Abstract
A recent surge of studies suggested that many novel genes arise de novo from previously non-coding DNA and not by duplication. However, since most studies concentrated on longer evolutionary time scales and rarely considered protein structural properties, it remains unclear how these properties are shaped by evolution, depend on genetic mechanisms and influence gene survival. Here we compare open reading frames (ORFs) from high coverage transcriptomes from mouse and another four mammals covering 160 million years of evolution. We find that novel ORFs pervasively emerge from intergenic and intronic regions but are rapidly lost again while relatively fewer arise from duplications but are retained over much longer times. Surprisingly, disorder and other protein properties of young ORFs do not change with gene age. Only length and nucleotide composition change, probably to avoid aggregation. Thus de novo genes resemble frozen accidents of randomly emerged ORFs which survived initial purging, likely because they are functional.
