Page 68 - GPD-1-2

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

systems. There are many issues in network biology. An (Figure 1A). A=(a ) is often denoted as the adjacency
ij n×n
important issue is how to construct biological networks, matrix of a complex network. If a > 0, there is an edge
ij
which is a reverse problem. This construction of biological between nodes i and j in which a represents the weight of
ij
networks is also known as network inference, network the edge. If a = 0, there is no edge between nodes i and j.
ij
construction, topological identification, and so on. In For an undirected network, the adjacency matrix A is
fact, network construction is the first step in network symmetrical, otherwise not. For an unweighted network,
biology [4,32-42] . Reliable biological networks guarantee a only takes 1 or 0. Different types of networks encompass
ij
the accuracy of results from network analysis and the different measures to describe their structural features.
subsequent real-world applications. The second issue is how Taking undirected and unweighted networks as
to effectively explore bioinformatics in biological networks. examples, some basic topological features of complex
This is known as biological network analysis or topological/ networks are introduced [1,2,4] . Some of the commonly used
structural analysis of biological networks [4,32,43-53] . The measures to evaluate the topological structure of complex
third issue is the applications of network biology. In fact, networks include average degree and degree distribution,
biological networks are essential for understanding the average clustering coefficient, average path length (APL),
cellular mechanisms of various phenotypes; they also have disassortativity, and so on [1,2,4] . Average degree is defined
wide applications in exploring relationships among human as the average neighbors of each node in the network. The
diseases, discovering new drug targets, guiding drug clustering coefficient is used to describe the degree to which
repositioning, and controlling biological systems [27,28,31] . the adjacent points of a node are connected to each other.
The applications of network biology rely on reliable APL is defined as the average number of steps along the
network construction and efficient modeling and analysis shortest paths for all possible pairs of nodes in the network.
of related biological networks [4,28,29,44-46,53] .
The assortativity coefficient is the Pearson correlation
The rapid development and wide applications of network coefficient (PCC) of degree between pairs of linked nodes.
biology encouraged us to conduct this review. This review If PCC > 0, the network is assortative; but if PCC < 0,
focuses on the aforementioned issues of network biology, the network is disassortative. In a disassortative network,
aiming at introducing some recent advances and challenges nodes with high degrees tend to connect with low-degree
of several basic research topics, principles, and applications nodes. This is a typical feature of biological networks . The
[4]
in network biology. The rest of the paper is organized as disassortative features of biological networks are different
follows: Section 2 briefly introduces the complex network from those of social networks. In social networks, high-
theory; Section 3 reviews some recent advances of network degree nodes tend to connect with high-degree ones. In
biology, including the recent progresses of network addition to the aforementioned features, there are many
construction, network-based identification of important other statistical indices to describe a network. For more
genes/proteins, and their related applications; Section 4 details, several references can be referred to Barabási, Chen
puts forward some of the challenges in network biology; et al., and Lü et al. [1,2,4]
and the final section comprises the conclusion remarks.
Extensive centrality measures such as degree centrality,
2. Complex network theory clustering coefficient centrality [1,2,54] , betweenness
centrality , k-shell , semi-local centrality , PageRank ,
[55]
[57]
[58]
[56]
A complex network consists of nodes and edges . Nodes LeaderRank , adaptive LeaderRank , SpectralRank ,
[1]
[61]
[60]
[59]
represent the concerned entities in the system, while edges and so on have been developed to measure the importance
denote the relationships among nodes. As to biological of nodes in a complex network. These measures are all
networks, nodes may be genes, ribonucleic acids (RNAs), based on the characteristics of the node and edge in the
microRNAs (miRNAs), proteins, metabolites, or other network. Different measures evaluate the importance of a
molecules . Edges indicate that there are physical or node from different aspects. For example, degree centrality
[4]
chemical interactions, chemical reactions, or co-expression measures how many neighbors a node has; betweenness
relationships among biomolecules. A weight can be assigned centrality evaluates how many shortest paths that path
to each edge to represent the strength of the interaction or through a node, and whether a node can act as a bottleneck
co-expression between two nodes. Depending on the types in the network ; PageRank and LeaderRank are all based
[54]
of nodes and the meaning of edges, a biological network on random walks on the networks; the recently proposed
can be directed or undirected and weighted or unweighted SpectralRank is based on the dominant eigenvector of
(Figure 1A). Different types of networks can be modeled the augmented network of the originally considered
and explored through different methods. network . The augmented network is obtained by adding
[61]
Mathematically, a complex network can be described a leader node that is bidirectionally connected with all the
by its adjacency matrix, edge list, or node-edge matrix nodes in the network. It shows that SpectralRank can well
Volume 1 Issue 2 (2022) 2 https://doi.org/10.36922/gpd.v1i2.101