The computer simulation of tumor – host ecosystems interacting with an adaptive immune system may serve as a tool for anti-cancer treatment optimization, but requires appropriate mathematical models. Regarding the tasks of the adaptive immune system (antigen pattern recognition and classification), a perceptron can be used as a conceptual structure representing corresponding biological structures for antigen pattern recognition and classification such as Antigen Presenting Cells (APC’s) and their interaction with effector cells in lymph nodes. Regarding the topology of the lymph vessel network, the adaptive immune system may be represented by several perceptrons receiving information about antigen patterns from different tissue compartments. In this study, two scenarios of lymph node arrangement have been investigated. In both scenarios, a tumor-host tissue compartment is treated with ionizing radiation and a second compartment with host tissue and a tumor metastasis is not irradiated. The results exhibit a dependence of the immune response onto the lymph node arrangement, indicating that the topology of the lymph node network is important for an optimal adaptive immune response. The presented simplistic model structure does not allow for a perfect classification between tumor and host tissue. Instead of a single perceptron which is related to the interaction of immune cells in a corresponding lymph node as suggested in this study, networks of locally interacting units may be considered as layers building a deep (convolutional) neural network - like structure.