Header

UZH-Logo

Maintenance Infos

Development of network-analysis tools and applications in biochemistry


Schütz, P. Development of network-analysis tools and applications in biochemistry. 2009, University of Zurich, Faculty of Science.

Abstract

Networks have been widely used in the last decade to analyze large-scale
data sets. This popularity originates from the possibility both to represent
pairwise interactions among arbitrary objects in simple, two dimensional
plots and to treat any type of data with the same formalism. In the first
part of this thesis, network technology is applied to examine the free-energy
surface of a protein using only a time series of an one-dimensional signal, e.g.
an intramolecular distance. In the second part, a new procedure to identify
tightly connected communities in large networks is presented. Common to
both procedures is the attempt to infer structures in the examined data by
analyzing the associated networks.
A complete characterisation of the free-energy surface is essential to understand
the folding mechanisms of a protein. With the experimental technique
“F¨orster Resonance Energy Transfer” (FRET) a single intramolecular
distance can be monitored with high time resolution over an extended period
of time. To assess the information content on the free-energy surface of a
FRET experiment, a new method called FESST (Free Energy Surface from
Single-molecule Time series) to extract free-energy basins from the time series
of a single distance is presented. The central assumption behind FESST
is that the distribution of the signal in small time windows is characteristic
for the actual free-energy basin. Applied to Beta3s, a 20-residue peptide with
native three stranded antiparallel �-sheet conformation, FESST extracts the
native state with more than 96% accuracy from a time series of the distance
between two C�-side chain atoms taken from a molecular dynamics simulation.
Furthermore, FESST extracts the barrier between folded and unfolded
state correctly up to a difference of two percent and detects three additional
free-energy basins stabilized mainly enthalpically. Extrapolating the
amount of data required by FESST to single molecule FRET experiments,
the free-energy basins of proteins with a folding time of few milliseconds can
be detected by FESST.
I
In the second part, a new procedure to detect modules in a large network
is presented. To identify groups of nodes with many internal and few
inter-community connections, the partition with highest “modularity” has to
be identified. This scoring function is used, because it allows an objective
and algorithm independent definition of a community. Here, an optimization
strategy called “MSG-VM” is presented that is both effective and efficient.
For multiple large benchmark networks, MSG-VM improves the literature
values for the highest modularity found. Furthermore, a new benchmark
network is suggested that represents coappearing words in the title of publications
authored by the famous physico-chemist Martin Karplus. Despite the
large overlap in vocabulary of the various topics of the work of M. Karplus,
the identified groups of words could be attributed to the different fields with
ease.

Abstract

Networks have been widely used in the last decade to analyze large-scale
data sets. This popularity originates from the possibility both to represent
pairwise interactions among arbitrary objects in simple, two dimensional
plots and to treat any type of data with the same formalism. In the first
part of this thesis, network technology is applied to examine the free-energy
surface of a protein using only a time series of an one-dimensional signal, e.g.
an intramolecular distance. In the second part, a new procedure to identify
tightly connected communities in large networks is presented. Common to
both procedures is the attempt to infer structures in the examined data by
analyzing the associated networks.
A complete characterisation of the free-energy surface is essential to understand
the folding mechanisms of a protein. With the experimental technique
“F¨orster Resonance Energy Transfer” (FRET) a single intramolecular
distance can be monitored with high time resolution over an extended period
of time. To assess the information content on the free-energy surface of a
FRET experiment, a new method called FESST (Free Energy Surface from
Single-molecule Time series) to extract free-energy basins from the time series
of a single distance is presented. The central assumption behind FESST
is that the distribution of the signal in small time windows is characteristic
for the actual free-energy basin. Applied to Beta3s, a 20-residue peptide with
native three stranded antiparallel �-sheet conformation, FESST extracts the
native state with more than 96% accuracy from a time series of the distance
between two C�-side chain atoms taken from a molecular dynamics simulation.
Furthermore, FESST extracts the barrier between folded and unfolded
state correctly up to a difference of two percent and detects three additional
free-energy basins stabilized mainly enthalpically. Extrapolating the
amount of data required by FESST to single molecule FRET experiments,
the free-energy basins of proteins with a folding time of few milliseconds can
be detected by FESST.
I
In the second part, a new procedure to detect modules in a large network
is presented. To identify groups of nodes with many internal and few
inter-community connections, the partition with highest “modularity” has to
be identified. This scoring function is used, because it allows an objective
and algorithm independent definition of a community. Here, an optimization
strategy called “MSG-VM” is presented that is both effective and efficient.
For multiple large benchmark networks, MSG-VM improves the literature
values for the highest modularity found. Furthermore, a new benchmark
network is suggested that represents coappearing words in the title of publications
authored by the famous physico-chemist Martin Karplus. Despite the
large overlap in vocabulary of the various topics of the work of M. Karplus,
the identified groups of words could be attributed to the different fields with
ease.

Statistics

Downloads

226 downloads since deposited on 14 Jan 2010
27 downloads since 12 months
Detailed statistics

Additional indexing

Item Type:Dissertation
Referees:Caflisch A, Schuler B
Communities & Collections:04 Faculty of Medicine > Department of Biochemistry
07 Faculty of Science > Department of Biochemistry
Dewey Decimal Classification:570 Life sciences; biology
Language:English
Date:9 July 2009
Deposited On:14 Jan 2010 11:35
Last Modified:06 Dec 2017 22:56
Number of Pages:107
Related URLs:http://opac.nebis.ch/F/?local_base=NEBIS&con_lng=GER&func=find-b&find_code=SYS&request=005928112

Download

Download PDF  'Development of network-analysis tools and applications in biochemistry'.
Preview
Filetype: PDF
Size: 6MB