PDE4 inhibitors have been largely studied because of their promising therapeutic effects concerning inflammation and neurodegenerative dysfunctions, such as depression, schizophrenia and Alzheimer's diseases. In this context, the PDE4B isoform proved to be particularly involved in the activation of inflammatory responses, while the PDE4D subfamily is more associated with neuropathologies. The clinical use of PDE4 inhibitors was restricted by the presence of prominent side effects probably due to their non-specific action across the different isoforms. Therefore, this work deals with the development of 3D-QSAR models, supported by molecular docking studies, to identify the key requirements underlying selective PDE4B or PDE4D inhibition. The results highlighted the ligand-based approach as a promising tool to guide the rational design of novel PDE4 inhibitors endowed with high affinity and selectivity profiles. The alignment of compound 1-85 and the model A statistical results are depicted.