New tools for quantitative analysis of nuclear architecture
Author(s)
Russell, Richard Anthony
Type
Thesis
Abstract
The cell nucleus houses a wide variety of macromolecular substructures including
the cell’s genetic material. The spatial configuration of these substructures is
thought to be fundamentally associated with nuclear function, yet the architectural
organisation of the cell nucleus is only poorly understood. Advances in microscopy
and associated fluorescence techniques have provided a wealth of nuclear image
data. Such images offer the opportunity for both visualising nuclear substructures
and quantitative investigation of the spatial configuration of these objects. In this
thesis, we present new tools to study and explore the subtle principles behind nuclear
architecture.
We describe a novel method to segment fluorescent microscopy images of nuclear
objects. The effectiveness of this segmentation algorithm is demonstrated using
extensive simulation. Additionally, we show that the method performs as well as
manual-thresholding, which is considered the gold standard. Next, randomisationbased
tests from spatial point pattern analysis are employed to inspect spatial interactions
of nuclear substructures. The results suggest new and interesting spatial
relationships in the nucleus. However, this approach probes only relative nuclear
organisation and cannot readily yield a description of absolute spatial preference,
which may be a key component of nuclear architecture.
To address this problem we have developed methodology based on techniques
employed in statistical shape analysis and image registration. The approach proposes
that the nuclear boundary can be used to align nuclei from replicate images
into a common coordinate system. Each nucleus and its contents can therefore be
registered to the sample mean shape using rigid and non-rigid deformations. This
aggregated data allows inference regarding global nuclear spatial organisation. For
example, the kernel smoothed intensity function is computed to return an estimate
of the intensity function of the registered nuclear object. Simulation provides evidence
that the registration procedure is sensible and the results accurate.
Finally, we have investigated a large database of nuclear substructures using
conventional methodology as well as our new tools. We have identified novel spatial
relationships between nuclear objects that offer significant clues to their function.
We have also examined the absolute spatial configuration of these substructures
in registered data. The results reveal dramatic underlying spatial preferences and
present new and clear insights into nuclear architecture.
the cell’s genetic material. The spatial configuration of these substructures is
thought to be fundamentally associated with nuclear function, yet the architectural
organisation of the cell nucleus is only poorly understood. Advances in microscopy
and associated fluorescence techniques have provided a wealth of nuclear image
data. Such images offer the opportunity for both visualising nuclear substructures
and quantitative investigation of the spatial configuration of these objects. In this
thesis, we present new tools to study and explore the subtle principles behind nuclear
architecture.
We describe a novel method to segment fluorescent microscopy images of nuclear
objects. The effectiveness of this segmentation algorithm is demonstrated using
extensive simulation. Additionally, we show that the method performs as well as
manual-thresholding, which is considered the gold standard. Next, randomisationbased
tests from spatial point pattern analysis are employed to inspect spatial interactions
of nuclear substructures. The results suggest new and interesting spatial
relationships in the nucleus. However, this approach probes only relative nuclear
organisation and cannot readily yield a description of absolute spatial preference,
which may be a key component of nuclear architecture.
To address this problem we have developed methodology based on techniques
employed in statistical shape analysis and image registration. The approach proposes
that the nuclear boundary can be used to align nuclei from replicate images
into a common coordinate system. Each nucleus and its contents can therefore be
registered to the sample mean shape using rigid and non-rigid deformations. This
aggregated data allows inference regarding global nuclear spatial organisation. For
example, the kernel smoothed intensity function is computed to return an estimate
of the intensity function of the registered nuclear object. Simulation provides evidence
that the registration procedure is sensible and the results accurate.
Finally, we have investigated a large database of nuclear substructures using
conventional methodology as well as our new tools. We have identified novel spatial
relationships between nuclear objects that offer significant clues to their function.
We have also examined the absolute spatial configuration of these substructures
in registered data. The results reveal dramatic underlying spatial preferences and
present new and clear insights into nuclear architecture.
Date Issued
2010-01
Date Awarded
2010-05
Copyright Statement
Attribution NoDerivatives 4.0 International Licence (CC BY-ND)
Advisor
Freemont, Paul
Sponsor
Medical Research Council
Creator
Russell, Richard Anthony
Publisher Department
Mathematics
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)