Many real-world systems can be represented as networks, including biological networks.

Networks are found in biological systems of varying scales, from

the evolutionary tree of life, ecological networks, to expression networks, and regulatory networks.

Can you name any other biological phenomena that can be represented as a network?

Protein or metabolic interactions in a call can be represented also as a graph or a network.

In biological networks, nodes and edges can represent different things.

Nodes can be proteins, peptides, or non protein biomolecules.

Edges can be any biological

relationship such as interactions, regulations, or transformations.

One can construct

bipartite or three-partite networks or graphs, for instance, between genes, proteins, and drugs.

Do you recall the definition of a bipartite graph?

Protein-protein interaction networks,

transcriptional regulatory networks, and metabolic networks, have been studied and their structural properties are reported to be similar.

The observed properties are not in agreement with traditional random graph models for complex networks.

Their degree distribution follows a power law degree distribution, basically

rich get richer. They have a small average shortest path length.

They are very resilient and have a strong resistance to failure on random attacks, but vulnerable to targeted attacks.

They have hierarchical modularity and a few central nodes and

also as we mentioned earlier, it has been observed that in biological networks

more central nodes these cannot be done without it.

The graph shown here is a protein-protein interaction network or PPI and

has been constructed using three different public databases.

This protein interaction network has more than 1,500 proteins and

337,000 interactions.

PPI are typically modeled as undirected graphs in which nodes represent proteins and edges represent

interactions.

The frequency of proteins having interactions with exactly K other proteins follows a power law.

PPI networks exhibits the small world phenomena. One can reach any node within a small number of hops,

usually four or five hops.

Although these networks are very sparse, it's very hard to be studied without breaking them into smaller pieces.

Biological networks are modular. Therefore, one can try to find highly clustered

subgraphs of the biological network to get a high-level overview of it.

This also allows us to predict unknown gene or protein

functions based on its communities.

Diverse algorithms and methods range from simple k-mean clustering to Markov chain based models have been

suggested in literature to find these modules and community in a complex network.

One of the main challenges in community detection is the number of communities.

Most of these methods have a set of parameters that needs to be said before running them.

Changing the exact value of these parameters

affects the number of detected communities or modules even if it has not been asked explicitly.

For instance, what's shown here is a module of the PPI Network found by the MCL method.

Many of the genes here are known to belong to a particular biological pathway.

As another example of biological networks here, we have a drug-drug interaction network.

This network has been reconstructed based on similarity of the effect of each drug on a set of genes.

This type of network is very useful in drug repurposing.

The construction of biological networks directly from experimental data

using statistical or machine learning approaches has received significant attention in the last decade.

There is a wide variety of different approaches available that can be used to infer biological networks from experimental data.

Approaches employed include: Bayesian networks, auto regressive models, correlation based,

mutual information based models, clustering techniques, or differential equation models.

Most of these methods use perturbation to validate the reconstructed network.

This can be performed by removing a node, gene, protein, or metabolite from a system or by preventing

interaction among particular ones.

Biological networks are very complex and diverse

so we cannot analyze them only by relying on statistical based methods. In the remaining part of this course

you will learn about a new way of studying and exploring such complex networks.