Bioinformatics and the identification of imprinted genes in mammals

With the identification of DNA sequence features that are unusually distributed in regions undergoing genomic imprinting, the basis is created for the bioinformatic prediction of the remaining imprinted genes in mammalian genomes, believed to number several hundreds. It is technically challenging to prove that a gene is imprinted, so any technique that narrows down the candidates for analysis is of obvious value. Genome sequence annotations can be mined to create large datasets for analysis, introducing a number of statistical challenges. We discuss how these challenges can be addressed, allowing every gene in the genome to be assigned a relative likelihood of imprinting on the basis of their similarity to known imprinted genes in terms of their most discriminatory sequence characteristics.