Dissertationen zum Thema „Cancer bioinformatics“

Cancer is a complex and multifaceted disease, and a vast amount of time and effort has been spent on characterizing its behaviors, identifying its weaknesses, and discovering effective treatments. Two major obstacles stand in the way of progress toward effective precision treatment for the majority of patients.

First, cancer's extraordinary heterogeneity—both between and even within patients—means that most patients present with a disease slightly different from every previously recorded case. New methods are necessary to analyze the growing body of patient data so that we can classify each new patient with as much accuracy and precision as possible. In chapter 2 I present a method that integrates data from multiple genomics platforms to identify axes of variation across breast cancer patients, and to connect these gene modules to potential therapeutic options. In this work we find modules describing variation in the tumor microenvironment and activation of different cellular processes. We also illustrate the challenges and pitfalls of translating between model systems and patients, as many gene modules are poorly conserved when moving between datasets.

A second problem is that cancer cells are constantly evolving, and many treatments inevitably lead to resistance as new mutations arise or compensatory systems are activated. To overcome this we must find rational combinations that will prevent resistant adaptation before it can start. Starting in chapter 3 I present a series of projects in which we used a high-throughput proteomics approach to characterize the activity of a large proportion of protein kinases, ending with the discovery of a promising drug combination for the treatment of breast cancer in chapter 8.

Years of sequence feature curation by UniProtKB/Swiss-Prot, PIR-PSD, NCBI-CDD, RefSeq and other database biocurators has led to a rich repository of information on functional sites of genes and proteins. This information along with variation-related annotation can be used to scan human short sequence reads from next-generation sequencing (NGS) pipelines for presence of non-synonymous single-nucleotide variations (nsSNVs) that affect functional sites. This and similar workflows are becoming more important because thousands of NGS data sets are being made available through projects such as The Cancer Genome Atlas (TCGA), and researchers want to evaluate their biomarkers in genomic data. BioMuta, an integrated sequence feature database, provides a framework for automated and manual curation and integration of cancer-related sequence features so that they can be used in NGS analysis pipelines. Sequence feature information in BioMuta is collected from the Catalogue of Somatic Mutations in Cancer (COSMIC), ClinVar, UniProtKB and through biocuration of information available from publications. Additionally, nsSNVs identified through automated analysis of NGS data from TCGA are also included in the database. Due to the petabytes of data and sequence information present in NGS primary databases, a High-performance Integrated Virtual Environment (HIVE) platform for storing, analyzing, computing and curating NGS data and associated metadata has been developed. Using HIVE, 31,979 nsSNVs were identified in TCGA-derived NGS data from breast cancer patients. All variations identified through this process are stored in a Curated Short Read archive, and the nsSNVs from the tumor samples are included in BioMuta. Currently, BioMuta has 26 cancer types with 13,896 small scale and 308,986 large scale study-derived variations. Integration of variation data allows identifications of novel or common nsSNVs that can be prioritized in validation studies.

In the last century cancer has become increasingly prevalent and is the second largest killer in the United States, estimated to afflict 1 in 4 people during their life. Despite our long history with cancer and our herculean efforts to thwart the disease, in many cases we still do not understand the underlying causes or have successful treatments. In my graduate work, I’ve developed two approaches to the study of cancer genomics and applied them to the whole genome sequencing data of cancer patients from The Cancer Genome Atlas (TCGA). In collaboration with Dr. Ewing, I built a pipeline to detect retrotransposon insertions from paired-end high-throughput sequencing data and found somatic retrotransposon insertions in a fifth of cancer patients.

My second novel contribution to the study of cancer genomics is the development of the CN-AVG pipeline, a method for reconstructing the evolutionary history of a single tumor by predicting the order of structural mutations such as deletions, duplications, and inversions. The CN-AVG theory was developed by Drs. Haussler, Zerbino, and Paten and samples potential evolutionary histories for a tumor using Markov Chain Monte Carlo sampling. I contributed to the development of this method by testing its accuracy and limitations on simulated evolutionary histories. I found that the ability to reconstruct a history decays exponentially with increased breakpoint reuse, but that we can estimate how accurately we reconstruct a mutation event using the likelihood scores of the events. I further designed novel techniques for the application of CN-AVG to whole genome sequencing data from actual patients and applied these techniques to search for evolutionary patterns in glioblastoma multiforme using sequencing data from TCGA. My results show patterns of two-hit deletions, as we would expect, and amplifications occurring over several mutational events. I also find that the CN-AVG method frequently makes use of whole chromosome copy number changes following by localized deletions, a bias that could be mitigated through modifying the cost function for an evolutionary history.

Lung cancer is the cancer form that has the highest mortality worldwide and inaddition the survival of lung cancer is very low. Only 15% of the patients are alivefive years from set diagnosis. More research is needed to understand the biologyof lung cancer and thus make it possible to discover the disease at an early stage.Early diagnosis leads to an increased chance of survival. In this thesis 179 lungcancer- and 116 control samples of blood serum were analyzed for identificationof metabolomic biomarkers. The control samples were derived from patients withbenign lung diseases.Data was gained from GC/TOF-MS analysis and analyzed with the help ofthe multivariate analysis methods PCA and OPLS/OPLS-DA. In this thesis it isinvestigated how to pre-treat and analyze the data in the best way in order todiscover biomarkers. One part of the aim was to give directions for how to selectsamples from a biobank for further biological validation of suspected biomarkers.Models for different stages of lung cancer versus control samples were computedand validated. The most influencing metabolites in the models were selected andconfoundings with other clinical characteristics like gender and hemoglobin levelswere studied. 13 lung cancer biomakers were identified and validated by raw dataand new OPLS models based solely upon the biomarkers.In summary the identified biomarkers are able to separate fairly good betweencontrol samples and late lung cancer, but are poor for separation of early lungcancer from control samples. The recommendation is to select controls and latelung cancer samples from the biobank for further confirmation of the biomarkers.NyckelordLung cancer is the cancer form that has the highest mortality worldwide and inaddition the survival of lung cancer is very low. Only 15% of the patients are alivefive years from set diagnosis. More research is needed to understand the biologyof lung cancer and thus make it possible to discover the disease at an early stage.Early diagnosis leads to an increased chance of survival. In this thesis 179 lungcancer- and 116 control samples of blood serum were analyzed for identificationof metabolomic biomarkers. The control samples were derived from patients withbenign lung diseases.Data was gained from GC/TOF-MS analysis and analyzed with the help ofthe multivariate analysis methods PCA and OPLS/OPLS-DA. In this thesis it isinvestigated how to pre-treat and analyze the data in the best way in order todiscover biomarkers. One part of the aim was to give directions for how to selectsamples from a biobank for further biological validation of suspected biomarkers.Models for different stages of lung cancer versus control samples were computedand validated. The most influencing metabolites in the models were selected andconfoundings with other clinical characteristics like gender and hemoglobin levelswere studied. 13 lung cancer biomakers were identified and validated by raw dataand new OPLS models based solely upon the biomarkers.In summary the identified biomarkers are able to separate fairly good betweencontrol samples and late lung cancer, but are poor for separation of early lungcancer from control samples. The recommendation is to select controls and latelung cancer samples from the biobank for further confirmation of the biomarkers.Nyckelord

Data mining approaches have been increasingly used in recent years in order to find patterns and regularities in large databases. In this study, the C4.5 decision tree approach was used for mining of Gene Logic database, containing biological data. The decision tree approach was used in order to identify the most relevant genes and risk factors involved in breast cancer, in order to separate healthy patients from breast cancer patients in the data sets used. Four different tests were performed for this purpose. Cross validation was performed, for each of the four tests, in order to evaluate the capacity of the decision tree approaches in correctly classifying ‘new’ samples. In the first test, the expression of 108 breast related genes, shown in appendix A, for 75 patients were used as input to the C4.5 algorithm. This test resulted in a decision tree containing only four genes considered to be the most relevant in order to correctly classify patients. Cross validation indicates an average accuracy of 89% in classifying ‘new’ samples. In the second test, risk factor data was used as input. The cross validation result shows an average accuracy of 87% in classifying ‘new’ samples. In the third test, both gene expression data and risk factor data were put together as one input. The cross validation procedure for this approach again indicates an average accuracy of 87% in classifying ‘new’ samples. In the final test, the C4.5 algorithm was used in order to indicate possible signalling pathways involving the four genes identified by the decision tree based on only gene expression data. In some of cases, the C4.5 algorithm found trees suggesting pathways which are supported by the breast cancer literature. Since not all pathways involving the four putative breast cancer genes are known yet, the other suggested pathways should be further analyzed in order to increase their credibility.

In summary, this study demonstrates the application of decision tree approaches for the identification of genes and risk factors relevant for the classification of breast cancer patients

Cancer is a leading cause of morbidity and mortality, and one in three individuals in the U.S. will be diagnosed with cancer in their lifetime. At the molecular level, cancer is driven by the activity of oncogenes and the loss of activity of tumor suppressors. The availability of genomic data from large sets of tumor tissue have facilitated the identification of subgroups of patients whose tumors share molecular patterns of expression. These molecular signatures, in turn, can help identify clinically-useful patient subgroups and inform potential therapeutic strategies against cancer.

In chapter 1, I review the current theories behind carcinogenesis, the molecular factors that regulate gene expression, and statistical methods for analyzing genomic data. In chapter 2, I describe an approach, termed oncomix, developed to identify oncogene candidates from expression data obtained from tumor and adjacent normal tissue. I apply oncomix to breast cancer expression data and identify an oncogene candidate, CBX2, whose expression is gained in a subset of breast tumors. CBX2 is expressed at low levels in most normal adult tissue, and the CBX2 protein contains a drug-targetable chromodomain, both of which are desirable properties in a potential therapeutic target. We then provide the first experimental evidence that CBX2 regulates the growth of breast cancer cells. In chapter 3, I develop a method for identifying nuclear hormone receptors whose expression is lost in endometrial cancers relative to normal tissue. I report, for the first time, that the loss of expression of Thyroid Hormone Receptor Beta (THRB) is associated with better 5-year survival in endometrial cancer. The loss of THRB expression is independent of the loss of estrogen and progesterone receptor expression, two genes whose loss of expression is known to be associated with poor survival. THRB expression could be considered as a biomarker to risk-stratify endometrial cancer patients. In Chapter 4, I develop a user-friendly application for visualizing chromosomal copy number state obtained from three types of copy number input in single cells – fluorescence in situ hybridization (FISH), spectral karyotyping (SKY), and whole genome sequencing (WGS). This web application, termed aneuvis, automatically creates novel visualizations and summary statistics from a set of user-uploaded files that contain chromosomal copy number information.

In this thesis, I develop new computational approaches for identifying candidate molecular regulators of cancer. I also develop a new user-friendly tool to enable biological researchers to identify aneuploidy and chromosomal instability within populations of single cells. Applying these tools to breast and endometrial cancer genomic datasets has highlighted novel aspects of breast and endometrial cancer biology and may inform novel therapeutic strategies based on molecular patterns of genomic heterogeneity. The freely available software developed as part of these projects has the potential to enable other researchers to advance our understanding of cancer genomics and to inform novel therapeutic strategies against cancer.

MicroRNAs (miRNAs) are a class of small non-coding RNAs that control gene expression by targeting mRNAs and triggering either translation repression or RNA degradation. Several groups pointed out that miRNAs play a major role in several diseases, including cancer. This is assumed since the expression level of several miRNAs differs between normal and cancerous cells. Further, it has been shown that miRNAs are involved in cell proliferation and cell death.

Because of this role it is suspected that miRNAs could serve as biomarkers to improve tumor classification, therapy selection, or prediction of survival. In this context, it is questioned, among other things, whether miRNA deregulations in cancer cells occur according to some pattern or in a rather random order. With this work we contribute to answering this question by adapting two approaches (Beerenwinkel et al. (J Comput Biol, 2005) and Höglund et al. (Gene Chromosome Canc, 2001)), developed to derive graph models of oncogenesis for chromosomal imbalances, to miRNA expression data and applying them to a breast cancer data set. Further, we evaluated the results by comparing them to results derived from randomly altered versions of the used data set.

We could show that miRNA deregulations most likely follow a rough temporal order, i.e. some deregulations occur early and some occur late in cancer progression. Thus, it seems to be possible that the expression level of some miRNAs can be used as indicator for the stage of a tumor. Further, our results suggest that the over expression of mir-21 as well as mir-102 are initial events in breast cancer oncogenesis.

Additionally, we identified a set of miRNAs showing a cluster-like behavior, i.e. their deregulations often occur together in a tumor, but other deregulations are less frequently present. These miRNAs are let-7d, mir-10b, mir-125a, mir-125b, mir-145, mir-206, and mir-210.

Further, we could confirm the strong relationship between the expression of mir-125a and mir-125b.

Somatic copy number alterations (SCNAs) are a class of alterations that lead to deviations from diploidy in developing and established tumors. A feature that distinguishes SCNAs from other alterations is their genomic footprint. The large genomic footprint of SCNAs in a typical cancer's genome presents both a challenge and an opportunity to find targetable vulnerabilities in cancer. Because a single event affects many genes, it is often challenging to identify the tumorigenic targets of SCNAs. Conversely, events that affect multiple genes may provide specific vulnerabilities through "bystander" genes, in addition to vulnerabilities directly associated with the targets.

We approached the goal of understanding how the structure of SCNAs may lead to dependency in two ways. To improve our understanding of how SCNAs promote tumor progression we analyzed the SCNAs in 4934 primary tumors in 11 common cancers collected by the Cancer Genome Atlas (TCGA). The scale of this dataset provided insights into the structure and patterns of SCNA, including purity and ploidy rates across disease, mechanistic forces shaping patterns of SCNA, regions undergoing significantly recurrent SCNAs, and correlations between SCNAs in regions implicated in cancer formation.

In a complementary approach, we integrating SCNA data and pooled RNAi screening data involving 11,000 genes across 86 cell lines to find non-driver genes whose partial loss led to increased sensitivity to RNAi suppression. We identified a new set of cancer specific vulnerabilities predicted by loss of non-driver genes, with the most significant gene being PSMC2, an obligate member of the 26S proteasome. Biochemically, we found that PSMC2 is in excess of cellular requirement in diploid cells, but becomes the stoichiometric limiting factor in proteasome formation after partial loss of this gene.

In summary, my work improved our understanding of the structure and patterns of SCNA, both informing how cancers develop and predicting novel cancer vulnerabilities. Our characterization of the SCNAs present across 5000 tumors uncovered novel structure in SCNAs and significant regions likely to contain driver genes. Through integrating SCNA data with the results of a functional genetic screen, we also uncovered a new set of vulnerabilities caused by unintended loss of non-driver genes.

Another important problem in bioinformatics is the question of how the inferred networks’ quality can be evaluated. The current best practice is a two step procedure. In the first step, the highest scoring interactions are compared to known interactions stored in biological databases. The inferred networks passes this quality assessment if there is a large overlap with the known interactions. In this case, a second step is carried out in which unknown but high scoring and thus promising new interactions are validated ’by hand’ via laboratory experiments. Unfortunately when integrating prior knowledge in the inference procedure, this validation procedure would be biased by using the same information in both the inference and the validation. Therefore, it would no longer allow an independent validation of the resulting network.

The main contribution of this thesis is a complete computational framework that uses experimental knock down data in a cross-validation scheme to both infer and validate directed networks. Its components are i) a method that integrates genomic data and prior knowledge to infer directed networks, ii) its implementation in an R/Bioconductor package and iii) a web application to retrieve prior knowledge from PubMed abstracts and biological databases. To infer directed networks from genomic data and prior knowledge, we propose a two step procedure: First, we adapt the pairwise feature selection strategy mRMR to integrate prior knowledge in order to obtain the network’s skeleton. Then for the subsequent orientation phase of the algorithm, we extend a criterion based on interaction information to include prior knowledge. The implementation of this method is available both as part of the prior retrieval tool Predictive Networks and as a stand-alone R/Bioconductor package named predictionet.

Furthermore, we propose a fully data-driven quantitative validation of such directed networks using experimental knock-down data: We start by identifying the set of genes that was truly affected by the perturbation experiment. The rationale of our validation procedure is that these truly affected genes should also be part of the perturbed gene’s childhood in the inferred network. Consequently, we can compute a performance score
Doctorat en Sciences
info:eu-repo/semantics/nonPublished

Metastatic, castrate-resistant prostate cancer (mCRPC) is one of the most prevalent cancers and is the third leading cause of cancer death among men. Several treatment options have been developed to combat mCRPC, however none have produced any tangible benefits to patients' overall survivability. As part of a crowd-sourced algorithm development competition, participants were asked to develop new prognostic models for mCRPC patients treated with docetaxel. Such results could potentially assist in clinical decision making for future mCRPC patients.

In this thesis, we present a new ensemble prognostic model to perform risk prediction for mCRPC patients treated with docetaxel. We rely on traditional survival analysis model like the Cox Proportional Hazard model, as well as more recently developed boosting model that incorporates smooth approximation of the concordance index for direct optimization. Our model performs better than the the current state-of-the-art mCRPC prognostic models for the concordance index performance measure and is competitive with these models on the integrated time-dependent area under the receiver operating characteristic curve.