Due to the tremendous cost of the traditional mutation-accumulation approach
(the
Bateman–Mukai technique), data are rare for deleterious mutation
parameters
such as genomic
mutation rate, selection and dominance coefficients. Two alternative approaches
have been
developed (the Morton–Charlesworth and Deng–Lynch techniques).
Except for the Deng–Lynch
method, the statistical properties (bias and sampling variance) of these
techniques are poorly
understood; therefore we investigated them using computer simulation. With
constant fitness
effects of mutations, the Bateman–Mukai (assuming additive effects)
and Deng–Lynch (assuming
multiplicative effects) techniques are unbiased; the Morton–Charlesworth
technique (assuming
multiplicative effects) is very biased if fitness is used in
the regression to estimate h, but slightly
biased if the logarithm of fitness is used. With variable fitness
effects,
all techniques are biased. The
Deng–Lynch technique is statistically better than the others except
when fitness is used to estimate
the average degree of dominance in selfing populations with the
Morton–Charlesworth technique.
If fitness effects are multiplicative but additivity is assumed, the
Bateman–Mukai technique is
biased under constant fitness effects, and less biased under variable
fitness effects relative to when
fitness effects are additive (as assumed by the technique). Our
study not only quantifies the degree
of bias under the biologically plausible situations investigated,
thus forming a basis for correct
inference of the true parameters by using these techniques, but
also provides insights into the
relative efficiencies of these techniques when the same number of genotypes
are handled
experimentally.