Hybrid set of optimally trained feed-forward, Hop-field and Elman neural networks were used as computational tools and were applied to immunoinformatics. These neural networks enabled a better understanding of the functions and key components of the adaptive immune system. A functional block representation was also created in order to summarize the basic adaptive immune system and the appropriate neural networks were employed to solve them. Training and learning accuracy of all neural networks were very good. Polymorphism, inheritance and encapsulation (PIE) learning concepts were adopted in order to predict the static and temporal behavior of adaptive immune system interactions in response to typical virus attacks