3.3 Other Methods Based on the fact that special disease has close association with
special tissue, Liu et al. [65] proposed a computational framework
to predict lncRNA-disease interactions by integrating lncRNA
expression profiles, gene expression profiles, and human disease-
associated gene data. The lncRNA expression profiles were used to
divide lncRNA into tissue-specific lncRNAs and non-tissue-specific
lncRNAs. For tissue-specific lncRNA, they considered this lncRNA
might be associated with diseases that are related to this specific
tissue. For non-tissue-specific lncRNA, the Spearman rank correla-
tion coefficients between the integral expression level of this
lncRNA and the corresponding genes in the common human tissue
or cell types were calculated. Then, a certain cutoff score was used
to select a set of genes co-expressed with this lncRNA. Finally, the
hypergeometric distribution test was employed to find significantly
enriched diseases in the corresponding co-expressed gene set for
each lncRNA.
In addition, statistics methods are also used to predict lncRNA-
disease interactions. Chen et al. [66] proposed a novel model,
lncRNA-disease association inference (HGLDA), to predict
lncRNA-disease associations based on hypergeometric distribution.
The lncRNA functional similarity was calculated by integrating
disease semantic similarity, miRNA-disease associations, and
miRNA-lncRNA interactions. The hypergeometric distribution
test was employed for each lncRNA-disease pair by examining
whether this lncRNA and disease significantly shared common
miRNAs which can interact with both of them. Li et al. [67]
developed a computational method to identify the relationship
between lncRNAs and gastric cancers. They constructed lncRNA
and protein-coding gene co-expressed network by integrating
Spearman correlation coefficient and Pearson correlation coeffi-
cient between the expression profiles of lncRNAs and coding
genes. In addition, they utilized Wilcoxon signed-rank test to
detect differentially expressed genes and lncRNAs in gastric cancer
versus normal tissue. Finally, the hypergeometric distribution was
used to find statistical significance of the differentially expressed
genes among the coding gene set co-expressed with each differen-
tially expressed lncRNA.
Based on assumption that functional similar lncRNAs tend to
compose module in network, some researchers try to mine similar
module of network to identify disease-related lncRNAs. Cogill et al.
[68] proposed a computational method for lncRNA-cancer associ-
ation prediction based on gene expression profiles and lncRNA
expression profiles. In the first step, they constructed a gene
co-expression network by computing Pearson product-moment
correlation. Then, the co-expression modules were discovered
and analyzed by using 1-Topological Overlap Matrix (TOM) and
Database for Annotation, Visualization and Integrated Discovery
(DAVID). Finally, 37 co-expression modules were found, and two
216 Wei Lan et al.