Konstantina Kourou

Machine learning applications in cancer prognosis and prediction.

Cancer has been characterized as a heterogeneous disease consisting of many different subtypes. The early diagnosis and prognosis of a cancer type have become a necessity in cancer research, as it can facilitate the subsequent clinical management of patients. The importance of classifying cancer patients into high or low risk groups has led many research teams, from the biomedical and the bioinformatics field, to study the application of machine learning (ML) methods. Therefore, these techniques have been utilized as an aim to model the progression and treatment of cancerous conditions. In addition, the ability of ML tools to detect key features from complex datasets reveals their importance. A variety of these techniques, including Artificial Neural Networks (ANNs), Bayesian Networks (BNs), Support Vector Machines (SVMs) and Decision Trees (DTs) have been widely applied in cancer research for the development of predictive models, resulting in effective and accurate decision making. Even though it is evident that the use of ML methods can improve our understanding of cancer progression, an appropriate level of validation is needed in order for these methods to be considered in the everyday clinical practice. In this work, we present a review of recent ML approaches employed in the modeling of cancer progression. The predictive models discussed here are based on various supervised ML techniques as well as on different input features and data samples. Given the growing trend on the application of ML methods in cancer research, we present here the most recent publications that employ these techniques as an aim to model cancer risk or patient outcomes.

Prediction of time dependent survival in HF patients after VAD implantation using pre- and post-operative data

Heart failure is one of the most common diseases worldwide. In recent years, Ventricular Assist Devices (VADs) have become a valuable option for patients with advanced HF. Although it has been shown that VADs improve patient survival rates, several complications persist during left VAD (LVAD) support. The stratification scores currently employed are mostly generic, i.e. not specifically built for LVAD patients, and are based on pre-implantation patient data. In this work we apply data mining approaches for the prediction of time dependent survival in patients after LVAD implantation. Moreover, the predictions acquired with the use of pre-implantation data are enriched by employing post-implantation data, i.e. follow-up data. Different clinical scenarios have been depicted and the subsequent conditions are tested in order to identify the optimal set of pre- and post-implant features, as well as the most suitable algorithms for feature selection and prediction. The proposed approach is applied to a real dataset of 71 patients, reporting an accuracy of 84.5%, sensitivity of 87% and specificity of 82%. Based on the reported results, expert cardio-surgeons can be supported in planning the treatment of VAD patients.

A Bayesian Network-based approach for discovering oral cancer candidate biomarkers.

Oral cancer can arise in the head and neck region. Due to the aggressive nature of the disease, which often leads to poor prognosis, Oral Squamous Cell Carcinoma (OSCC) constitutes the 8th most common neoplasms in humans. In the present work we formulate gene interaction network from oral cancer genomic data using Dynamic Bayesian Networks (DBNs). Four modules were extracted after applying a clustering technique to the network. We consequently explore them by applying topological and functional analysis methods in order to identify significant network nodes. Our analysis revealed that these important nodes may correspond to candidate biomarkers of the disease.

Prediction of oral cancer recurrence using dynamic Bayesian networks

We propose a methodology for predicting oral cancer recurrence using Dynamic Bayesian Networks. The methodology takes into consideration time series gene expression data collected at the follow-up study of patients that had or had not suffered a disease relapse. Based on that knowledge, our aim is to infer the corresponding dynamic Bayesian networks and subsequently conjecture about the causal relationships among genes within the same time-slice and between consecutive time-slices. Moreover, the proposed methodology aims to (i) assess the prognosis of patients regarding oral cancer recurrence and at the same time, (ii) provide important information about the underlying biological processes of the disease.We propose a methodology for predicting oral cancer recurrence using Dynamic Bayesian Networks. The methodology takes into consideration time series gene expression data collected at the follow-up study of patients that had or had not suffered a disease relapse. Based on that knowledge, our aim is to infer the corresponding dynamic Bayesian networks and subsequently conjecture about the causal relationships among genes within the same time-slice and between consecutive time-slices. Moreover, the proposed methodology aims to (i) assess the prognosis of patients regarding oral cancer recurrence and at the same time, (ii) provide important information about the underlying biological processes of the disease.

Gene-based pathway enrichment analysis of oral squamous cell carcinoma patients

Oral Cancer has been characterized as a complex disease and involves dynamic genomic changes. These changes reveal the ability to explore the interactions of the molecules and of differentially expressed genes that are involved in cancer progression. Moreover, based on this knowledge the identification of differentially expressed genes and pathways is of great importance. The last decade pathway analysis has become a first choice in order to study the underlying biology of differentially expressed genes associated with a disease. In the present study we exploit genes that have been characterized as oral cancer risk associated in order to further perform pathway enrichment analysis. According to our results we found significant pathways in which the disease associated genes have been identified as strongly enriched.

Modeling Biological Data Through Dynamic Bayesian Networks for Oral Squamous Cell Carcinoma Classification

We propose a computational approach for modeling the progression of Oral Squamous Cell Carcinoma (OSCC) through Dynamic Bayesian Network (DBN) models. RNA-Seq transcriptomics data, available from public functional genomics data repositories, are exploited to find genes related to disease progression (i.e. recurrence or no recurrence). Our primary aim is to perform a computational analysis based on the differentially expressed genes identified. More specifically, a search for putative transcription factor binding sites (TFBSs), in the promoters of the input gene set, as well as an analysis of the pathways of the suggested transcription factors is conducted. Activities of transcription factors which are regulated by upstream signaling cascades are further discovered. These activities converge in certain nodes, representing molecules which are potential regulators of OSCC progression. The resulting gene list is further exploited for the inference of their causal relationships and for disease classification in terms of DBN models. The structure and the parameters of the models are defined subsequently, revealing the changes in gene-gene interactions with reference to disease recurrence after surgery. The objectives of the proposed methodology are to: (i) accurately estimate OSCC progression, and (ii) provide better insights into the regulatory mechanisms of the disease. Moreover, we can conjecture about the interactions among genes based on the inferred network models. The proposed approach implies that the resulting regulatory molecules along with the differentially expressed genes extracted, can be considered as new targets, and are candidates for further experimental and in silico validation.

Identification of differentially expressed genes through a meta-analysis approach for oral cancer classification

We propose a meta-analysis scheme for identifying differentially expressed genes in Oral Squamous Cell Carcinoma (OSCC) from different microarray studies. We detect a subset of relevant features and further classify samples under two experimental conditions (i.e healthy and cancer samples) for better patient stratification. A well-established meta-analysis method is adopted and gene expression data sets are derived from a public functional genomics data repository. Our primary aim is the accurate identification of up- and down-regulated genes in order to extract valuable biological information concerning the changes in expression between healthy and cancer samples. According to our results and the extracted informative gene list, a high classification accuracy of healthy and OSCC tumors is achieved with as few genes as possible. Furthermore, the proposed scheme implies that the combination of datasets from different origins may reduce the estimated percentage of false predictions, while the power of gene identification and disease classification is increased.

A computational pipeline for deciphering the molecular mechanisms of oral cancer progression

In this work we present a computational pipeline in order to further study the molecular mechanisms underlying the progression of Oral Squamous Cell Carcinoma. Microarray gene expression data of oral cancer patients are exploited aiming to identify the differentially expressed genes between relapse and non-relapse samples. Furthermore, gene ontology analysis of the significant genes is performed and the associated functional groups are annotated. Thus, we are able to interpret the differential expression results in terms of biological processes, molecular functions and cellular components. In addition, pathway enrichment analysis is also considered based on a curated database of molecular pathways. According to our analysis results, genes that exhibit remarkable changes in expression, such as the YES1 oncogene has been predicted to interact closely with genes implicated in oral cancer. Concerning the pathway enrichment analysis, Ribosome and Tyrosine metabolism pathways are identified and should be considered for further analysis.

Integration of Pathway Knowledge and Dynamic Bayesian Networks for the Prediction of Oral Cancer Recurrence

Oral squamous cell carcinoma has been characterized as a complex disease which involves dynamic genomic changes at the molecular level. These changes indicate the worth to explore the interactions of the molecules and especially of differentially expressed genes that contribute to cancer progression. Moreover, based on this knowledge the identification of differentially expressed genes and related molecular pathways is of great importance. In the present study, we exploit differentially expressed genes in order to further perform pathway enrichment analysis. According to our results we found significant pathways in which the disease associated genes have been identified as strongly enriched. Furthermore, based on the results of the pathway enrichment analysis we propose a methodology for predicting oral cancer recurrence using dynamic Bayesian networks. The methodology takes into consideration time series gene expression data in order to predict a disease recurrence. Subsequently, we are able to conjecture about the causal interactions between genes in consecutive time intervals. Concerning the performance of the predictive models, the overall accuracy of the algorithm is 81.8% and the area under the ROC curve 89.2% regarding the knowledge from the overrepresented pre-NOTCH Expression and processing pathway.

Sequence Patterns Mediating Functions of Disordered Proteins

Disordered proteins lack specific 3D structure in their native state and have been implicated with numerous cellular functions as well as with the induction of severe diseases, e.g., cardiovascular and neurodegenerative diseases as well as diabetes. Due to their conformational flexibility they are often found to interact with a multitude of protein molecules; this one-to-many interaction which is vital for their versatile functioning involves short consensus protein sequences, which are normally detected using slow and cumbersome experimental procedures. In this work we exploit information from disorder-oriented protein interaction networks focused specifically on humans, in order to assemble, by means of overrepresentation, a set of sequence patterns that mediate the functioning of disordered proteins; hence, we are able to identify how a single protein achieves such functional promiscuity. Next, we study the sequential characteristics of the extracted patterns, which exhibit a striking preference towards a very limited subset of amino acids; specifically, residues leucine, glutamic acid, and serine are particularly frequent among the extracted patterns, and we also observe a nontrivial propensity towards alanine and glycine. Furthermore, based on the extracted patterns we set off to infer potential functional implications in order to verify our findings and potentially further extrapolate our knowledge regarding the functioning of disordered proteins. We observe that the extracted patterns are primarily involved with regulation, binding and posttranslational modifications, which constitute the most prominent functions of disordered proteins.