Gene Prediction by Markov Models & Complementary Methods

描述（由申请人提供）：本项目的目标是为真核基因发现带来一种自我训练的范例，这种范例是我们小组以前成功开发的，用于原核基因发现算法GeneMark和GeneMark.hmm。考虑到进行大规模测序工作的真核生物基因组的数量不断增加，这一目标具有最重要的实际意义。更好地了解这些生物体（包括人类）的基因组生物学，对于研究、控制和预防人类疾病至关重要。我们计划开发一个从头算算法的基因识别匿名真核生物基因组的基因模型建立的无监督学习过程。此外，我们计划开发一个扩展，该算法将现有的外部信息（cDNA，蛋白质序列数据）整合到基因识别和自我训练程序。最后，我们计划开发新的算法工具，从现有的序列数据产生额外的外在信息。这些工具将有助于自我训练的基因发现算法以及新预测的蛋白质的表征。

DESCRIPTION (provided by applicant): The goal of this project is bring to eukaryotic gene finding a paradigm of self-training successfully developed previously by our group for prokaryotic gene finding algorithms GeneMark and GeneMark.hmm. This objective is of the outmost practical importance given the growing number of genomes of eukaryotic organisms that undergo large scale sequencing efforts. Better understanding of genome based biology of these organisms, including human, has critical importance for studying, controlling and preventing human diseases. We plan to develop an ab initio algorithm for gene identification for anonymous eukaryotic genomes with gene models built by an unsupervised learning procedure. Also we plan to develop an extension of this algorithm integrating existing extrinsic information (cDNA, protein sequence data) into the gene identification and self-training procedures. Finally we plan to develop novel algorithmic tools for generating additional extrinsic information from available sequence data. These tools will be useful for both the self-trained gene finding algorithms as well as for the characterization of newly predicted proteins.