Page 73 - GPD-1-2

Gene & Protein in Disease Recent advances and challenges of network biology

phenotypes and genotypes. A tool called CIPHER, which In addition to the above works, based on the guilt-
integrates human PPIs, disease phenotype similarities, by-rewiring principle, many other methods have been
and known gene-phenotype associations, was developed developed to explore crucial genes/proteins in biological
to predict and prioritize disease genes; CIPHER has been systems, such as those based on differential co-expression
shown to be applicable to genetically uncharacterized analysis (Figure 3B) [49,103-111] or hidden Markov random
phenotypes and effective in genome-wide scans of field (HMRF) models . For instance, dynamical network
[52]
[49]
[73]
disease genes. In 2016, based on BioGRID , HPRD , biomarker (DNB) and single sample DNB [109] have been
[67]
and literatures [118] , we constructed a large-scale human developed to detect critical gene/protein sets that dominant
PPI network and explored the topological features of the development of complex diseases. The basic idea of
essential genes, viable genes, conserved genes, disease DNBs is hinged on the guilt-by-rewiring principle, where
genes, house-keeping genes, and tissue-specific genes . genes/proteins that extensively altered their co-expression
[13]
It was found that the lethal, conserved, house-keeping, patterns between treated and control samples are
and tissue-specific genes had hallmark graphical features. closely related to the causal phenotypes (Figure 3B).
With regard to degree, k-shell, eigenvector centrality, By incorporating differential rewiring signals from the
closeness, and other centrality measures, essential genes genome-wide association study (GWAS) and the HMRF
can be distinguished from viable ones with an accuracy model, Hou et al. developed a method to prioritize disease
as high as approximately 70%. Closeness, semi-local, and genes . We have recently proposed a new framework
[52]
eigenvector centralities can distinguish house-keeping to construct a gene differential co-expression network
genes from tissue-specific ones with an accuracy of about (GDCN) based on several RNA-Seq data [107] . Further basing
82%. Based on topological properties of disease genes in on the topological structures of the constructed GDCN,
the PPI network, an improved random forest classifier has three measures have been designed to explore important
been proposed to detect disease-related genes [119] . genes that are closely related to phenotypic changes
between treatments and controls. The proposed GDCN-
Besides the traditional centrality measures for complex based approach, which integrates the guilt-by-association
networks, some motif-based methods have also been and guilt-by-rewiring principles, provides alternative tools
developed. It has been reported that biological networks for omics data analysis and network biology.
consist of functional building blocks, known as network
motifs [5,120,121] . Theoretical and experimental perspectives 3.2.2. Applications of network biology
have proven that network motifs have critical biological Network biology is widely applied in various scientific
[5]
functions . Koschützki et al. developed a motif-based fields, including molecular biology, systems biology,
centrality based on network motifs [117] . For a given motif ecology, and network medicine (Figure 4). Some of its
in a biological network, they tallied the frequency of applications are discussed in this section.
each node involved in motifs, matching the given motif,
and they ranked the nodes according to their frequency. Network biology can be used to predict functions
The proposed motif-based method was applied to the of genes/proteins. Conventionally, functions of gene/
GRN of Escherichia coli, yielding interesting results protein can be inferred by sequence alignment analysis [122]
about key regulators. In 2014, we also proposed a motif or biological experiments [123] . Two genes/proteins with
centrality . Different from the motif-based centrality in high sequence similarity tend to have similar biological
[47]
another study [117] , we considered the frequency of each functions; therefore, the functions of genes/proteins in
node involved in the 2-node, 3-node, and 4-node motifs one species can be predicted through sequence alignment
in a biological network and developed a novel index based with genes/proteins in model organisms. Different from
on principal component analysis. The motif centrality traditional approaches, network biology, based on guilt-by-
was applied to the neural network for Caenorhabditis association and guilt-by-rewiring principles, provides an
elegans, TRNs for E. coli and Yeast, the Drosophila
developmental transcriptional network, as well as the
human signal transduction network. The results revealed
that the proposed motif centrality can effectively identify
command interneurons and key transcriptional factors in
the neural network and TRNs, respectively. By integrating
various known centrality measures, and based on principal
component analysis, an integrative measure to identify
structurally dominant proteins in PPI networks has been
proposed . Figure 4. Applications of network biology.
[82]

Volume 1 Issue 2 (2022) 7 https://doi.org/10.36922/gpd.v1i2.101