MIMLRBF预测谷物蛋白质功能方法的改进

Improvement of MIMLRBF Algorithm for Predicting Function of Grain Proteins

摘要:

随着人们对营养与保健功能的关注，谷物蛋白质功能预测已经成为当前研究热点。面对大量已完成测序的谷物蛋白质基因组数据，利用计算方法来预测谷物蛋白质功能已经成为主流。从谷物蛋白质结构域序列出发，首次将MIMLRBF算法运用到蛋白质功能预测，并在此算法基础上，提出了多种改进后的谷物蛋白质功能预测模型。其中，针对平均Hausdorff距离削弱了两种蛋白质之间最短结构域距离所起作用的问题，为平均Hausdorff距离引入一个自动调节系数来计算蛋白质之间的相似性。同时，为提高MIMLRBF算法的预测效果，利用改进后的混合径向基核函数进行激活，得到了改进后的MIMLRBF算法模型。最终利用主流的评价标准对预测结果进行评价，可以发现改进后的MIMLRBF比传统的预测效果更好，证明了所建模型的优越性。

Abstract:

With people's attention to nutrition and health care functions, the prediction of grain protein function has become a research hotspot. Faced with large amounts of sequenced cereal protein genome data, the use of computational methods to predict grain protein function has become the mainstream. For the first time, the MIMLRBF algorithm was applied to protein function prediction from the grain protein domain sequence. Based on this algorithm, several improved grain protein function prediction models were proposed. Among them, an automatic adjustment coefficient for the average Hausdorff distance was introduced to calculate the similarity between proteins aiming to solve the problem that the average Hausdorff distance weakened the shortest domain distance between two proteins. At the same time, in order to improve the prediction effect, the improved MIMLRBF algorithm model was obtained using the improved hybrid radial basis kernel function to activate. Finally, the improved MIMLRBF was confirmed to be better than the traditional prediction based on the evaluation of prediction results using mainstream evaluation criteria, which proved the superiority of the proposed model.