摘要
LncRNA-疾病关联预测的计算方法是解决传统生物学实验昂贵且费时的有效途径,其中基于机器学习的计算方法是当前研究热点,但其存在着未充分考虑lncRNA-疾病关联矩阵的局部结构和全局结构的问题.因此,本文提出了一种lncRNA与疾病潜在关联的多层线性投影预测方法(MLPLDA:Multi-layer linear projection for predicting lncRNA-disease association).MLPLDA利用组合加权整合lncRNA和疾病的两种相似性,然后用WKNKN重构原始的lncRNA-疾病关联矩阵,最后使用堆叠层策略的多层线性投影进行lncRNA-疾病关联预测.在留一和五折交叉验证实验中,MLPLDA获得的AUC分别是0.8807和0.8563±0.0045,体现了其可靠的性能.在3种疾病(肺癌,乳腺癌和骨肉瘤)的案例研究中,MLPLDA能够有效预测与3种疾病有关系的lncRNA.
The calculation method of lncRNA-disease association prediction is an effective way to solve traditional biological experiments,which is expensive and time-consuming.Among them,the calculation method based on machine learning is the current research hotspot,but its existence does not fully consider the local structure and the global structure of lncRNA-disease association matrix.Therefore,this paper proposes a multi-layer linear projection for predicting lncRNA-disease association(MLPLDA:Multi-layer linear projection for predicting lncRNA-disease association).MLPLDA uses combination weighting to integrate two similarities between lncRNA and disease,and then uses WKNKN Reconstruct the original lncRNA-disease association matrix,and finally use the multi-layer linear projection of the stacked layer strategy to predict the lncRNA-disease association.In the leave-one-out cross validation and 5-fold cross validation,the AUC obtained by MLPLDA is 0.8807 and 0.8563±0.0045,respectively,reflecting its reliable performance.In the case studies of three diseases(lung cancer,breast cancer and osteosarcoma),it can effectively predict the lncRNAs related to the three diseases.
作者
谢国波
韩玉琼
林志毅
XIE Guo-bo;HAN Yu-qiong;LIN Zhi-yi(School of Computer Science,Guangdong University of Technology,Guangzhou 510006,China)
出处
《小型微型计算机系统》
CSCD
北大核心
2022年第10期2084-2089,共6页
Journal of Chinese Computer Systems
基金
国家自然科学基金项目(618002072,62002070)资助
广东省自然科学基金项目(2018A030313389)资助
广东省科技计划项目(2018B030323026)资助
广州市科技计划项目(201902020012,201907010021)资助.
