240

Bioinformatics of the Brain

(AD), Autism, Bipolar Disorder (BD), Parkinson’s Disease (PD), Multiple

Sclerosis (MS), and schizophrenia is a current functional connectome research

area in neuroscience attracting many researchers.

The topological properties of a normal brain network are characterized by

being a small-world, scale-free network with a hierarchical structure consisting

of modules connected by hubs as outlined. On the other hand, neurological

disorders result in structural and functional brain network changes as shown

in various clinical research studies [32–34]. Neurological disorders such as AD,

schizophrenia and PD result in deviation from these properties as outlined in

the following sections. The organization of normal structural and functional

brain networks is described and the network studies of neurological disorders

with a focus on AD, MS, and epilepsy to discover possible common patterns

in the analysis of diseases are presented in [35]. The author proposes that hub

overload and failure which results in the separation of the hierarchical brain

network structure may be a common characteristic of several neurological

disorders.

9.7.1

AD Connectome

AD is a progressive neurological disorder prevalent mostly in elderly popula-

tions. The structure of a connectome in AD is affected resulting in the loss

of small-world, scale-free and hierarchical modular structure of normal brain

networks as shown in various studies.

Functional brain networks of 15 AD patients and 13 control subjects were

investigated in [36] by forming connectivity matrices of beta band-filtered elec-

troencephalography (EEG) channels and then analyzing the equivalent graphs

for characteristic path length L and cluster coefficient C. The authors report

that for various threshold values in converting the matrices to graphs, L was

significantly longer in AD patients than normal subjects although clustering

coefficients were similar and concluded that these results indicate the loss of

small-world feature in AD which may be used to diagnose this disease.

A sub-network kernel to validate the similarity between a pair of connec-

tomes to classify brain diseases is presented in [37]. The local to global topolog-

ical properties of brain network nodes are considered to evaluate similarities of

connectomes. The proposed method is tested on subjects with baseline func-

tional magnetic resonance imaging (fMRI) data obtained from the Alzheimer’s

Disease Neuroimaging Initiative (ADNI) database. The authors state that the

results indicate that their method outperforms several graph-based methods

in mild cognitive impairment classification.

The effective connectivity of default mode network in AD patients and

normal controls is investigated in [38] to find that the intensity and quantity

of connections in AD were decreased when compared to the control subjects.

In particular, the authors note that posterior cingulated cortex (PCC) was

strongly connected to most of the default mode network regions in control