Dynamic matching algorithm for viral structure prediction.

Most viruses have RNA genomes, their biological functions are expressed more by folded architecture than by sequence. Among the various RNA structures, pseudoknots are the most typical. In general, RNA secondary structures prediction doesn't contain pseudoknots because of its difficulty in modeling. Here we present an algorithm of dynamic matching to predict RNA secondary structures with pseudoknots by combining the merits of comparative and thermodynamic approaches. We have tested and verified our algorithm on some viral RNA. Comparisons show that our algorithm and loop matching method has similar accuracy and time complexity, and are more sensitive than the maximum weighted matching method and Rivas algorithm. Among the four methods, our algorithm has the best prediction specificity. The results show that our algorithm is more reliable and efficient than the other methods.

Characteristics of equipartition for RNA structure.

BACKGROUND: With the continuous discovery of novel RNA molecules with key cellular functions and of novel pathways and interaction networks, the need for structural information of RNA is still increasing. In order to predict structure of long RNA and understand its natural folding mechanism, exploring the characteristic of RNA structure is an important issue. METHODS: The real RNA secondary structures of all 480 sequences from the database of RNA strand, validated by nuclear magnetic resonance or x-ray are selected. For one sequence with multiple domains, the length ratios of these domains to the sequence are computed. For one sequence with one domain and multiple sub-domains, the length ratios of these sub-domains to the domain are computed. Then the ratios are compared and analyzed to seek the partition characteristic of domains and subdomains. RESULTS: For most RNAs, the length ratios of multiple domains to its sequence are close to equal, and those of sub-domains to its domain are also nearly identical. Most RNAs with multiple domains have two domains, so the length ratios of the domains to its sequence are close to 0.5. For sequence with one domain and no sub-domain or one sub-domain, the centre of domain and sub-domain is close to that of the sequence. CONCLUSIONS: A novel finding is given that RNA folding accords with the characteristic of equipartition based on statistical analysis. The characteristic reflects the folding rules of RNA from a new angle, which maybe more close to natural folding.

Characteristics and prediction of RNA structure.

RNA secondary structures with pseudoknots are often predicted by minimizing free energy, which is NP-hard. Most RNAs fold during transcription from DNA into RNA through a hierarchical pathway wherein secondary structures form prior to tertiary structures. Real RNA secondary structures often have local instead of global optimization because of kinetic reasons. The performance of RNA structure prediction may be improved by considering dynamic and hierarchical folding mechanisms. This study is a novel report on RNA folding that accords with the golden mean characteristic based on the statistical analysis of the real RNA secondary structures of all 480 sequences from RNA STRAND, which are validated by NMR or X-ray. The length ratios of domains in these sequences are approximately 0.382L, 0.5L, 0.618L, and L, where L is the sequence length. These points are just the important golden sections of sequence. With this characteristic, an algorithm is designed to predict RNA hierarchical structures and simulate RNA folding by dynamically folding RNA structures according to the above golden section points. The sensitivity and number of predicted pseudoknots of our algorithm are better than those of the Mfold, HotKnots, McQfold, ProbKnot, and Lhw-Zhu algorithms. Experimental results reflect the folding rules of RNA from a new angle that is close to natural folding.