About 1.5% of the genome codes for proteins, and we understand proteins (and the genetic code) well enough to analyze mutations in those coding regions. About 70% of nucleotide changes in codons change the amino acid sequence. Most of those changes are deleterious.
According to Kryukov and coauthors [Kryukov et al 2007, AJHG ] , about 27% of those missense mutations are neutral, while 20% cause a loss of protein function. The remaining 53% have slight deleterious effects – their estimate is that the selective disadvantage ranges from 0.001 to 0.003.
Note that loss-of-function mutations are typically more deleterious, but many still have smallish effects in single copy. In fruit flies, the average recessive lethal only decreases fitness by 1 or 2 percent in heterozygotes.
This is all aimed at understanding the spectrum of fresh mutations. I would like to know what fraction decrease fitness by 50%, 20%, 10%, 5%, 1%, etc. If some population adopted customs that resulted in increased paternal age, the numbers of mutations in those classes would all increase, but at different rates. If the mutation rate went up by 50%, the number of dominant lethals would reach the new higher equilibrium in one generation, but it would take on the order of 100 generations for the 1% deleterious mutations to approach their new higher equilibrium, So, if we knew more about the mutational spectrum, we could better understand the impact in a population that had had high average paternal age for 100, 1000, or 10,000 years. Most of the old-dad cultures I am reading about look (to me) as if they came into existence in farming cultures, so are probably not incredibly old. The Australian aborigine pattern of gerontocratic polygyny could be well be older, but I don’t know how archaeology could resolve that question. Ancient Viagra bottles? Genetics might tell us: if the Australians have higher-than-usual numbers of deleterious mutations with small effects, more so than other populations, we would have to suspect that they’d been at it a long time.
One big problem is that most of the genome that does something significant and can thus mutate in ways that cause harm is noncoding. We don’t understand that component as well. Certainly I don’t. A rough estimate would be that 1.5% of the genome makes proteins while another 3.5% does something else functional - regulation of protein expression, regulation of other regulators, etc. The effect size of such mutations is probably biased towards small numbers, since few genetic syndromes serious enough to be called Mendelian diseases are caused by changes in noncoding parts of the genome. Such stretches of the genome are conserved enough that they’re obviously experiencing purifying selection, but most don’t cause flagrant disease when altered. Think of it this way: a loss-of-function mutation ruins the protein, but a regulatory mutation might just change the amount made or the timing of its expression. Sounds less serious.
Of course, as the functional fraction of the genome increases, there are that many more things to go wrong. Genetic load will be bigger. If selection is relaxed – as it is today – the human race will deteriorate more rapidly. It is not just that people in developed countries are experiencing selection for lower intelligence (as they are) – you have to think about mutational accumulation. While anyone still can think at all.
In the same way, the larger the fraction of the genome that plays a functional role in a particular organ, the more that organ’s efficiency will be compromised by genetic load. That’s what we see with the one-generation effects of advanced paternal age – they mess up the brain, as Kondrashov has pointed out.