基因位点预测的一种特征选择优化算法

目的剪接位点是真核细胞生物基因序列中外显子和内含子的相邻区域,如果能准确预测基因序列中的剪接位点,就能将基因中的表达区域和非表达区域分开.方法从机器学习的角度出发,提出了一种有效的特征选择算法用于剪接位点的建模和预测.该算法首先将初始链模型中每一对父子节点作为特征量提取,然后通过遗传算法和最大后验分类器进行特征选择.结果及结论对剪接位点数据的预测结果显示,这种新算法能够有效地优化链模型的结构,提高对剪接位点的预测能力.同时,经过优化的模型也有助于了解真核细胞中基因转录和表达的过程.

An Effective Feature Selection Optimization Algorithm For Gene Splice Junction Sites Prediction

Objective Splice junction sites are the boundaries between exons and introns in eukaryotic gene sequences. If the splice junction sites in a gene sequence could be predict correctly, the coding regions in a gene would be separated from the non-coding regions. Method This paper proposes a machine learning approach for the modeling and prediction of splice junction sites by an effective feature selection algorithm. This algorithm selects every pair of parent and child node in original chain models for splice junction sites as features, and then uses genetic algorithm and a MAP(Maximum A Posteriori) classifier to select the features. Results and Conclusion The experiment result shows that the new algorithm can optimize the chain models and improve the prediction accuracy of splice junction sites. Besides the architecture of the optimized chain models can also help to understand the procedures of gene translation and expression in eukaryotic cells.