He uses the SRC gene family as an example case. Summarizing his explanation, the SRC genes code the family of proteins (8 in total) that activate cellular division.
This gives 2 wrong impressions, firstly that activating cellular division is all the SRC gene family proteins do, which it isn't, and secondly that they are the only genes that activate cell division which they aren't. If these were true it would greatly strengthen his argument about redundancy, but they aren't true.
Cell proliferation is only one of the many biological functions the SRC genes are involved in, they also regulate aspects of the immune response, cellular differentiation, motility and adhesion as well as a number of other things.
Non-synonymous point mutations in these genes are often lethal because they code proteins that don't "turn-off" leading to cancerous cell division.
As Percy says, it is very hard to find anything that actually backs up this statement. Certainly we do know about mutations that produce constitutively active forms of SRC especially, since it has massive historical importance in our understanding of cancer and oncogenes, but to turn that into a blanket statement about many more mutations in all of the SRC family genes extends well beyond where the evidence takes us.
A more usable approach than looking at orthologs in one genome would be to look at homologues in different genomes. NCBI have helpfully already calculated the similarities for a whole lot of pairwise comparisons
here. If the chimpanzee SRC protein can differ by 40 amino acids why should we imagine that those substitutions are not tolerable in the human form of the protein? What does this say about the genes being 'highly selective against mutation'?
As an aside this data once again nicely demonstrates that protein similarity is distinct from genetic similarity since the chimpanzee has the lowest protein identity score compared to the human protein of any of the mammals studied, lower even than the chicken's score, but the chimpanzee's DNA identity score is still the highest.
How is it possible for evolution to generate redundancy within genetic systems?
The very same way that Borger dismisses, via gene duplication. As long as one looks at the reality of the biology rather than presupposing that the genes essentially can't tolerate any non-synonymous mutations, then the problem suddenly ceases to exist.
Percy made several other important suggestions for how such redundancy might come about even allowing that all non-synonymous mutations might be lethals.
TTFN,
WK