Optimization based clustering and classification algorithms in analysis of microarray gene expression data sets
- Authors: Mardaneh, Karim
- Date: 2007
- Type: Text , Thesis , PhD
- Full Text:
- Description: Doctor of Philosophy
- Description: Bioinformatics and computational biology are relatively new areas that involve the use of different techniques including computer science, informatics, biochemistry, applied math and etc., to solve biological problems. In recent years the development of new molecular genetics technologies, such as DNA microarrays led to the simultaneous measurement of expression levels of thousands and even tens of thousands of genes. Microarray gene expression technology has facilitated the study of genomic structure and investigation of biological systems. Numerical output of this technology is shown as microarray gene expression data sets. These data sets contain a very large number of genes and a relatively small number of samples and their precise analysis requires a robust and suitable computer software. Due to this, only a few existing algorithms are applicable to them, so more efficient methods for solving clustering, gene selection and classification problems of gene expression data sets are required and those methods need to be computationally applicable and less expensive. The aim of this thesis is to develop new algorithms for solving clustering, gene selection and data classification problems on gene expression data sets. Clustering in gene expression data sets is a challenging problem. The increasing use of DNA microarray-based tumour gene expression profiles for cancer diagnosis requires more efficient methods to solve clustering problems of these profiles. Different algorithms for clustering of genes have been proposed, however few algorithms can be applied to the clustering of samples. k-means algorithm, among very few clustering algorithms is applicable to microarray gene expression data sets, however these are not efficient for solving clustering problems when the number of genes is thousands and this algorithm is very sensitive to the choice of a starting point. Additionally, when the number of clusters is relatively large, this algorithm gives local minima which can differ significantly from the global solution. Over the last several years different approaches have been proposed to improve global ii Abstract Abstract search properties of k-means algorithm. One of them is the global k-means algorithm, however this algorithm is not efficient when data are sparse. In this thesis we developed a new version of the global k-means algorithm, the modified global k-means algorithm which is effective for solving clustering problems in gene expression data sets. In a microarray gene expression data set, in many cases only a small fraction of genes are informative whereas most of them are non-informative and make noise. Therefore the development of gene selection algorithms that allow us to remove as many non-informative genes as possible is very important. In this thesis we developed a new overlapping gene selection algorithm. This algorithm is based on calculating overlaps of different genes. It considerably reduces the number of genes and is efficient in finding a subset of informative genes. Over the last decade different approaches have been proposed to solve supervised data classification problems in gene expression data sets. In this thesis we developed a new approach which is based on the so-called max-min separability and is compared with the other approaches. The max-min separability algorithm is an equivalent of piecewise linear separability. An incremental algorithm is presented to compute piecewise linear functions separating two sets. This algorithm is applied along with a special gene selection algorithm. In this thesis, all new algorithms have been tested on 10 publicly available gene expression data sets and our numerical results demonstrate the efficiency of the new algorithms that were developed in the framework of this research
- Authors: Mardaneh, Karim
- Date: 2007
- Type: Text , Thesis , PhD
- Full Text:
- Description: Doctor of Philosophy
- Description: Bioinformatics and computational biology are relatively new areas that involve the use of different techniques including computer science, informatics, biochemistry, applied math and etc., to solve biological problems. In recent years the development of new molecular genetics technologies, such as DNA microarrays led to the simultaneous measurement of expression levels of thousands and even tens of thousands of genes. Microarray gene expression technology has facilitated the study of genomic structure and investigation of biological systems. Numerical output of this technology is shown as microarray gene expression data sets. These data sets contain a very large number of genes and a relatively small number of samples and their precise analysis requires a robust and suitable computer software. Due to this, only a few existing algorithms are applicable to them, so more efficient methods for solving clustering, gene selection and classification problems of gene expression data sets are required and those methods need to be computationally applicable and less expensive. The aim of this thesis is to develop new algorithms for solving clustering, gene selection and data classification problems on gene expression data sets. Clustering in gene expression data sets is a challenging problem. The increasing use of DNA microarray-based tumour gene expression profiles for cancer diagnosis requires more efficient methods to solve clustering problems of these profiles. Different algorithms for clustering of genes have been proposed, however few algorithms can be applied to the clustering of samples. k-means algorithm, among very few clustering algorithms is applicable to microarray gene expression data sets, however these are not efficient for solving clustering problems when the number of genes is thousands and this algorithm is very sensitive to the choice of a starting point. Additionally, when the number of clusters is relatively large, this algorithm gives local minima which can differ significantly from the global solution. Over the last several years different approaches have been proposed to improve global ii Abstract Abstract search properties of k-means algorithm. One of them is the global k-means algorithm, however this algorithm is not efficient when data are sparse. In this thesis we developed a new version of the global k-means algorithm, the modified global k-means algorithm which is effective for solving clustering problems in gene expression data sets. In a microarray gene expression data set, in many cases only a small fraction of genes are informative whereas most of them are non-informative and make noise. Therefore the development of gene selection algorithms that allow us to remove as many non-informative genes as possible is very important. In this thesis we developed a new overlapping gene selection algorithm. This algorithm is based on calculating overlaps of different genes. It considerably reduces the number of genes and is efficient in finding a subset of informative genes. Over the last decade different approaches have been proposed to solve supervised data classification problems in gene expression data sets. In this thesis we developed a new approach which is based on the so-called max-min separability and is compared with the other approaches. The max-min separability algorithm is an equivalent of piecewise linear separability. An incremental algorithm is presented to compute piecewise linear functions separating two sets. This algorithm is applied along with a special gene selection algorithm. In this thesis, all new algorithms have been tested on 10 publicly available gene expression data sets and our numerical results demonstrate the efficiency of the new algorithms that were developed in the framework of this research
Best practice data life cycle approaches for the life sciences
- Griffin, Philippa, Khadake, Jyoti, LeMay, Kate, Lewis, Suzanna, Orchard, Sandra, Pask, Andrew, Pope, Bernard, Roessner, Ute, Russell, Keith, Seemann, Torsten, Treloar, Andrew, Tyagi, Sonika, Christiansen, Jeffrey, Dayalan, Saravanan, Gladman, Simon, Hangartner, Sandra, Hayden, Helen, Ho, William, Keeble-Gagnère, Gabriel, Korhonen, Pasi, Neish, Peter, Prestes, Priscilla, Richardson, Mark, Watson-Haigh, Nathan, Wyres, Kelly, Young, Neil, Schneider, Maria
- Authors: Griffin, Philippa , Khadake, Jyoti , LeMay, Kate , Lewis, Suzanna , Orchard, Sandra , Pask, Andrew , Pope, Bernard , Roessner, Ute , Russell, Keith , Seemann, Torsten , Treloar, Andrew , Tyagi, Sonika , Christiansen, Jeffrey , Dayalan, Saravanan , Gladman, Simon , Hangartner, Sandra , Hayden, Helen , Ho, William , Keeble-Gagnère, Gabriel , Korhonen, Pasi , Neish, Peter , Prestes, Priscilla , Richardson, Mark , Watson-Haigh, Nathan , Wyres, Kelly , Young, Neil , Schneider, Maria
- Date: 2018
- Type: Text , Journal article
- Relation: F1000 Research Vol. 6, no. (2018), p. 1-28
- Full Text:
- Reviewed:
- Description: Throughout history, the life sciences have been revolutionised by technological advances; in our era this is manifested by advances in instrumentation for data generation, and consequently researchers now routinely handle large amounts of heterogeneous data in digital formats. The simultaneous transitions towards biology as a data science and towards a 'life cycle' view of research data pose new challenges. Researchers face a bewildering landscape of data management requirements, recommendations and regulations, without necessarily being able to access data management training or possessing a clear understanding of practical approaches that can assist in data management in their particular research domain. Here we provide an overview of best practice data life cycle approaches for researchers in the life sciences/bioinformatics space with a particular focus on 'omics' datasets and computer-based data processing and analysis. We discuss the different stages of the data life cycle and provide practical suggestions for useful tools and resources to improve data management practices. © 2018 Griffin PC et al.
- Authors: Griffin, Philippa , Khadake, Jyoti , LeMay, Kate , Lewis, Suzanna , Orchard, Sandra , Pask, Andrew , Pope, Bernard , Roessner, Ute , Russell, Keith , Seemann, Torsten , Treloar, Andrew , Tyagi, Sonika , Christiansen, Jeffrey , Dayalan, Saravanan , Gladman, Simon , Hangartner, Sandra , Hayden, Helen , Ho, William , Keeble-Gagnère, Gabriel , Korhonen, Pasi , Neish, Peter , Prestes, Priscilla , Richardson, Mark , Watson-Haigh, Nathan , Wyres, Kelly , Young, Neil , Schneider, Maria
- Date: 2018
- Type: Text , Journal article
- Relation: F1000 Research Vol. 6, no. (2018), p. 1-28
- Full Text:
- Reviewed:
- Description: Throughout history, the life sciences have been revolutionised by technological advances; in our era this is manifested by advances in instrumentation for data generation, and consequently researchers now routinely handle large amounts of heterogeneous data in digital formats. The simultaneous transitions towards biology as a data science and towards a 'life cycle' view of research data pose new challenges. Researchers face a bewildering landscape of data management requirements, recommendations and regulations, without necessarily being able to access data management training or possessing a clear understanding of practical approaches that can assist in data management in their particular research domain. Here we provide an overview of best practice data life cycle approaches for researchers in the life sciences/bioinformatics space with a particular focus on 'omics' datasets and computer-based data processing and analysis. We discuss the different stages of the data life cycle and provide practical suggestions for useful tools and resources to improve data management practices. © 2018 Griffin PC et al.
A review of analytical techniques and their application in disease diagnosis in breathomics and salivaomics research
- Beale, David, Jones, Oliver, Karpe, Avinash, Dayalan, Saravanan, Oh, Ding, Kouremenos, Konstantinos, Ahmed, Warish, Palombo, Enzo
- Authors: Beale, David , Jones, Oliver , Karpe, Avinash , Dayalan, Saravanan , Oh, Ding , Kouremenos, Konstantinos , Ahmed, Warish , Palombo, Enzo
- Date: 2017
- Type: Text , Journal article
- Relation: International Journal of Molecular Sciences Vol. 18, no. 1 (2017), p. 1-26
- Full Text:
- Reviewed:
- Description: The application of metabolomics to biological samples has been a key focus in systems biology research, which is aimed at the development of rapid diagnostic methods and the creation of personalized medicine. More recently, there has been a strong focus towards this approach applied to non-invasively acquired samples, such as saliva and exhaled breath. The analysis of these biological samples, in conjunction with other sample types and traditional diagnostic tests, has resulted in faster and more reliable characterization of a range of health disorders and diseases. As the sampling process involved in collecting exhaled breath and saliva is non-intrusive as well as comparatively low-cost and uses a series of widely accepted methods, it provides researchers with easy access to the metabolites secreted by the human body. Owing to its accuracy and rapid nature, metabolomic analysis of saliva and breath (known as salivaomics and breathomics, respectively) is a rapidly growing field and has shown potential to be effective in detecting and diagnosing the early stages of numerous diseases and infections in preclinical studies. This review discusses the various collection and analyses methods currently applied in two of the least used non-invasive sample types in metabolomics, specifically their application in salivaomics and breathomics research. Some of the salient research completed in this field to date is also assessed and discussed in order to provide a basis to advocate their use and possible future scientific directions. © 2016 by the authors; licensee MDPI, Basel, Switzerland.
- Authors: Beale, David , Jones, Oliver , Karpe, Avinash , Dayalan, Saravanan , Oh, Ding , Kouremenos, Konstantinos , Ahmed, Warish , Palombo, Enzo
- Date: 2017
- Type: Text , Journal article
- Relation: International Journal of Molecular Sciences Vol. 18, no. 1 (2017), p. 1-26
- Full Text:
- Reviewed:
- Description: The application of metabolomics to biological samples has been a key focus in systems biology research, which is aimed at the development of rapid diagnostic methods and the creation of personalized medicine. More recently, there has been a strong focus towards this approach applied to non-invasively acquired samples, such as saliva and exhaled breath. The analysis of these biological samples, in conjunction with other sample types and traditional diagnostic tests, has resulted in faster and more reliable characterization of a range of health disorders and diseases. As the sampling process involved in collecting exhaled breath and saliva is non-intrusive as well as comparatively low-cost and uses a series of widely accepted methods, it provides researchers with easy access to the metabolites secreted by the human body. Owing to its accuracy and rapid nature, metabolomic analysis of saliva and breath (known as salivaomics and breathomics, respectively) is a rapidly growing field and has shown potential to be effective in detecting and diagnosing the early stages of numerous diseases and infections in preclinical studies. This review discusses the various collection and analyses methods currently applied in two of the least used non-invasive sample types in metabolomics, specifically their application in salivaomics and breathomics research. Some of the salient research completed in this field to date is also assessed and discussed in order to provide a basis to advocate their use and possible future scientific directions. © 2016 by the authors; licensee MDPI, Basel, Switzerland.
Analysis of Classifiers for Prediction of Type II Diabetes Mellitus
- Barhate, Rahul, Kulkarni, Pradnya
- Authors: Barhate, Rahul , Kulkarni, Pradnya
- Date: 2018
- Type: Text , Conference proceedings , Conference paper
- Relation: 4th International Conference on Computing, Communication Control and Automation, ICCUBEA 2018
- Full Text:
- Reviewed:
- Description: Diabetes mellitus is a chronic disease and a health challenge worldwide. According to the International Diabetes Federation, 451 million people across the globe have diabetes, with this number anticipated to rise up to 693 million people by 2045. It has been shown that 80% of the complications arising from type II diabetes can be prevented or delayed by early identification of the people who are at risk. Diabetes is difficult to diagnose in the early stages as its symptoms grow subtly and gradually. In a majority of the cases, the patients remain undiagnosed until they are admitted for a heart attack or begin to lose their sight. This paper analyzes the different classification algorithms based on a patient's health history to aid doctors identify the presence of as well as promote early diagnosis and treatment. The experiments were conducted on Pima Indian Diabetes data set. Various classifiers used include K Nearest Neighbors, Logistic Regression, Decision Trees, Random Forest, Gradient Boosting, Support Vector Machine and Neural Network. Results demonstrate that Random Forests performed well on the data set giving an accuracy of 79.7%. © 2018 IEEE.
- Description: E1
- Authors: Barhate, Rahul , Kulkarni, Pradnya
- Date: 2018
- Type: Text , Conference proceedings , Conference paper
- Relation: 4th International Conference on Computing, Communication Control and Automation, ICCUBEA 2018
- Full Text:
- Reviewed:
- Description: Diabetes mellitus is a chronic disease and a health challenge worldwide. According to the International Diabetes Federation, 451 million people across the globe have diabetes, with this number anticipated to rise up to 693 million people by 2045. It has been shown that 80% of the complications arising from type II diabetes can be prevented or delayed by early identification of the people who are at risk. Diabetes is difficult to diagnose in the early stages as its symptoms grow subtly and gradually. In a majority of the cases, the patients remain undiagnosed until they are admitted for a heart attack or begin to lose their sight. This paper analyzes the different classification algorithms based on a patient's health history to aid doctors identify the presence of as well as promote early diagnosis and treatment. The experiments were conducted on Pima Indian Diabetes data set. Various classifiers used include K Nearest Neighbors, Logistic Regression, Decision Trees, Random Forest, Gradient Boosting, Support Vector Machine and Neural Network. Results demonstrate that Random Forests performed well on the data set giving an accuracy of 79.7%. © 2018 IEEE.
- Description: E1
- «
- ‹
- 1
- ›
- »