No description available.

Standard survival analysis focuses on population-based studies. The objective of our work, survival prediction, is different: to find the most accurate model for predicting the survival times for each individual patient. We view this as a regression problem, where we try to map the features for each patient to his/her survival time. This is challenging in medical data due to the presence of irrelevant features, outliers, and missing class labels. Our approach consists of two major steps: (1) apply various grouping methods to segregate patients, and (2) apply different regression to each sub-group we obtained from the first step. We focus our experiments on a data set of 2402 patients (1260 censored). Our final predictor can obtain an average relative absolute error < 0.54. The experimental results verify that we can effectively predict survival times with a combination of statistical and machine learning approaches.

This thesis deals with a classification of glioma grade in high and low aggressive tumours and overall survival prediction based on magnetic resonance imaging. Data used in this work is from BRATS challenge 2019 and each set contains information from 4 weighting sequences of MRI. Thesis is implemented in PYTHON programming language and Jupyter Notebooks environment. Software PyRadiomics is used for calculation of image features. Goal of this work is to determine best tumour region and weighting sequence for calculation of image features and consequently select set of features that are the best ones for classification of tumour grade and survival prediction. Part of thesis is dedicated to survival prediction using set of statistical tests, specifically Cox regression

The area of data imputation, which is the process of replacing missing data with substituted values, has been covered quite extensively in recent years. The literature on the practical impact of data imputation however, remains scarce. This thesis explores the impact of some of the state of the art data imputation methods on HCC survival prediction and classification in combination with data-level methods such as oversampling. More specifically, it explores imputation methods for mixed-type datasets and their impact on a particular HCC dataset. Previous research has shown that, the newer, more sophisticated imputation methods outperform simpler ones when evaluated with normalized root mean square error (NRMSE). Contrary to intuition however, the results of this study show that when combined with other data-level methods such as clustering and oversampling, the differences in imputation performance does not always impact classification in any meaningful way. This might be explained by the noise that is introduced when generating synthetic data points in the oversampling process. The results also show that one of the more sophisticated imputation methods, namely MICE, is highly dependent on prior assumptions about the underlying distributions of the dataset. When those assumptions are incorrect, the imputation method performs poorly and has a considerable negative impact on classification.

The area of data imputation, which is the process of replacing missing data with substituted values, has been covered quite extensively in recent years. The literature on the practical impact of data imputation however, remains scarce. This thesis explores the impact of some of the state of the art data imputation methods on HCC survival prediction and classification in combination with data-level methods such as oversampling. More specifically, it explores imputation methods for mixed-type datasets and their impact on a particular HCC dataset. Previous research has shown that, the newer, more sophisticated imputation methods outperform simpler ones when evaluated with normalized root mean square error (NRMSE). Contrary to intuition however, the results of this study show that when combined with other data-level methods such as clustering and oversampling, the differences in imputation performance does not always impact classification in any meaningful way. This might be explained by the noise that is introduced when generating synthetic data points in the oversampling process. The results also show that one of the more sophisticated imputation methods, namely MICE, is highly dependent on prior assumptions about the underlying distributions of the dataset. When those assumptions are incorrect, the imputation method performs poorly and has a considerable negative impact on classification. / Forskningen kring data imputation, processen där man ersätter saknade data med substituerade värden, har varit omfattande de senaste åren. Litteraturen om den praktiska inverkan som data imputation metoder har på klassificering är dock otillräcklig. Det här kandidatexamensarbetet utforskar den inverkan som de nyare imputation metoderna har på HCC överlevnads klassificering i kombination med andra data-nivå metoder så som översampling. Mer specifikt, så utforskar denna studie imputations metoder för heterogena dataset och deras inverkan på ett specifikt HCC dataset. Tidigare forskning har visat att de nyare, mer sofistikerade imputations metoderna presterar bättre än de mer enkla metoderna när de utvärderas med normalized root mean square error (NRMSE). I motsats till intuition, så visar resultaten i denna studie att när imputation kombineras med andra data-nivå metoder så som översampling och klustring, så påverkas inte klassificeringen alltid på ett meningsfullt sätt. Detta kan förklaras med att brus introduceras i datasetet när syntetiska punkter genereras i översampling processen. Resultaten visar också att en av de mer sofistikerade imputation metoderna, nämligen MICE, är starkt beroende på tidigare antaganden som görs om de underliggande fördelningarna i datasetet. När dessa antaganden är inkorrekta så presterar imputations metoden dåligt och har en negativ inverkan på klassificering.

Proteins and their interactions are fundamental to cellular life. Disruption of protein-protein, protein-RNA, or protein-DNA interactions can lead to disease, by affecting the function of protein complexes or by affecting gene regulation. A better understanding of these interactions on the molecular level gives rise to new methods to predict protein interaction, and is critical for the rational design of new therapeutic agents that disrupt disease-causing interactions. This thesis consists of three parts that focus on various aspects of protein interactions and their prediction in the context of disease. In the first part of this thesis, we classify interfaces of protein-protein interactions. We do so by systematically computing all binding sites between protein domains in protein complex structures solved by X-ray crystallography. The result is SCOPPI, the Structural Classification of Protein Protein Interfaces. Clustering and classification of geometrically similar interfaces reveals interesting examples comprising viral mimicry of human interface binding sites, gene fusion events, conservation of interface residues, and diversity of interface localisations. We then develop a novel method to predict protein interactions which is based on these structural interface templates from SCOPPI. The method is applied in three use cases covering osteoclast differentiation, which is relevant for osteoporosis, the microtubule-associated network in meiosis, and proteins found deregulated in pancreatic cancer. As a result, we are able to reconstruct many interactions known to the expert molecular biologist, and predict novel high confidence interactions backed up by structural or experimental evidence. These predictions can facilitate the generation of hypotheses, and provide knowledge on binding sites of promising disease-relevant candidates for targeted drug development. In the second part, we present a novel algorithm to search for protein binding sites in RNA sequences. The algorithm combines RNA structure prediction with sequence motif scanning and evolutionary conservation to identify binding sites on candidate messenger RNAs. It is used to search for binding sites of the PTBP1 protein, an important regulator of glucose secretion in the pancreatic beta cell. First, applied to a benchmark set of mRNAs known to be regulated by PTBP1, the algorithm successfully finds significant binding sites in all benchmark mRNAs. Second, collaborators carried out a screen to identify changes in the proteome of beta cells upon glucose stimulation while inhibiting gene expression. Analysing this set of post-transcriptionally controlled candidate mRNAs for PTBP1 binding, the algorithm produced a ranked list of 11 high confident potential PTBP1 binding sites. Experimental validation of predicted targets is ongoing. Overall, identifying targets of PTBP1 and hence regulators of insulin secretion may contribute to the treatment of diabetes by providing novel protein drug targets or by aiding in the design of novel RNA-binding therapeutics. The third part of this thesis deals with gene regulation in disease. One of the great challenges in medicine is to correlate genotypic data, such as gene expression measurements, and other covariates, such as age or gender, to a variety of phenotypic data from the patient. Here, we address the problem of survival prediction based on microarray data in cancer patients. To this end, a computational approach was devised to find genes in human cancer tissue samples whose expression is predictive for the survival outcome of the patient. The central idea of the approach is the incorporation of background knowledge information in form of a network, and the use of an algorithm similar to Google s PageRank. Applied to pancreas cancer, it identifies a set of eight genes that allows to predict whether a patient has a poor or good prognosis. The approach shows an accuracy comparable to studies that were performed in breast cancer or lymphatic malignancies. Yet, no such study was done for pancreatic cancer. Regulatory networks contain information of transcription factors that bind to DNA in order to regulate genes. We find that including background knowledge in form of such regulatory networks gives highest improvement on prediction accuracy compared to including protein interaction or co-expression networks. Currently, our collaborators test the eight identified genes for their predictive power for survival in an independent group of 150 patients. Under a therapeutic perspective, reliable survival prediction greatly improves the correct choice of therapy. Whereas the live expectancy of some patients might benefit from extensive therapy such as surgery and chemotherapy, for other patients this may only be a burden. Instead, for this group, a less aggressive or different treatment could result in better quality of the remaining lifetime. Conclusively, this thesis contributes novel analytical tools that provide insight into disease-relevant interactions of proteins. Furthermore, this thesis work contributes a novel algorithm to deal with noisy microarray measurements, which allows to considerably improve prediction of survival of cancer patients from gene expression data.

Proteins and their interactions are fundamental to cellular life. Disruption of protein-protein, protein-RNA, or protein-DNA interactions can lead to disease, by affecting the function of protein complexes or by affecting gene regulation. A better understanding of these interactions on the molecular level gives rise to new methods to predict protein interaction, and is critical for the rational design of new therapeutic agents that disrupt disease-causing interactions. This thesis consists of three parts that focus on various aspects of protein interactions and their prediction in the context of disease. In the first part of this thesis, we classify interfaces of protein-protein interactions. We do so by systematically computing all binding sites between protein domains in protein complex structures solved by X-ray crystallography. The result is SCOPPI, the Structural Classification of Protein Protein Interfaces. Clustering and classification of geometrically similar interfaces reveals interesting examples comprising viral mimicry of human interface binding sites, gene fusion events, conservation of interface residues, and diversity of interface localisations. We then develop a novel method to predict protein interactions which is based on these structural interface templates from SCOPPI. The method is applied in three use cases covering osteoclast differentiation, which is relevant for osteoporosis, the microtubule-associated network in meiosis, and proteins found deregulated in pancreatic cancer. As a result, we are able to reconstruct many interactions known to the expert molecular biologist, and predict novel high confidence interactions backed up by structural or experimental evidence. These predictions can facilitate the generation of hypotheses, and provide knowledge on binding sites of promising disease-relevant candidates for targeted drug development. In the second part, we present a novel algorithm to search for protein binding sites in RNA sequences. The algorithm combines RNA structure prediction with sequence motif scanning and evolutionary conservation to identify binding sites on candidate messenger RNAs. It is used to search for binding sites of the PTBP1 protein, an important regulator of glucose secretion in the pancreatic beta cell. First, applied to a benchmark set of mRNAs known to be regulated by PTBP1, the algorithm successfully finds significant binding sites in all benchmark mRNAs. Second, collaborators carried out a screen to identify changes in the proteome of beta cells upon glucose stimulation while inhibiting gene expression. Analysing this set of post-transcriptionally controlled candidate mRNAs for PTBP1 binding, the algorithm produced a ranked list of 11 high confident potential PTBP1 binding sites. Experimental validation of predicted targets is ongoing. Overall, identifying targets of PTBP1 and hence regulators of insulin secretion may contribute to the treatment of diabetes by providing novel protein drug targets or by aiding in the design of novel RNA-binding therapeutics. The third part of this thesis deals with gene regulation in disease. One of the great challenges in medicine is to correlate genotypic data, such as gene expression measurements, and other covariates, such as age or gender, to a variety of phenotypic data from the patient. Here, we address the problem of survival prediction based on microarray data in cancer patients. To this end, a computational approach was devised to find genes in human cancer tissue samples whose expression is predictive for the survival outcome of the patient. The central idea of the approach is the incorporation of background knowledge information in form of a network, and the use of an algorithm similar to Google s PageRank. Applied to pancreas cancer, it identifies a set of eight genes that allows to predict whether a patient has a poor or good prognosis. The approach shows an accuracy comparable to studies that were performed in breast cancer or lymphatic malignancies. Yet, no such study was done for pancreatic cancer. Regulatory networks contain information of transcription factors that bind to DNA in order to regulate genes. We find that including background knowledge in form of such regulatory networks gives highest improvement on prediction accuracy compared to including protein interaction or co-expression networks. Currently, our collaborators test the eight identified genes for their predictive power for survival in an independent group of 150 patients. Under a therapeutic perspective, reliable survival prediction greatly improves the correct choice of therapy. Whereas the live expectancy of some patients might benefit from extensive therapy such as surgery and chemotherapy, for other patients this may only be a burden. Instead, for this group, a less aggressive or different treatment could result in better quality of the remaining lifetime. Conclusively, this thesis contributes novel analytical tools that provide insight into disease-relevant interactions of proteins. Furthermore, this thesis work contributes a novel algorithm to deal with noisy microarray measurements, which allows to considerably improve prediction of survival of cancer patients from gene expression data.

No description available.