Evolutionary computation and applications in bioinformatics

本研究计划的目标是：1)开发改进的组合算法来解决生物信息学中的问题，特别是生物分子的结构预测，以我实验室迄今为止开发的算法为基础。在我的实验室开发的基于进化计算的RnaPredict算法已被证明能够以非常高的精度（高达90%）预测某些已知的Rna二级结构，并且在某些短和中等长度序列上具有比mfold更高的预测精度，mfold是目前从头开始Rna二级结构预测的基准。特别是，rnappredict不像mfold那样容易预测误报。rnappredict使用许多高级操作符和表示。我计划进一步评估、改进和扩展rnappredict。我已经开发了一个RNA二级结构的可视化工具，它比现有的工具有很多好处，包括：广泛支持输入和输出格式，能够生成高质量的图形进行复制，交互式输出以供最终用户进一步操作，能够在经典结构图中显示伪结，以及能够覆盖两种不同的结构进行比较。jViz。Rna还可以在两种结构之间进行定量和定性比较。定量比较目前主要考虑敏感性、特异性和f值。我打算扩展jViz。利用对偶图结合基于共同子结构的结构比较。

The objectives of this research program are:1) To develop improved combinatorial algorithms for problems in Bioinformatics, particularly for structure prediction of biomolecules, building on the algorithms that have been developed in my lab so far2) To extend jViz.Rna, a visualization tool for RNA secondary structureThe algorithm RnaPredict, developed in my lab and based on evolutionary computation, has been demonstrated to predict certain known RNA secondary structures with very high accuracy (up to 90%) and to have a higher prediction accuracy on certain short and medium length sequences than mfold, the current benchmark for ab initio RNA secondary structure prediction. Particularly, RnaPredict is not as prone to predicting false positives as mfold is. Rnapredict uses a number of advanced operators and representations. I plan to further evaluate, improve, and extend RnaPredict. I have developed a visualization tool for RNA secondary structure which has a number of benefits over existing tools, including: wide support of input and output formats, ability to produce high quality graphics for reproduction, interactive output for further manipulation by the end user, ability to display pseudoknots in a classical structure plot, and the ability to overlay two different structures for comparison. jViz.Rna also produces quantitative and qualitative comparisons between two structures. The quantitative comparison is currently taking into account sensitivity, specificity, and F-measure. I plan to extend jViz.RNA to incorporate a comparison of structures based on common substructures using dual graphs.