Sorry, but it is simply not possible than over 50% of de novo mutations are highly deleterious. As you mentioned above, on average, humans acquire ~74 de novo single nucleotide variants (SNVs) per genome per generation. That would mean that every individual has over 37 highly deleterious mutations,
in addition to those inherited from each parent. To keep it simple, let's say even 35 highly deleterious mutations, which would be an underestimation (under 50%) according to what Kimura-san claims. A person would therefore be born with 35 de novo highly deleterious mutations but inherited about 35 highly deleterious mutations from his/her parents (half of each parent's 35 mutations but x2 as they are 2 parents). By that logic, each generation would ineluctably add 35 highly deleterious mutations to the germ line. It quickly rises to the thousands of highly deleterious mutations per individual. Mildly deleterious mutation can cause poor vision, allergies, low intelligence, frequent colds, and so on. Highly deleterious mutations are things that can cause foetuses to self-abort to people to be born with severe genetic conditions like cystic fibrosis or Huntington's disease, which lead to deaths at a relatively young age. It just doesn't make any sense to say that all people are born with thousands of highly deleterious mutations, including at least 35 de novo ones. That would be the end of the species.
Anyway, anyone who has studied biology should know that statistically most mutations are synonymous because of the high redundancy in the way amino acids are encoded by DNA (or actually RNA during
translation). To illustrate for non biologists, here is a table of RNA translation to amino acids. This is the way our genome encodes proteins. Each amino acid is encoded by a three genetic bases (A, C, G or T in DNA, which become U, G, C or A after
transcription to RNA). For example, if you want to produce Methionine, the amino acid that signal the beginning any protein sequence, you will need the RNA sequence AUG. If a mutation occurs in any of these three 'letters', it won't produce Methionine by another amino acid (e.g. Isoleucine if the final G becomes an A), and the protein won't be made. This would be an example of deleterious mutation, as the body cannot produce one type of protein. In most cases this is bad enough to guarantee the non-viability of the foetus.
However, as you can see on the table below, other amino acids can be encoded by any of 2, 3, 4 or even 6 different sequences in the case of Leucine. So if a mutation occurs in a CUA sequence for Leucine, chances are that the resulting protein will still be Leucine. Change the C into a U or an A and you still get Leucine. Change the final A into a C, G or U, and you still get Leucine. A mutation in the central U would result in a different amino acid. But even so, a mutation from U to C would give Proline, which is another hydrophobic amino acid with similar properties that is unlikely to cause major disruptions to the protein structure. So the only cases in which a mutation would be highly deleterious here is if the CUA sequences becomes CCA (Glycine) or CGA (Arginine), as they would turn a hydrophobic amino acid into a hydrophilic or a positively charged one. In this case, out of 9 possible mutations (3 for each letter), 7 are neutral and only 2 are deleterious. Considering that amino acids with similar properties (e.g. hydrophobic) have similar sequences, the chances of mutations changing those properties are not that high.
Any change that tempers with an initial Methionine or results in a stop codon (UAA, UGA or UAG) would be particularly deleterious. You can see here a few examples of highly deleterious mutations.
I haven't calculated all the possibilities of mutations for each amino acid, nor applied those results to the percentage of each sequence present in the human genome. That is the kind of humongous calculation that should be left to computers. But at first sight it looks like over half of mutations would be silent. More than mutations, it is deletions and insertions that tend to cause major problems, as they cause
frameshift mutations, potentially altering whole genes.
Furthermore, proteins are only encoded by our exome (coding DNA), which represents a mere 1% of the total genome. Any mutation that occurs in the 99% of non-coding DNA would in all likelihood have no effect whatsoever on fitness or phenotype. That is what we observe in Y-DNA mutations. Over 15,000 Y-DNA mutations have been identified at present, but
only a tiny minority of them seem to have been selected by evolution because they benefited the carriers.
