Data of experiment
|
In this study, we used the datasets that were used previously by Singh-Blom et al(2013).
PPI network: either the HumanNet(Mostafavi S and Morris Q, 2010) or HPRD network. HumanNet contains 733,836 edges between genes with non-zero weights integrated from 21 different data sources. HPRD is an unweighted and much sparser PPI network that contains only 56,661 associations. Gene-phenotype associations(eight non-human species): plants,worms,fruit flies, mice, yeast, Escherichia coli, zebrafish, and chickens. phenotype similarities for human were derived solely from Van Driel MA (2006). |
Methods of experiment
|
Here we propose two novel multipath methods, HeteSim_MultiPath (HSMP) and HeteSim_SVM (HSSVM), based
on the HeteSim measure.
HSMP uses the HeteSim measure to calculate the similarity between nodes in heterogeneous networks. Then, the HeteSim scores of different paths are combined with a constant that dampens contributions from longer paths. HSSVM also uses the HeteSim measure to calculate similarity. However, HSSVM uses a machine learning method instead of a constant to combine HeteSim scores. |
All datasets and methods used in this paper can be download here.