The structural requirements of pyrrolo[2,3-d]pyrimidine nucleoside (PPN) analogues as adenosine kinase (AK) inhibitors were in silico studied by using CoMSIA method. All models were trained with 32 compounds, after which they were evaluated for predictive ability with additional 5 compounds. Quantitative information on structure-activity trends is provided for further rational development and direction of selective synthesis. The best CoMSIA model included hydrophobic, H-bond donor and H-bond acceptor fields and had a good predictive quality according to internal validation criteria. In addition, this model predicted adequately the compounds contained in the test set. The analysis of the model gives a comprehensive qualitative and quantitative description of the molecular features at C4 and C5 positions of the pyrrolo[2,3-d]pyrimidine scaffold and C5-position of the beta-d-ribofuranose of PPN analogues, relevant for a high AK inhibitory activity.