|
Abstract:
|
A complete genotype-phenotype map is one of the major promises of statistical genomics to the practice of preventing or treating diseases. Since genes do not operate in vacuum, many researchers believe that prediction power can be significantly improved with genetic interactions appropriately modeled. However, as the number of combinations of genetic variants easily becomes astronomical, it is challenging to balance overfitting and underfitting when working with millions of genetic variants but limited number of subjects on the level of thousand.
Between genomes and phenotype, there are important cellular traits like gene expression, DNA methylation, and etc. These traits are closer to the phenotype (than genomes do) although not stably inherited. In theory, they are excellent middle-traits that bridge the gap between genotypes and phenotypes. In this talk, the models developed to leverage such in-between-omes will be presented, together with their applications in human and model organism data.
|
