Malaria is one of the leading causes of death by infectious disease worldwide. The widespread of resistance to the current antimalarial drugs makes urgent the search and discovery of new targets and new drugs. A potential target for the development of new antimalarial drugs is deoxyuridine triphosphatase (dUTPase), and it has been validated for Escherichia coli, Saccharomyces cerevisiae and Mycobacterium smegmatis. This enzyme plays an important role in maintaining the balance between 2'-deoxyuridine 5'-triphosphate (dUTP) and 2'- deoxythymidine 5'-triphosphate (dTTP), in order to avoid the erroneous incorporation uracil in the DNA tape. In this study, we developed robust conformation-independent fragment-based quantitative structure–activity (QSAR) and structure–selectivity relationship (QSSR) models for a series of β-branched acyclic nucleotides inhibitors of Plasmodium and human dUTPase, aiming to design new anti-malarial agents. The Hologram QSAR and QSSR models generated showed good robustness and external predictability, and is capable of predict affinity and selectivity of untested compounds inside the applicability domain. Therefore, the generated models can be used in virtual screening campaigns in the search of new potent and selective PfdUTPase inhibitors.