The blood glucose system is one of the most important systems of the human body, since energy transport is fulfilled through this complex endocrine control process. Because of its great importance (in case of failure, its primary consequence diabetes mellitus appears), many models were published, most of them with phenomenological approach. The current paper focuses on a new molecular model published recently. By modifying it to model the failure of pancreatic insulin secretion that characterizes Type 1 diabetes mellitus, soft computing based controller is designed. Rough rule base is generated with subtractive clustering which is followed by its refinement by parameter tuning. As a result, an Adaptive Neuro-Fuzzy Inference System (ANFIS) is created. Simulation results are in accordance with the behavior of the healthy human blood glucose system.