Deoxyribonucleoside kinases (dNKs) carry out the rate-determining step in the nucleoside salvage pathway within all domains of life where the pathway is present, and, hence, are an indication on whether or not a species/genus retains the ability to salvage deoxyribonucleosides. Here, a phylogenetic tree is constructed for the thymidine kinase 2-like dNK gene family in metazoa. Each enzyme class (deoxycytidine, deoxyguanosine, and deoxythymidine kinases, as well as the multisubstrate dNKs) falls into a monophyletic clade. However, in vertebrates, dCK contains an apparent duplication with one paralog lost in mammals, and a number of crustacean genomes (like Caligus rogercresseyi and Lepeophtheirus salmonis) unexpectedly contain not only the multisubstrate dNKs, related to Drosophila multisubstrate dNK, but also a TK2-like kinase. Additionally, crustaceans (Daphnia, Caligus, and Lepeophtheirus) and some insects (Tribolium, Danaus, Pediculus, and Acyrthosiphon) contain several multisubstrate dNK-like enzymes which group paraphyletically within the arthropod clade. This might suggest that the multisubstrate dNKs underwent multiple rounds of duplications with differential retention of duplicate copies between insect families and more complete retention within some crustaceans and insects. Genomes of several basal animalia contain more than one dNK-like sequence, some of which group outside the remaining eukaryotes (both plants and animals) and/or with bacterial dNKs. Within the vertebrates, the mammalian genomes do not contain the second dCK, while birds, fish, and amphibians do retain it. Phasianidae (chicken and turkey) have lost dGK, while it has been retained in other bird lineages, like zebra finch. Reconstruction of the ancestral sequence between the multisubstrate arthropod dNKs and the TK2 clade of vertebrates followed by homology modeling and discrete molecular dynamics calculations on this sequence were performed to examine the evolutionary path which led to the two different enzyme classes. The structural models showed that the carboxyl terminus of the ancestral sequence is more helical than dNK, in common with TK2, although any implications of this for enzyme specificity will require biochemical validation. Finally, rate-shift and conservation-shift analysis between clades with different specificities uncovered candidate residues outside the active site pocket which may have contributed to differentiation in substrate specificity between enzyme clades.