First, please update DataFrames.jl to 0.22 release to get the latest and bug free functionality.
Now regarding your code the simple transformation is (it assumes that DRE contains ONLY ASCII digits):
transform!(df, :DRE => ByRow(x -> "20"*x[2:3]) => :DRE)
or a general solution that correctly handles all UNICODE characters but is a bit verbose:
transform!(df, :DRE => ByRow(x -> "20"*last(first(x, 3), 2)) => :DRE)