Summary
Acute Myeloid Leukaemia (AML) is an aggressive cancer of the
bone marrow, affecting formation of blood cells during haematopoiesis.
This thesis presents investigation of AML using mRNA
gene expression profiles (GEP) of samples extracted from the bone
marrow of healthy and diseased subjects. Here GEPs from purified
healthy haematopoietic populations, with different levels of
differentiation, form the basis for comparison with diseased
samples. We present a mathematical transformation of mRNA
microarray data to make it possible to compare AML samples,
carrying expanded aberrant haematopoietic progenitor populations,
with their closest normal cellular counterpart. Analysing >
1000 AML patients using this framework resulted in precise genetic
signatures for known AML karyotypes and decomposition of the
large group of patients with no such classification. Additionally,
using a murine model to investigate the role of telomerase in AML,
we were able to translate the observed effect into human AML
patients and identify specific genes involved, which also predict
survival patterns in AML patients. During these studies we have
applied methods for investigating differentially expressed genes
and genetic signatures and for reducing dimensionally of gene
expression data. Next, we have used machine-learning methods to
predict survival and to assess important predictors based on these
results. General application of a number of these methods has been
implemented into two public query-based gene-lookup webservices,
called HemaExplorer and BloodSpot. These web-services
support the aim of making data and analysis of haematopoietic cells
from mouse and human accessible for researchers without bioinformatics
expertise. Finally, in order to aid the analysis of the very
limited number of haematopoietic progenitor cells obtainable from
bone marrow aspirations, this thesis presents a method developed
to investigate transcription factor binding and histone modifications
by ChIP-Seq using pico-scale amounts of DNA.