The Punctuated Clock

Arlin Stoltzfus arlin at
Sat Jun 14 22:21:54 EST 1997

Richard M Kliman wrote:
> In article <andrew.rambaut-ya023380001306971245430001 at>,
> Andrew Rambaut <andrew.rambaut at> wrote:
> >In article <5nmp30$ooh at>, rkliman at (Richard M
> >Kliman) wrote:
> >
> >> >Many mutations in protein coding DNA do not result in a change in amino
> >> >acid due to the redundancy of code. Thus these occur despite natural
> >> >selection as there is no phenotype to be selected.
> >>
> >> This is not always true.  There is considerable evidence for selection
> >> acting on silent substitutions, probably favoring those that increase the
> >> efficiency and/or fidelity of translation.
> >
> >The fact that many substitutions do not result in AA changes IS true. Some
> >of these may be select for or against due to tendancies related to G/C bias
> >etc. However it seems reasonably to say that most will simply be neutral -
> >i.e. they had no affect on phenotype and were fixed by drift. Indeed is
> >seems likely that many AA replacement substitutions will also be between
> >AAs that have a minimal effect on the protein structure (i.e. changes tend
> >to be withing AA groupings with respect to size and hydrophobicity).
> Agreed, sort of.  There is a difference between neutral in an
> evolutionary sense and neutral in a phenotypic sense.  Silent
> substitutions *do* seem to be subject to selection in many
> microorganisms, C. elegans and Drosophila.  However, being subject to
> selection (i.e., having a phenotypic effect) and being predictably
> influenced by selection are two different things.

I don't agree that such a distinction exists.  The definition of
does not refer explicitly to any phenotype other than fitness, and this 
is because there are no other restrictions on what is the phenotype 
of a neutral allele.  IMHO, there is alot of confusion about 
what "neutral" means, and what are its implications.  The chance that 
a newly arising advantageous allele with a selection coefficient of _s_ 
will be fixed by selection is approximately _s_ (most such alleles 
are quickly lost by drift).  In the strictly hypothetical (i.e., not 
real) instance of a newly arising allele that has absolutely no 
difference in fitness from the wild-type allele, the chance that 
it will be fixed by random genetic drift is approximately _1/N_ (where 
_N_ is the effective population size), that is, the chance is equal 
to its initial frequency.  Kimura's neutral theory is all about the 
interesting situation in which _s_ is so small that it is comparable 
to or smaller than _1/N_.  

If _|s|_ is much less than _1/N_, i.e., 

  s << 1/N

it doesn't even matter whether _s_ is positive (advantageous) or 
negative (disadvantageous), because its chance of fixation is dominated 
by the chance of fixation by drift.  This situation is mathematically 
tractable, because selection can be ignored, and this is what gives 
the neutral theory its power.  Accordingly, "neutral" (or "strictly 
neutral") is defined in terms of _|s| << 1/N_.  A common mistake 
is to assume that _s_ for a neutral allele must be equal to 
0.0000000000000000000000000000000000000000000000000 etc (i.e., 
zeros to an infinite number of decimal places), e.g.,.  
> Take-home message: for a variety of reasons, silent substitutions may
> often be effectively neutral, in the sense that one can not reject a
> neutral model.  However, that does not make them truly neutral - i.e.,
> they do not have a selection coefficient of zero.  I think there is
> considerable evidence for effective selection on silent mutations, and
> one should be careful about assuming neutrality.

More information about the Mol-evol mailing list