Global optimization analysis of the 38-atom sequence of CunAum (n+m=38) nanoalloy clusters was performed by means of a genetic algorithm and using the Gupta empirical potential. A subset of these global minimum structures was recalculated with more detail using density-functional theory (DFT) techniques, namely the sequence CunAu38-n with n=6,13,15,18,37. The density-functional theory analysis confirms some general qualitative trends observed in the cluster sequence using the genetic algorithm, namely: (a) the tendency for gold atoms to migrate to the outer shells of the cluster leaving a backbone of Cu atoms, and (b) rapid structural changes as pure Cu clusters are doped with increasing number of Au atoms (ranging from slightly distorted octahedron shapes for n=37,18 to oblate spheroids for the rest of the sequence). On the other hand, DFT analysis show that the investigated structures are more expanded and that their binding energies are higher than those obtained by the empirical method. Additionally, the separation between the outer shell of Au atoms and the inner core of Cu atoms is larger in the final relaxed structures. It seems therefore that in general the empirical Gupta potential has underestimated the internal stress when gold atoms are initially inserted inside the cluster. This cast some doubt about the parametrization of the Gupta potential that has been used for this kind of study. The cluster Cu3Au22 has also been investigated by means of DFT in order to clarify the different geometrical structures that were obtained empirically in previous studies. However, it revealed inconclusive for both the geometrical shape and binding energy of the global minimum structure.