Background: Identification of protein phenotypic annotations is an essential and challenging
problem in modern genetics. Such problem is related to some serious diseases, including
cancers, HIV and so on. The factors of genotype and environment increase the difficulties in determining
the phenotype of proteins. The experiment methods to achieve such a goal are always timeconsuming
Objective: The aim of this study was to design a quick and cheap method for determining the phenotypes
Method: In this study, we proposed a network computational method to identify novel phenotypic
annotations of proteins. To execute such method, a heterogeneous network was constructed, which
contained three sub-networks: protein network, phenotypic type network, and protein-phenotypic
type network. The method tried to find out all paths with limited length, which connected one protein
and one phenotypic type. A scoring scheme was adopted to count obtained paths and induced a
score to indicate the associations between them.
Results and Conclusion: The ROC and PR curve analyses were done to evaluate the performance
of the method, indicating the utility of the method. Our method was superior to other network
methods, which incorporated popular network algorithms.