You are on page 1of 30

Introduction the role of graph theory

in human diseases
Rozgonyi Kristóf
Department of Physics of Complex Systems
Eötvös Loránd University

November 15, 2014


Motivation

• To understand the processes behind diseases


• To know the structure of cells, struktural problems in ill cells
• To model diseases

Introduction Motivation 2 / 30
Networks in biology
• A biological network is
any network that
applies to biological
systems
• Food webs

• Neuronal networks

• Metabolic networks

• etc

Introduction Networks in biology 3 / 30


Mathematical modeling of networks

• The basic mathematical concept used to model networks is a graph


• By denition graph is a representation of a set of objects where some
pairs of objects are connected by links
• The interconnected objects are represented by mathematical
abstractions called vertices or nodes, and the links that connect some
pairs of vertices are called edges

Graph theory Dentions 4 / 30


Directed and Undirected graphs
• There are two main types of graph:
• Directed graph
• Undirected graph

Figure : An example of directed graph (left) and an undirected graph (right)

Graph theory Dentions 5 / 30


Directed graphs
• A nite directed graph G consist a set of vertices or nodes V (G )
together with an edge set E (G )

V (G ) = {v1 , . . . , vn }
E (G ) ⊆ V (G ) × V (G )
• Each edge (u , v ) ∈ E (G ) can be tought of as connecting the starting
node u to the terminating node v
• The maximum number of edges is n(n − 1) where n the number of
nodes (without self loops)
• Examples from biology: food web, metabolic networks . . . etc.

Graph theory Dentions 6 / 30


Undirected graphs
• A nite undirected graph G also consist a set of nodes V (G ) and an
edge set E (G )
• We can think of it as a directed graph which edge set consist
(v , u ) ∈ E (G ) if ∃ (u , v ) ∈ E (G )
• The maximum number of edges is n(n2−1) where n is the number of
nodes (without self loops)
• Examples from biology: protein-protein interaction networks,
metabolic networks . . . etc.

Graph theory Dentions 7 / 30


Adjacency matrix, neighborhood
• Suppose the vertices of a graph G are ordered as v1 , . . . , vn . Then the
adjacency matrix, A of G is given by

1 vi vi +1 ∈ E (G )

aij = 0
if

if vi vi +1 ∈
/ E (G )

• If the graph G is undirected, the matrix A is simmetric


• If we have an edge (u , v ) then the vertices u and v is said to be
adjancent or connected to each other
• In this case we also said that u and v are neighbours
• The set of all neighbours of u is denoted N (u )

Graph theory Dentions 8 / 30


Node-degree
• For an undirected graph G, we shall write deg (u ) for the degree of u
in V (G )
• Node-degree is the total numberof edges at u

deg (u ) = |N (G )|
• For a directed graph we shall write
• in-degree (degin (u )) of a vertex u is given by the number of edges that
terminate at u
• out-degree (degout (u )) of a vertex u is given by the number of edges
that start at u

Graph theory Dentions 9 / 30


Paths, path lenght
• Let u ,v be two vertices in graph G. Then a sequence of vertices

u = v1 , v2 , . . . , vk = v
where

(vi , vi +1 ) ∈ E (G )
vi 6= vj for i 6= j
• It is said to be a path with a length k − 1 from u to v
• The geodesic distance (or distance), δ(u , v ) from u to v is the lenght
of the shortest path from u to v in G

Graph theory Dentions 10 / 30


Connected graphs

Figure : Example connected graph, where k = 5 and δ(u , v ) = 4

• If for every pair of vertices (u , v ∈ V (G )), there is some path from u


to v , then we say G is connected

Graph theory Dentions 11 / 30


Clustering coecient
• Suppose u is a node of degree k in graph G and there are e edges
between the k neighbours of u
• Then the clustering coecient of u in G

2e
Cu =
k (k − 1 )

• Cu mesures the possibility that two neighbours of u is connected to


each other

Graph theory Dentions 12 / 30


Graph models
• Graphs can be used to
model many types of
relations and processes in
physical, biological. . . etc
systems

• Many practical problems


can be represented by
graphs

• We will see some example graph modeling methods

Graph theory Graph models 13 / 30


Erd®s-Rényi model
• In this model the graph is constructed by connecting nodes randomly
• This graph can be dened as G (n, p ), where
• n the number of vertices
• p is a real number, increases 0 to 1
• Each edge is included in the graph with probability p independent
from every other edge
• All graphs created from this model with n nodes and m edges have
equal probability Pr (m)
 n 
Pr (m) = pm (1 − p)(2)−m
n
2
m

This graph G (n, p ) has on average n2 p edges




Graph theory Graph models 14 / 30


Scale-free networks
• The degree distribution of the ER graph is binomial

k −1 k
 
P (deg (v ) = k ) = p (1 − p)n−1−k
k
• The degree distribution of many biological networks appear to follow
board-tailed power law (2 < γ < 3)

P (k ) ∼ k −γ

• This behaviour is inconsistens with the result of the ER model


⇒ we need a new model
• This type of graphs which follow board-tailed power law said to be
scale-free graphs
Graph theory Graph models 15 / 30
Barabási-Albert model
• Basicly the BA model consider a network as an evolving entity and it
is modeling the dynamics of network growth
⇒ it is a dynamic graph model
• To create a BA graph needed to dene the initial conditions
• Set the parameter m (m ∈ Z+ )
• Give an initial graph G0
• Then let the network evolve
• This evolution of a BA graph is a discrete-time process

Graph theory Graph models 16 / 30


BA model evolution rules
• These are the basic BA model evolution rules
• Growth: Each time j a new node of degree m is added to the network
• Preferential Attachment : For each node u in the existing network, the
probability that the new node connect to it is propotional to the degree
u
of

• Formally, writing Gj for the network at time j and P (u , j ) for the


probability that the new node is connected to u
deg (u )
P (u , j ) = P
v ∈V (G −1 ) deg (v )
j

• Where v is all the nodes ∈ V (Gj −1 )


• The degree distribution of a BA graph
P (k ) ∼ k −3

Graph theory Graph models 17 / 30


Problems with BA model
• BA model is not based on specic biological considerations
• The degree distribution is just one property of a network, there are
other important structural aspects
• We are in eect drawing conclusions about the topology of an entrie
network based on a shample of it nodes
• Subnetworks sampled from a scale-free network are not in general
scale-free
• It is possible for a sampled subnetwork of a network with
Poisson-distribution (which is certanly not scale-free) to appear to
scale-free

Graph theory Graph models 18 / 30


Levels of human diseases
• There are three levels of diseases

1 The top level of human diseases is the level of society

2 The second is the level of viruses, bacterials and enviromental diseases

3 The third is the gene-level diseases

• Graph theory widely used in all of the three levels

Human diseases Introduction 19 / 30


Gene-level diseases

• In this level is the most useful apply graph theory, so this presentation
will be about this level
• Many human diseases have a genetic component is said to be genetic
disorder
• A genetic disorder is an illness caused by one or more abnormalities in
the genome (genotype), especially a condition that is present from
birth

Human diseases Introduction 20 / 30


Genotype-phenotype

• Genotype : is the genetic makeup of a cell, an organism, or an


individual usually with reference to a specic characteristic under
consideration (genetic code), generally denes the phenotype
• Phenotype : is the composite of an organism's observable
characteristics or traits, such as its morphology, development,
biochemical or physiological properties, phenology, behavior, and
products of behavior

Human diseases Introduction 21 / 30


Cells as networks
• In the last decades the organizational view of the cell changed from
being merley a "bag of enzymes" to a web of interconnected organelles
• Cells can described as complex webs of macromolecular interactions
⇒We can model cells with networks
• The nodes are proteins, molecules. . . ect.
• The edges are physical interactions, reactions. . . ect.
• Once shown to be that cellular networks are statistically dierent from
randomized networks
• It is very dicult mapping intercellular networks

Human diseases Intercellular networks 22 / 30


Modeling gene-related diseases

Figure : Modeling of genotype-phenotype relationships and disease

• We representate diseases as graph perturbations in our models


• We can dene not just one network in a cell, and the same molecule
can be appear more networks

Human diseases Intercellular networks 23 / 30


Metabolic networks
• Metabolic network maps attempt to comprehensively describe all
possible biochemical reactions for a particular cell or organism
• There are many possibilities to representate metabolic networks
• Nodes are biochemical metabolities
• Edges are either the reactions that convert one metabolite into an
other or the enzymes that catalyze these reactions
• Edges can be directed or undirected, depending on wheter a given
reaction is reversible or not

• Metabolic networks are the most comprehensive of all biological


networks
• These maps are widely used in direct experimental investigation

Human diseases Intercellular networks 24 / 30


Other intercellular networks
• There are other possibilites
to representate a cell with
graphs
• For example the
protein-protein interaction
networks
• Nodes represent
proteins
• Edges are physical
interaction between two
proteins

Human diseases Intercellular networks 25 / 30


Hubs
• In scale-free networks most nodes have only a few interactions and
these coexist with a highly connected nodes the hubs, that holds the
whole network together
• This scale-free property has been found in all organism for which
protein-protein interaction and metabolic network map exist from
yeast to human
• Hubs seem to be wery important object within cells
• Based on measurments we expect that, in humans, hubs should
preferentially encode disease-relateg genes

Human diseases Intercellular networks 26 / 30


Origin of scale-free property
• The scale-free topology of biological networks likely originates from
gene duplication
• When a cell divide and the genom replicates, occasionally an extra
copy of genes or cromosoes get produced
• After the divide, both of the original and new extra copy can interact
with the same set of partners
• Consequently, each of the protein partners that interacted with the
anestor, gains a new interaction
• This process result in the "ritch-get-ritcher" phenomenon
• This mechanism generate hubs and and could be responsible for the
scale-free property

Human diseases Intercellular networks 27 / 30


Otherview

• Looked at the basics of graph theory


• Summaryzed how to model gene-related diseases
• Saw some example of biological networks
• Discussed the role of hubs and the origin of scale-free properties

Otherview 28 / 30
Thank you for your attention!

29 / 30
References

O.Mason, M.Verwoerd (2006) Graph theory and networks in biology


ArXiv: q-bio/0604006
M.Vidal, M.E.Cusick, A.Barabásy (2011) Interactome network and
human disease Doi:10.1016/j.cell.2011.02.016

30 / 30

You might also like