期刊名称：Lecture Notes in Engineering and Computer Science
印刷版ISSN：2078-0958
电子版ISSN：2078-0966
出版年度：2018
卷号：2233&2234
页码：57-62
出版社：Newswood and International Association of Engineers
摘要：Remote homology detection plays a pivotal role in
bioinformatics and can be used to detect functional and
structural relationships between proteins that have a low
sequence identity. While good discriminative methods for
remote homology detection have been developed recently, the
accurate representation of various protein features for
homology detection remains a challenge. A hybrid support
vector machine method (SVM-hybrid) for protein remote
homology detection that combines the support vector machine
auto-cross covariance (SVM-ACC) and support vector machine
physicochemical distance transformation (SVM-PDT) methods
was proposed. A distance transformation was used to extract
evolutionary and physicochemical data from protein sequences.
A mean receiver operating characteristic (ROC) of 0.959 was
achieved using the SCOP 1.53 benchmark datasets. A mean
accuracy of 95%, a specificity of 0.894, a sensitivity of 0.988 and
a Matthews correlation coefficient (MCC) score of a 0.887 were
obtained on opsin protein datasets. The SVM-hybrid method is
capable of remote homology detection and has the potential to
be used for further protein research.
关键词：protein remote homology; support vector
machine; protein family detection
