenotypes and phenocopies, could provide valuable insight into the genetic architecture of complex diseases.

Thyroiditis; Autoimmune; Genetics; Population; Computational Biology; Statistics

nije evidentirano