A wide variety of practical problems related to the interaction of agents can be examined using biological metaphors. This paper applies the theory of G-networks to agent systems by considering a biological metaphor based on three types of entities: normal cells C, cancerous or bad cells B, and immune defense agents A which are used to destroy the bad cells B, but which sometimes have the effect of being able to destroy the good cells C as well (autoimmune response). Cells of type C can mutate into cells of Type B, and vice-versa. In the presence of probabilities of correct detection and false alarm on the part of agents of Type A, we examine how the dose of agent A will influence the desired outcome which is that most bad cells B are destroyed while the damage to cells C is limited to an acceptable level. In a second part of the paper we illustrate how a similar model can be used to represent a mixture of agents with the ability to cooperate as well as to compete.