Applied Research Lab

Herbert Tsang


Project abstractApplied Research Lab (ARL) is dedicated to a board program of basic and applied research in the areas relates to engineering and computing science. The laboratory's mission is to advance the state-of-the-art applied research and develop technologies that benefit the world.In carrying out the above-mentioned objectives, ARL has a strong emphasis on theoretical modeling, simulation, experimental work, data analysis, and prototyping. We have expertise in the following areas: computational intelligence, bioinformatics, visualization, and immersive environment.Project descriptionRNA function is determined largely by its three-dimensional structure. Since X-ray diffraction and Nuclear Magnetic Resonance are often too costly to determine the 3D form of a long RNA from a single RNA sequence, computational methods offer a more practical alternative, with the ability to predict RNA secondary structure from the base sequence at lower cost. My team has investigated the use of Evolutionary Computation (EC) in an ensemble-based hybrid algorithm by combining a permutation-based stochastic optimization algorithm and machine learning techniques to RNA secondary structure prediction (SARNA-Ensemble-Predict).We are proposing a research project with focus on developing new algorithms based on evolutionary computation techniques for predicting RNA secondary structures. There are two specific goals for the student researcher: a) incorporate swarm intelligence into SARNA-Ensemble-Predict and b) evaluate and make improvement to the structure compare algorithm, RNADPCompare. The current proposal is based on the recent success in SARNA-Ensemble-Predict and extending it to include swarm intelligence, another EC technique. In addition, with ensemble-based approach we have found that there is a need of a better metric in structure comparison for RNA structure. Subsequently, we have invented an algorithm for structural comparison based on image processing technique, RNADPCompare, which shown promising results. There is increasing evidence that there is no single prediction algorithm that can consistently outperform all others. By exploiting the complementary strengths of multiple approaches, we can create a generic and powerful algorithm that outperforms existing secondary structure prediction algorithms in terms of prediction accuracies.