Nc values and positive selection

David Cutler dave at dmitri.ucdavis.edu
Thu Sep 23 11:40:46 EST 1999


MArtin John Cann <mjcann at med.cornell.edu> wrote:
: I am working on a cluster of genes in the fly that have very high Nc
: values (59.12, 59.29, and 58.45).  A selection of "surrounding" genes in
: the genome have lowered values of Nc of 34 to 49 reflecting a partial
: bias in codon usage.  To what extent can I use these high values of Nc
: as evidence for positive selection at the locus?  Values of pi and theta
: are lowered compared to the genome average and Tajimas D statistic is
: -ve but not sufficiently so to reject the neutral mutation hypothesis.
: Does this add up to reasonable evidence for positive selection?

: Any advice would be greatly appreciated.

Fascinating question.  I'm not sure I'm totally following your logic, though.
First how was Nc inferred? I'll assume Hudson's method correct me if I'm
wrong...although Wakley and Hey won't materially change any of the following.

At first blush inferred Nc values at 3 loci all within 1% of each other is
shocking.  So shocking that I would immediately suspect something strange
is going on.  You say these are "a cluster of genes."  Do you have reason
to suspect there has been gene conversion or recent duplication between the
three loci?  If so, we might reasonably think of all 3 values as simply a
single locus, repeated 3 times.

If this is a single locus repeated 3 times, I would suggest there is probably
no significant difference between Nc=59 and Nc=34.  The tails on Hudson's 
estimator are enormous, and one generally thinks of Nc=25 as not statistically 
distinguishable from Nc=100.  These are rules of thumb, and I could be
wrong, but at first blush I would be loath to make any inference that assumes
Nc=59 was in any way distinct from Nc=34.

Second, if we do assume that the apparently high Nc value is real (which I 
wouldn't), I'm not totally clear why you would conclude that this is 
evidence for positive selection.  A priori I would take it as evidence of
negative selection (if many sites are hanging around in mutation/selection
balance, there will be an apparent excess in sites that pass the "4 gamete"
test, and consequently an apparent increase in Nc).  Of course, this is
not exactly my specialty, and some obvious argument is probably escaping me.
Reduced values for theta are also consistent with negative selection.  Why
is that you suspect positive selection?

Good Luck,
dave cutler
djcutler at ucdavis.edu 




More information about the Mol-evol mailing list