Aubrey de Grey's COA (Cancelling out Argument) and why I disagree with it.

In Aubrey de Grey's 2015 Paper Do we have genes that exist to hasten aging? New data, new arguments, but the answer is still no he puts forward a number of arguments as to why he believes aging to be other than programmed aging (PA).

One of these argument is his "COA Cancelling out argument". I extract from the paper his reasons for this.

STRONGER CHALLENGES TO PA 
A simple but remarkably rarely stated argument against 
PA, which here I shall term the “cancelling-out argument” 
(COA), is that it is impossible for a species to maintain two 
sets of genetic pathways whose selected actions diametri
cally oppose each other. Specifically, since we clearly have a 
large amount of genetic anti-aging machinery (repairing and 
pre-empting the damaging side-effects of metabolism), we 
cannot also have pro-aging machinery that accelerates the 
accumulation of such damage, unless the latter machinery is 
selected for another purpose. The logic leading to this con
clusion is simple. A large proportion of mutations just 
slightly impair the function of a gene, thereby mildly degrad
ing the performance of systems in which the gene partici
pates. Thus, in a species that possesses both pro-aging and 
anti-aging machinery, it will frequently be the case that mild 
loss-of-function mutations will occur more-or-less simulta
neously in both. The net effect on the rate of accumulation of 
aging damage will be nil: there will be slightly less pre
emption and repair and also slightly less acceleration, so the 
two will cancel out. (Note that there is no implicit assump
tion here that any of these genes’ function is constant with 
age.) And if there is no phenotype – no difference, at any 
age, between the physiological performance of an individual 
with these new mutations and an individual without them – 
then there can be no selection, whether at the level of the 
individual or at the population level, to eliminate the muta
tions. Thus, their survival will be at the mercy of random 
genetic drift, just as for any single mutation that has no effect 
at all on the performance of its host gene. And that means 
that such pairs of mutations will continue to accumulate over 
time, progressively degrading the performance of both the 
pro-aging and the anti-aging machinery at equal rates, until 
one of them – namely, necessarily, the pro-aging machinery 
  – has mutated into oblivion. 
Can PA survive this argument? Two options seem avail
able. One is that the optimum distribution of ages in the 
population is one that cannot actually be delivered by dam
age accumulation on its own, such that pro-aging machinery 
is required to adjust it. In this scenario, the pro-aging ma
chinery would exert its effects in an age-dependent pattern 
that differs from the age-dependency of the anti-aging ma
chinery: for example, it could make the age distribution more 
“rectangular” (kill individuals off very efficiently above a 
certain age but hardly at all at younger ages). The other is 
that the presence of both types of machinery allows the spe
cies to respond more nimbly to changed extrinsic mortality 
risks, in essentially the same way that “futile cycles” of bio
chemical reactions permit a more rapid response to changes 
in substrate concentrations. Both these defences of PA 
against COA merit discussion, but they do not seem com
patible with available data. 
First, it appears that the distribution of ages at death seen 
in protected populations is absolutely consistent with NPA 
[9]. This is not merely a confirmation of NPA, but also a 
challenge to PA, because (as noted above) COA demon
strates that any putative pro-aging machinery needs to de
liver a phenotype, and if the phenotype is not at the level of 
mean lifespan (as discussed in the previous paragraph) it 
must be at the level of secondary features of the distribution 
of ages at death, i.e. the shape of the survival curve. In the 
wild, the picture is ostensibly murkier [10], but on closer 
inspection it can be seen that – especially in metazoans – the 
wide variety of patterns of mortality as a function of age in 
different species overwhelmingly occurs in the early and 
middle periods of life, when aging is not a major contributor 
to death and vagaries of ecological niche can dominate. 
The other defence of PA is perhaps even more starkly 
undermined by data, in the form of the observed rate at 
which successive generations of a population change their 
rate of aging in response to abruptly altered extrinsic mortal
ity conditions [11, 12]. While from other perspectives it 
might be considered remarkable that a population could 
halve its rate of aging over only a few hundred generations 
(far fewer than, for example, the number that separates us 
from our most recent common ancestor with chimpanzees, 
for example), from a pro-PA point of view is it distinctly 
underwhelming. Can the individual-benefit “hit” of a delib
erately shortened lifespan truly be outweighed by so sluggish 
a response, as is required by PA? The implausibility of this 
presumption strongly challenges the “futile cycle” defence 
against COA.

My view about this is that it is too narrow a definition to be used as an argument against programmed aging in general. We can actually start with the title of his paper. If you start with the hypothesis detailed here: https://citrate.science/2025poster/poster2025.html then as complexes 1 and 3 generate ROS and ROS damage accelerates aging through transversions and deletions in mtDNA then genes exist the modification of which accelerates aging by generating more ROS.

However, dealing with the COA ("Cancelling out Argument") itself:

• a) This only covers pathways which have no other purpose.
• b) This does not include pathways selected at different times. Hence if one of the pathways is the generation of ROS by complexes 1 and 3 (which is partially defined in mtDNA) then COA does not apply as the aspects defined in mtDNA (MT-D1-6 and MT-CYB) are selected during follicular atresia. This is perhaps the strongest argument against COA applying.
• c) This requires subtle balancing such that changes to pathways balance out.

Hence in a broader sense COA is quite narrow in its application and does not apply in particular to my own hypothesis. It should not be used as a general argument against Programmed Aging.