Heuristic Algorithm for Pseudoknotted RNA Structure Prediction

Pseudoknotted RNA structure prediction is an important problem in computational biology. Existing polynomial time algorithms can handle only limited types of pseudoknots, or have too high time or space to predict long sequences. In this paper a heuristic algorithm is presented to maximize stems and predict arbitrary pseudoknots with O(n³) time and O(n) space for a large scale of 5000 bases. Compared with maximum weighted matching algorithm, our algorithm reduce space complexity from O(n²) to O(n); and the experimental results show that its sensitivity is improved form 80% to 87.8%, and specificity is increased from 53.7% to 75.9%. Compared with genetic algorithm with the accuracy of 83.3% and simulated annealing algorithm with the accuracy of 79.7%, our algorithm increases the predicted accuracy to 87.5%.