Project VII: Simulation-based analysis of the spatial
configuration of nuclear bodies.
A key biological objective is to understand and represent structural
characteristics of cell nuclei in order to gain insight into the various
mechanisms of cell organization. This can
to be achieved using statistical hypothesis testing, inference and
classification to assess the evidence relating to the configuration of cellular
and nuclear objects.
Confocal
microscopy images of mammalian interphase cell nuclei stained with a variety of
nuclear components are available and can be used to investigate their spatial
configuration. Spatial probabilistic
models are now routinely used in biostatistics and epidemiology, and can be
used for the analysis of biological data.
Advances in the field of simulation have made simulation-based more
attractive, and this is the basis of the inference mechanisms in this project.
The set of squared distances
between pairs of loci in the nucleus can be used to define simulation-based
statistical tests. Simulation-based
hypothesis testing is characterized by need for a good choice of test
statistics to ensure high statistical power, and efficient computation of null distribution
through simulation; to test a specific
null hypothesis, simulation methods can be implemented. For tests on single
cells, Monte Carlo tests will often be sufficient, but when a random sample of
cells are available, it will be necessary and important to be able to combine
information across the samples in order to carry out (for example) likelihood
ratio tests or likelihood/Bayesian inference.