原核基因识别中的一种负样本生成算法

在基因识别的两类算法中,判别算法通常需要正负两类样本来训练参数.在原核生物的基因组中,由于可充当负样本的基因间序列太少,如何产生负样本便成为原核基因识别中的一个问题.本文提供了一种基于"自相似映射"的负样本生成算法,与通常使用的随机生成算法不同,该算法不需要生成随机数.本文给出了两种负样本生成算法的比较,并初步讨论了自相似性对于DNA序列分析的意义.

A self-similarity-map-based algorithm for generating negative samples and its application in prokaryotic gene recognition

In the two types of gene recognition algorithms discriminant and clustered, the discriminant algorithms usually need two groups of samples as training parameters. For the prokaryotic genomes, few intergenic sequence acts as negative samples. Therefore,an unsolved problem is how to generate negative samples for prokaryotic gene recognition. In the paper, a simple algorithm for generating negative samples is presented based on the self-similar map. In contrast to widely used randomness-based algorithms,our algorithm doesn't need random numbers. A comparison between self-similarity-based and randomness-based algorithms is given , and the significance and utility of self-similarity in the analysis of DNA sequence is also disscussed.