In the last few years Graph Neural Networks (GNNs), a family of neural architectures used for irregularly structured data, gained a rising interest of the graph processing research community. Previous work has shown the potential of GNNs in simulations. In the field of bioinformatics the simulation of human body movements is of great interest. State of the art approaches used for simulations, like the Finite Element Method (FEM), are very time and resource consuming. The human body, in particular the muscle system, can be represented in a graph structure. This offers a great potential of applying Deep Learning approaches for graph structures on muscle data to optimize the simulation. This work presents a proof of concept to apply GNNs on muscle data and find effective network properties. A Deep Learning Feed Forward Neural Network combined with a Graph Convolutional Network (GCN) is used to learn the deformations of the muscles. The model can predict every node inside the muscle. The results are compared to an existing approach which is using a Feed Forward Neural Network to predict a fixed set of nodes inside a muscle.