A structural split in the human genome.

Abstract

BACKGROUND: Promoter-associated CpG islands (PCIs) mediate methylation-dependent gene silencing, yet tend to co-locate to transcriptionally active genes. To address this paradox, we used data mining to assess the behavior of PCI-positive (PCI+) genes in the human genome. RESULTS: PCI+ genes exhibit a bimodal distribution: (1) a 'housekeeping-like' subset characterized by higher GC content and lower intron length/number, and (2) a 'pseudogene paralog' subset characterized by lower GC content and higher intron length/number (pTpA mutation and intron insertion. We propose a model of evolving biological complexity in which environmentally-selected gains or losses of PCI methylation respectively favor positive or negative selection, thus polarizing PCI+ gene structures around a genomic core of ancestral PCI- genes.