Comparison of result differences in multiple implementations of a stochastic motif finding algorithm

The computational difficulty of DNA motif sequence finding algorithms is a well studied problem. One of the most promising directions in dealing with the challenges of a huge possible solution space are heuristic based methods. However, although there are various alternatives based on a stochastical model, there are few studies that look at differences across implementation methods. Generally recursive implementations are considered less desirable, due to the slower throughput rate and increased memory usage of recursion thus they are overlooked. In this article the goal is to illustrate differences in results of recursive versus iterative implementation of a stochastic, Monte-Carlo method based, nucleotide motif finding algorithm. To deal with the possible differences that may arise due to the internal construction of the pseudorandom function we have run the experiment on four implementations of the same algorithm, one iterative and one recursive on two programming languages. The experimental results presented in this article show that although both implementation tend to have the same central tendency for result value, recursive algorithms have better local optimum values and can potentially yield better results.