In this paper, we present a concentration control method that may become a new framework of DNA computing. In this method, the concentration of each DNA is used as input and output data. By encoding the numeric data into concentrations of DNAs, a shortest path problem, which is a combinatorial optimization problem, can be solved. The method also enables local search among all candidate solutions instead of a exhaustive search. Furthermore, we can reduce the costs of some experimental operations in detecting process of DNA computing, because we have only to extract and analyze relatively intensive bands. Solutions of a shortest path problem by using a simulator and by laboratory experiments are presented to show the effectiveness of the concentration control method.