Advanced

Bioinformatic approaches to gene expression in leukemia. Networks and deconvolution

Järvstråt, Linnea LU (2017)
Abstract
The aim of this thesis is develop methods to extract information from high-dimensionality data and to apply them in looking at gene-gene and gene-protein interactions in Acute Myeloid Leukemia (AML). The in silico methods developed can be used with data from other systems. The thesis gives an overview of the field for network inference and deconvolution methods, as well as a brief background on the biology concerning AML.

Paper I develops a method, Ultranet, for inferring Gaussian Graphical Models in an efficient manner. The models can created by solving the sparse inverse covariance selection (SICS) problem be used to identify gene networks from data containing a large number of variables but a proportionately low sample... (More)
The aim of this thesis is develop methods to extract information from high-dimensionality data and to apply them in looking at gene-gene and gene-protein interactions in Acute Myeloid Leukemia (AML). The in silico methods developed can be used with data from other systems. The thesis gives an overview of the field for network inference and deconvolution methods, as well as a brief background on the biology concerning AML.

Paper I develops a method, Ultranet, for inferring Gaussian Graphical Models in an efficient manner. The models can created by solving the sparse inverse covariance selection (SICS) problem be used to identify gene networks from data containing a large number of variables but a proportionately low sample number. We apply Ultranet to data from several blood disorders and show that the models capture blood related gene interactions.

In paper II, we investigate how the cellular heterogeneity of many tissue samples can be taken into consideration when examining gene expression of said samples. Cell hierarchies are thought to be present in leukemia and some solid tumors, sustained by cancer stem cells (CSCs). We developed a computational approach that extracts gene expression patterns and cell type proportions in silico.

Paper III and IV looks at how the DEK and WT1 proteins, respectively, interact with DNA using chromatin immunoprecipitation followed by sequencing. The proteins are known to be altered in a range of different cancer forms, including Acute Myeloid Leukemia.

We find, in paper III, that DEK binds close to transcription start sites of actively transcribed genes as indicated by the presence of the histone markers. DEK binds to genes that are ubiquitously expressed across tissues, represented by presence of RNA polymerase II binding sites in cell lines. We also show that knockdown of DEK by shRNA results in both significant down- and up-regulations of DEK-bound genes, suggesting complex interactions.

In paper IV, we study how the binding patterns of the WT1 isoforms differ. There is an alternative splice site between zinc finger three and four which leads to inclusion or exclusion of three amino acids (+/-KTS). We show that WT1 -KTS binds to transcription start sites, while WT1 +KTS bind inside gene bodies. (Less)
Please use this url to cite or link to this publication:
author
supervisor
opponent
  • professor Persson, Bengt, Uppsala University
organization
alternative title
Bioin
publishing date
type
Thesis
publication status
published
subject
keywords
bioinformatics, acute myeloid leukemia, networks, deconvolution, microarrays, ChIP-seq, DEK, WT1
pages
42 pages
publisher
Lund University, Faculty of Medicine
defense location
Föreläsningssal BMC I1345, Sölvegatan 19 i Lund
defense date
2017-11-09 13:00
ISBN
978-91-7619-521-5
language
English
LU publication?
yes
id
314ede26-4174-4ea4-8183-91ebeefc7724
date added to LUP
2017-10-18 09:23:07
date last changed
2017-10-23 10:05:13
@phdthesis{314ede26-4174-4ea4-8183-91ebeefc7724,
  abstract     = {The aim of this thesis is develop methods to extract information from high-dimensionality data and to apply them in looking at gene-gene and gene-protein interactions in Acute Myeloid Leukemia (AML). The <i>in silico</i> methods developed can be used with data from other systems. The thesis gives an overview of the field for network inference and deconvolution methods, as well as a brief background on the biology concerning AML. <br/><br/>Paper I develops a method, Ultranet, for inferring Gaussian Graphical Models in an efficient manner. The models can created by solving the sparse inverse covariance selection (SICS) problem be used to identify gene networks from data containing a large number of variables but a proportionately low sample number. We apply Ultranet to data from several blood disorders and show that the models capture blood related gene interactions. <br/><br/>In paper II, we investigate how the cellular heterogeneity of many tissue samples can be taken into consideration when examining gene expression of said samples. Cell hierarchies are thought to be present in leukemia and some solid tumors, sustained by cancer stem cells (CSCs). We developed a computational approach that extracts gene expression patterns and cell type proportions<i> in silico</i>. <br/><br/>Paper III and IV looks at how the DEK and WT1 proteins, respectively, interact with DNA using chromatin immunoprecipitation followed by sequencing. The proteins are known to be altered in a range of different cancer forms, including Acute Myeloid Leukemia. <br/><br/>We find, in paper III, that DEK binds close to transcription start sites of actively transcribed genes as indicated by the presence of the histone markers. DEK binds to genes that are ubiquitously expressed across tissues, represented by presence of RNA polymerase II binding sites in cell lines. We also show that knockdown of DEK by shRNA results in both significant down- and up-regulations of DEK-bound genes, suggesting complex interactions. <br/><br/>In paper IV, we study how the binding patterns of the WT1 isoforms differ. There is an alternative splice site between zinc finger three and four which leads to inclusion or exclusion of three amino acids (+/-KTS). We show that WT1 -KTS binds to transcription start sites, while WT1 +KTS bind inside gene bodies.},
  author       = {Järvstråt, Linnea},
  isbn         = {978-91-7619-521-5},
  keyword      = {bioinformatics,acute myeloid leukemia,networks,deconvolution,microarrays,ChIP-seq,DEK,WT1},
  language     = {eng},
  pages        = {42},
  publisher    = {Lund University, Faculty of Medicine},
  school       = {Lund University},
  title        = {Bioinformatic approaches to gene expression in leukemia. Networks and deconvolution},
  year         = {2017},
}