• adenosine receptor antagonists
• model validation
• QSAR
• training set selection

Over the last years, an increasing number of applications of QSAR (Quantitative Structure-Activity Relationships) appeared in the literature, not only in lead finding and lead optimization, but also in other fields related to drug discovery, such as ADMET predictions. Neverthless, many questions are still open and they supplied the starting point of this research. Attention was focused on characteristics usually intended as “pitfalls” of QSAR itself. In this work, each step of the QSAR model development process was handled with a rational and rigorous approach, and the classic QSAR strategies were implemented with new protocols