Genome-wide prediction of imprinted murine genes
Press Reports: Wall Street Journal
31 May 2005: A machine learning approach was used to both identify imprinted gene candidates and predict their parental expression preference across the entire mouse genome. This is the first large-scale prediction of imprinted genes based on DNA sequence characteristics alone.
Of 23,788 annotated autosomal mouse genes, 600 (2.5%) were predicted with high probability to be imprinted, 64% of which are predicted to exhibit maternal expression. The number, type, and relative orientation of repeated elements flanking a gene were found to be particularly important in predicting whether a gene is imprinted.
This subset of genes is particularly important medically becasue imprinted genes could function as epigenetically labile targets for linking environmental exposures during pregnancy to the susceptibility of developing chronic disorders as adults (Waterland and Jirtle, Nutrition 20: 63-68, 2004). Consistent with imprinted genes potentially being mechanistically involved in the developmental origins of adult diseases (Barker, J Am Coll Nutr 23(6 Suppl): 588S-595S, 2004) was our ability to identify strong candidate genes for complex human conditions where parent-of-origin inheritance is involved, including Alzheimer disease, autism, bipolar disorder, diabetes, male sexual orientation, obesity, and schizophrenia.
For example, the mouse homologue of human NKX6B at 10q26 (IDDM, 605955) was predicted to be imprinted and maternally expressed, making it a good candidate for maternal transmission of male homosexuality linked to this chromosomal region (Mustanski et al., Hum. Genet. 116: 272-278, 2005). Similarly, the mouse homologue of human GAD2 at 10p12 (IDDM, 138275) was predicted to be imprinted and maternally expressed. It is therefore a good candidate for maternal transmission of increased risk for obesity linked to this chromosomal region (Dong et al., Am. J. Hum. Genet. 76: 427-437, 2005).