It is now widely recognized that extracting knowledge from large-scale data sets requires mathematical modeling of the underlying processes. Traditional modeling methods in biochemistry and cellular biology are bottom-up approaches where models are constructed by piecing together information about individual components (usually obtained in vitro). The systems biology approach, however, lends itself naturally to top-down approaches where models are built directly from data obtained from intact cells. We propose two independent methods to build models directly from data, one continuous and based on calculus, and the other discrete and based on computational algebra. Each of these methods has strong and weak points but these are largely non-overlapping, making the idea of combining them very attractive. Here I will briefly describe the two methods and the ongoing efforts to make their application robust.