Finding Genes in DNA using Machine Learning Algorithms

自动高效地识别DNA序列中的基因是目前计算分子生物学中的一个核心问题。本研究项目利用基因序列数据库中的基因识别问题作为机器学习研究的驱动力。它建立在几个有希望的最新结果的基础上，研究了决策树方法、基于记忆的推理和隐马尔可夫模型。这些方法每一种都有其独特的优点。PI最近的研究发现，决策树可以实现比任何以前在非常短的DNA子序列中寻找蛋白质编码区的方法都更高的准确性。将决策树方法嵌入到动态规划算法中可以用来产生更通用的系统。被发现最有效的技术将被整合到一个完整的基因发现系统中，该系统可以高精度地识别基因，并增加了一个用户友好的界面，并向生物学和计算机科学研究界免费提供。

The automatic and efficient recognition of genes in DNA sequences is now widely recognized as a central problem in computational molecular biology. This research project uses the gene recognition problem in databases of gene sequence data as a driver for research in machine learning. It builds on several promising recent results, investigating decision tree methods, memory-based reasoning, and hidden Markov models. Each of these approaches has particular strengths. Recent research by the PI has found that decision trees can achieve consistently higher accuracy that any previous methods for finding protein coding regions in very short DNA subsequences. Embedding of the decision tree approach in a dynamic programming algorithm can be used to produce a more general system. The techniques discovered to be most effective will be incorporated into a complete gene finding system that can identify genes with high accuracy, augmented with a user-friendly interface, and made freely available to the biology and computer science research communities.