摘要
对大鼠的分子调控网络进行了系统化的建模分析和预测。选取了7类具有显著生物学含义的数据类型(证据),采用数据整合的方法对蛋白质-蛋白质相互作用(PPI)和蛋白质-DNA相互作用加以系统性分析,并利用支持向量机(SVM)预测大鼠全基因组的PPI和PDI。通过对阈值的设定,有效的控制假阳性率,预测结果与权威数据库BOND和MINT中的记录部分重合。实验结果表明数据整合方法可以弥补单一证据的局限性,提高预测的准确性,为后续的实验验证提供理论支持。
An integration methodology to the analysis and prediction of rat＇s molecular regulatory networks was introduced. 7 kinds of data types which are called evidences with significant biological senses were selected these heterogeneous data were integrated to systematically analyze protein-protein interactions （PPI） and protein-DNA interactions（PDl）. Support vector machines were used to predict all the possible PPIs and PDIs in rat＇s genome and the contribution of each evidence was calculated. The false positive rate was effectively controlled by setting proper threshold, and the predictions were partially overlapped with the records in BOND and MINT. The result shows that data integration makes up the drawbacks of using one data type alone, improves the accuracy of prediction and serves the downstream experimental verifications.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2009年第5期1479-1483,1494,共6页
Journal of System Simulation
