I know there is extensive research into the taxonomy and evolution of protein domains and many connections are, and are continuing to be, found between numerous domains. However, as far as I can see, new domains always arise from preexisting ones. I have seen no paper, in any field, showing the development of a new protein domain from a random piece of DNA which just mutates a start codon and synthesizes a new protein with no apparent function. I know this is theoretically possible but has anyone seen any evidence it has happened in any form of life in any geological period of time? OR are there some other mechanisms that could generate novel domains from non-coding DNA that I have overlooked when searching the journals?

My thinking is that, if de novo protein domain generation doesn't occurr, does it mean that all protein domains are a kind of "progeny" of a single domain that arose at the onset of life as we know it (or maybe even was the first lifeform on Earth). It seems very strange to me that this hasn't been pointed out by now since it's a relatively simple deduction. Thus, I'm asking either for a reference paper showing evidence for the evolution of de novo protein domains from a random piece of DNA that somehow managed to turn into a gene or for a reference to a paper discussing the idea that all proteins originate from a common ancestor existing at the onset of life and even, possibly, a single domain as old as life itself?