Most DNA computing algorithms for mathematical problems start with combinatorial generation of an initial pool. Several methods for initialpool generation have been proposed, including hybridization/ligation and mix/split methods. Here, we implement and compare parallel overlap assembly with the hybridization/ligation method. We applied these methods to the molecular algorithm to solve an instance of the graph problem with weighted edges. Our experimental results show that parallel overlap assembly is a better choice in terms of generation speed and material consumption than the hybridization/ligation method. Simulation of parallel overlap assembly was performed to investigate the potential and the limitation of the method.