CAJ | 학술논문

利用机器学习方法对单个氨基酸突变引起的蛋白质稳定性变化进行精确地预测,对蛋白质的结构和功能方面的研究具有重要的价值,并且对设计新的蛋白质及蛋白质工程学具有一定的指导意义.通过对蛋白质网络拓扑特征的研究,发现网络拓扑特征对于蛋白质突变稳定性影响具有较高的准确率.基于蛋白质网络拓扑特征的随机森林算法,能较好的对蛋白质单点突变所造成的稳定性改变进行预测,总准确率达到86%,MCC值达到0.67,并优于文献报道的预测结果.
이용궤기학습방법대단개안기산돌변인기적단백질은정성변화진행정학지예측,대단백질적결구화공능방면적연구구유중요적개치,병차대설계신적단백질급단백질공정학구유일정적지도의의.통과대단백질망락탁복특정적연구,발현망락탁복특정대우단백질돌변은정성영향구유교고적준학솔.기우단백질망락탁복특정적수궤삼림산법,능교호적대단백질단점돌변소조성적은정성개변진행예측,총준학솔체도86%,MCC치체도0.67,병우우문헌보도적예측결과.
Protein stability changes by single amino acid substitutions are important for the understanding of the relationship between protein structure and function.Previous prediction models were constructed using protein sequence and structural characteristics to predict the change of free energy stability on mutant.Such models were also valuable for designing and engineering new proteins.In this study,we presented a random forest algorithm combined with protein network topology properties to predict the change of free energy stability caused by the single point mutation.Amino acid residues around a mutation were also applied to characterize its environment.This method achieved total prediction accuracy(ACC) of 0.86 and Matthew's correlation coefficient(MCC) of 0.67,which are slightly higher to those reported previously.