Gene expression data in biology is becoming important as the amount and quality of the data rapidly increases. However, the amount generated can be daunting and its direct interpretation is often difficult. The interaction of the genes and the number involved can be large. Is there a dynamical system at play? This paper discusses modeling gene expression data as a computational reconstruction of a dynamical system. The problem is a classic inverse problem – given the data what is the model? A phenomenological model based on a extension of generalized Lotka-Volterra models is developed. One advantage of these models is that they are readily amenable to biological interpretation. The reconstruction is ill-posed and subject to numerical instability problems when is there not enough data of sufficient quality. We will discuss how these problems can affect the results and how we might overcome them. Lastly, we will present some preliminary results and some applications of the reconstructions