population history
Wall JD, Lohmueller KE & Plagnol V 2009 Detecting ancient admixture and estimating demographic parameters in multiple human populations. Mol Biol Evol 26:1823-1827.
- our analysis also corrects a misuse of the popular coalescent simulator ms (Hudson 2002) in our previous study (Plagnol and Wall 2006)
- this error changes the qualitative results found earlier
- an updated version of ms is available from R. Hudson's web site
- Plagnol and Wall (2006) considered a two-population model where there was migration between the populations, and then at some time in the past, the two populations merged into the same ancestral population
- in ms (cf. Hudson 2002), two populations are joined (going backwards in time) using the "-ej" flag
- this flag moves lineages from one population into the other population
- in the version of ms used in Plagnol and Wall (2006) the "-ej" flag did not set the migration rate (going backwards in time) equal to zero at the time the two populations merged
- instead, the migration rate was left unchanged
- the "-eM" flag would have to be used to explicitly set the migration matrix equal to zero at the time the two populations merged
- thus, in Plagnol and Wall (2006), it was assumed that all lineages were in the same ancestral population, when in reality, some of those lineages could migrate back to the other population
- the net effect of this is to model population structure in the ancestral population without intending to do so
- if there was no migration between the two populations at all, then the "-ej" flag would correctly move the lineages from one population to the other without allowing any migration into the second population
- simply because there is no migration being allowed
- the effect that this error had on demographic parameter estimates is substantial