真核生物多基因预测结果整合算法

Algorithm for Integration of Multiple Eukaryotic Gene Prediction Results

针对独立基因预测算法可靠性较差的缺点,提出了真核生物多基因预测结果整合算法(Algorithm for integration of multiple eukaryotic gene prediction results,AIMEGPR)。该算法在综合分析各种预测算法结果的基础上,首先用极大似然法估计各种预测算法的性能参数,然后利用这些性能参数计算基因证据区间上各个基因片段归属于各种基因元件类型的后验概率,最后采用动态规划法在基因证据区间上确定最优的一致基因结构。AIMEGPR既不需要人工定制整合规则,也不需要复杂的训练学习,因此AIMEGPR尤其在基因组新测序物种上进行编码基因注释时具有十分显著的优越性。实验结果表明,利用AIMEGPR算法对多基因预测结果整合可以明显提高一致基因的可靠性。

Considering the disadvantage of poor reliability of the single prediction algorithm,the paper presents an algorithm for integration of multiple eukaryotic gene prediction results （AIMEGPR）.AIMEGPR took account of the results of multiple prediction software and used the maximum likelihood method to estimate the performance parameters of various prediction algorithms.Then AIMEGPR used these performance parameters to calculate the posterior probability which class the gene segment in gene evidence region belongs to.Finally,AIMEGPR determined an optimal consensus gene structure for gene evidence region using dynamic programming method.AIMEGPR does not require integration of the rules offered by experts,and also does not require complex training study.Therefore,AIMEGPR has a significant superiority on annotation for the new sequencing genome.The experimental results show that AIMEGPR can improve the reliability of consensus genes.