Day 1 :
Keynote Forum
En Bing Lin
Central Michigan University, USA
Keynote: Big data analysis in bioinformatics
Biography:
En-Bing Lin is a Professor of Mathematics at Central Michigan University, USA. He has been associated with several institutions including Massachusetts Institute of Technology, University of Wisconsin-Milwaukee, University of California, Riverside, University of Toledo, UCLA, and University of Illinois at Chicago. He has received his PhD from Johns Hopkins University. His research interests include Data Analysis, Applied and Computational Mathematics, and Mathematical Physics. He has Supervised a number of graduate and undergraduate students. He serves on the Editorial Boards of several journals. He has organized several special sessions at regional IEEE conferences and many other professional meetings
Abstract:
With the increasing use of advanced technology and the exploding amount of data in bioinformatics, it is imperative to introduce effective and efficient methods to handle Big data using the distributed and parallel computing technologies. Big data analytics can examine large data sets, analyze and correlate genomic and proteomic information. In this presentation, we begin with an overview of Big data and Big data analytics, we then address several challenging and important tasks in bioinformatics such as analyzing coding, noncoding regions and finding similarities for coding and noncoding regions as well as many other issues. We further study mutual information-based gene or feature selection method where features are wavelet-based; the bootstrap techniques employed to obtain an accurate estimate of the mutual information and other new methods to analyze data. Given the multi-scale structure of most biological data, several methods will be presented to achieve improvements in the quality of mathematical or statistical analysis of such data. In a DNA strand, it is essential to find sequences, which can be transcribed to complementary parts of the DNA strand. We will mention several methods to identify protein coding regions. We also use some special variance and entropy to analyze similarities among coding and noncoding regions of several DNA sequences respectively and compare the resulting data. We will address the use of big data analytics in many phases of the bioinformatics analysis pipeline
Keynote Forum
Abdel-Salam Gomaa
Qatar University, Qatar
Keynote: Model robust profile monitoring for the generalized linear mixed model for phase I analysis
Biography:
Abdel-Salam Gomaa holds BS and MS (2004) degrees in Statistics from Cairo University and MS (2006) and PhD (2009) degrees in Statistics from Virginia Polytechnic Institute and State University (Virginia Tech, USA). Prior to joining Qatar University as an Assistant Professor and a Coordinator of the Statistical Consulting Unit and Coordinator for the Statistics Programs, he taught at Faculty of Economics and Political Science (Cairo University), Virginia Tech, and Oklahoma State University. Also, he worked at JPMorgan Chase Co. as Assistant Vice President in Mortgage Banking and Business Banking Risk Management Sectors. He has published several research papers and delivered numerous talks and workshops. He has awarded couples of the highest prestige awards such as Teaching Excellence from Virginia Tech, Academic Excellence Award, Freud International Award, and Mary G Natrella Scholarship from American Statistical Association (ASA) and American Society for Quality (ASQ), for outstanding graduate study of the theory and application of Quality Control, Quality Assurance, Quality Improvement, and Total Quality Management. He is a Member of the ASQ and ASA
Abstract:
There are so many applications for detecting the changes in the relationship between the response variable and explanatory variable (s) may be the most important consideration rather than detecting the changes in univariate or multivariate quality characteristics. This relationship between the response variable and one or more explanatory variables is called a profile. The act of using various techniques to statistically monitor the process or product profiles is known as profile monitoring. The study introduces two mixed model methods to monitor profiles from the exponential family: a nonparametric (NP) regression method based on the penalized spline regression technique and a semi-parametric (SP) method (Model robust profile monitoring for the generalized linear mixed model (MRGLMM)) which combines the advantages of both the parametric and nonparametric methods. A correctly specified parametric (P) model will have the most power in detecting the profile shift, while a NP method can give improved performance for any type of profile. The MRGLMM method gives results similar to the P method when P model is correctly specified and it gives results similar to the NP method if the proposed P model is badly misspecified. The MRGLMM method gives results that are superior to either the P method or the NP method if the P model provides some useful information regarding profile behavior. Thus, the MRGLMM method is robust to model misspecification. The performances of P, NP and MRGLMM methods are compared through a simulation study using binary data
Keynote Forum
Hong Lin
University of Houston-Downtown, USA
Keynote: Information technologies: opportunities and challenges in personal healthcare systems
Biography:
Hong Lin was a Postdoctoral Research Associate at Purdue University; an Assistant Research Officer at the National Research Council, Canada, and a Software Engineer at Nokia, Inc. He is currently a Professor with UHD. His research interests include human-centered computing, parallel/distributed computing, grid computing, multi-agent systems, and high level computational models. He is a Co-supervisor of the Grid Computing Lab at UHD
Abstract:
The well-being of a person consists of 2 aspects: the physical body well-being and the mind well-being (the perception or the feeling of well-being). Technology development makes it possible to massively produce cheap sensors for personal use. The data collected, if being properly analyzed, can provide objective and comprehensive personal health information. The information helps us to understand the well-being of the person and then further offers the opportunity to develop a high quality personal healthcare system for the well-being of the person. In this talk, I will report our preliminary findings in applying modern information technology to personal healthcare systems. We construct a brain activity level model by using EEG signals to objectively measure the effectiveness of meditation, detect mental fatigue and boredom, and comprehend human emotions. Also, we have used accelerometer and GPS data to assess sports performance and training enhancement, leg muscle injury prevention and recovery monitoring, and fall prevention for aged people. In addition, the ubiquitous nature of accelerometer and GPS technology make it possible to deliver personal healthcare services for people in physical excise. Then, we exploit the potential of Kinect device in monitoring the movements of aged persons in their houses to prevent falls. Finally, we point out some remaining challenges and possible opportunities in using information technologies to deliver personal health care
- Clinical Biostatistics | Statistical Methods | Bayesian statistics | Biostatistics applications | Big Data Analytics | Structural Bioinformatics | Systems Biology in Bioinformatics | Regression Analysis
Chair
Abdel-Salam Gomaa
Qatar University, Qatar
Co-Chair
Meenakshi Nadimpalli
Reliable Software Resources Inc., USA
Session Introduction
Pratool Bharti
University of South Florida, USA
Title: HuMAn: complex activity recognition with multi-modal multi-positional body sensing
Biography:
Pratool Bharti is pursuing his PhD at University of South Florida, Tampa. He is also a Graduate Student Ambassador for artificial intelligence at Intel corp. He has completed his undergraduate in Computer Science and Engineering from Kalyani Government Engineering College, India in 2010. Before starting his PhD in 2014, he has worked 4 years as Software Developer in machine learning. His current research is in finding pervasive solution for healthcare problems in society. He has published multiple journals and conference papers on activity recognition and smart healthcare
Abstract:
Current state-of-the-art systems in the literature using wearables are not capable of distinguishing many fine-grained and/or complex human activities, which may appear similar but with vital differences in context, such as lying on floor vs. lying on bed vs. lying on sofa. This paper fills this gap by proposing a novel system, called HuMAn, that recognizes and classifies complex at-home activities of humans with wearable sensing. Specifically, HuMAn makes such classifications feasible by leveraging selective multi-modal sensor suites from wearable devices, and enhances the richness of sensed information for activity classification by carefully leveraging placement of the wearable devices across multiple positions on the human body. The HuMAn system consists of the following components: Practical feature set extraction from specifically selected multi-modal sensor suites; a novel two-level structured classification algorithm that improves accuracy by leveraging sensors in multiple body positions; and improved refinement in classification of complex activities with minimal external infrastructure support (e.g., only a few Bluetooth beacons used for location context). The proposed system is evaluated with 10 users in real home environments. Experimental results demonstrate that the HuMAn can detect a total of 21 complex at-home activities with high degree of accuracy. For same-user evaluation strategy, the average activity classification accuracy is as high as 97% over all the 21 activities. For the case of 10-fold cross- validation evaluation strategy, the average classification accuracy is 94%, and for the case of leave-one-out cross-validation strategy, the average classification accuracy is 76%.
Rashid Ahmed
University of North Dakota, USA
Title: Cost analysis of metastatic non-small cell lung cancer (NSCLC) after completion of chemotherapy with the introduction of erlotinib
Biography:
Rashid Ahmed has completed his PhD in Bio-Statistics from University of Waterloo; ON; Canada. He has strong background in epidemiology and biostatistics and has experience in the development of statistical methods for the design of community-based interventions and the analysis of longitudinal health data. Currently, he is developing diagnostic measures for joint models for longitudinal and survival data in the presence of non-ignorable missing data. Currently, he is working as an Associate Dean for Research with the College of Nursing and Professional Disciplines at the University of North Dakota. He has published more than 30 papers in reputed journals and has been serving as reviewer of several journals
Abstract:
Erlotinib has been funded for use as a second and third line treatment of advanced NSCLC since 2006 in Manitoba, Canada. Prior research examined, the cost effectiveness of Erlotinib across various countries, yet results may not be generalizable to the healthcare system within Canada. The present paper outlines the cost effectiveness of Erlotinib by examining population-based total costs, service utilization, and clinical outcomes of patients with metastatic NSCLC receiving Erlotinib were explored from the time they completed chemotherapy till the end of follow-up date (31 Dec 2012), death, or relocation. Metastatic NSCLC patients, who were approved for Erlotinib in Manitoba between June 2006 and 31 Dec 2012, were selected. The Manitoba Cancer Registry (MCR) and chart review were used to capture the information on treatment and clinical outcome. Service utilization information and direct cost information were extracted from the electronic health records known as ARIA, the Physicians Claims, the Hospital Discharge database, and the Drug Program Information Network. Survival rate, using the Kaplan-Meier method, was calculated from the date of the Erlotinib approval date till end of follow-up date, Dec 31, 2012, or death. The median survival rate was 40.1 weeks. The average cost per patient was CA$ 30,503 and CA$ 987 per patient-week for patients who received Erlotinib. Ninety percent of the cost was accounted for by hospital stays and drug costs. General demographic patterns per cost suggested that current smokers tended to incur higher costs compared to non-smokers. The results of this study appear to replicate patterns of other studies examining the cost effectiveness such that Erlotinib appears to have a high survival rate paired with lower costs related to side effect management
Owen P L Mtambo
Namibia University of Science and Technology, Namibia
Title: Spatio-temporal quantile interval regression using R-INLA with applications to childhood overweight and obesity in Sub-Saharan Africa
Biography:
Owen P L Mtambo is a Lecturer in Statistics at Namibia University of Science and Technology since March 2014. He was a Lecturer in Statistics at University of Malawi from July 2007 to March 2014. He was Secondary School Teacher in Mathematics and Physics from January 2002 to June 2007. He holds an MSc (Biostatistics) with credit (2012), a BSc (Mathematical Sciences) with distinction (2007), and a DipEdu (Sciences) with distinction (2001); all obtained from University of Malawi. He has more than 8 publications. Currently, he is pursuing his PhD in Statistics at University of South Africa (UNISA) since February 2016.
Abstract:
Childhood malnutrition has serious adverse effects on a child, a family and the development of a country. It leads to more than 30% of deaths in children below five years in sub-Saharan African countries. A malnourished child is more likely to be sick and die. Malnutrition can lead to stunted growth, overweight and obesity, impaired cognitive and behavior development, poor school performance, lower working capacity and lower income. It can slow down economic growth and increase level of poverty. Furthermore, it can prevent a society or a nation from meeting its full potential through loss in productivity, cognitive capacity and increased cost in health care. The indicators of malnutrition range from stunting, wasting and underweight to overweight and obesity. In the past, childhood undernutrition was used to be the most malnutrition burden over the past two decades across the sub-Saharan Africa and is still remaining a burden to date. Doubly surprising, childhood overnutrition is alarmingly becoming the most prevalent parallel to still existing undernutrition burden in sub-Saharan Africa. Overweight and obesity rates are reaching epic proportions in sub-Saharan Africa. The prevalence of childhood overweight and obesity in sub-Saharan Africa was 8.5% in 2010 and is expected to reach 12.7% by 2020. The consequences of overnutrition can be more devastating than those for undernutrition because it leads to chronic failure problems which in turn lead to increased medical expenditure. For this reason, only childhood overweight and obesity were analyzed in this study, in order to assess socio-demographic and socio-economic determinants of childhood overweight and obesity in sub-Saharan Africa. This study also assessed the geographical variation of childhood overweight and obesity in sub-Saharan Africa with more emphasis on both spatial and spatio-temporal effects. All available Demographic and Health Survey (DHS) datasets since 2000 were used and the statistical inference was fully Bayesian using R-INLA package in the selected countries. Almost all studies on spatial quantile modelling of childhood overweight and obesity have emphasized on selecting only one specific response quantile level of interest and report the recommendations based on the only chosen response quantile. Unlike mean response modelling, quantile regression yields model estimates which are stochastic functions of quantile levels such that. This implies that quantile regression modelling using estimates based on only one chosen quantile level might be inefficient and not robust enough. In this study, we used weighted mean estimates based on all quantiles in the quantile interval which corresponded to modelling childhood overweight and obesity. We found out that the significant determinants of childhood overweight and obesity ranged from socio-demographic factors such as type of residence to child and maternal factors such as child age, duration of breastfeeding and maternal BMI. We also observed significant positive structured spatial effects on childhood overweight and obesity mostly in the regions in the center of Namibia.
Biography:
Abdul Basit is the PhD Research Scholar in the discipline of Statistics in National College of Business Administration and Economics Lahore, Pakistan. He has completed his MS in Social Sciences from SZABIST Karachi, Pakistan in 2014. Currently, he is serving as Deputy Director in Research Cluster of State Bank of Pakistan. He has published 07 research papers in journals and many articles were presented in national and international conferences
Abstract:
Survival function and hazard rate are very informative and reliable characteristics of any distribution. Entropy is a tool to measure the maximum information of any distribution. In this study, we use the hazard function to develop a new entropy measure. We also introduce the modified forms of Renyi, Tsallis and Havrda and Charvat entropy. The main properties and characteristics and applications associated with these modified entropies are established. We made a comparison between modified entropies by applying them on health indicators Infants Mortality Rate, Crude Death Rate and Life Expectancy of Pakistan. New entropy measures are very useful for censored data. We also introduced a new methodology for the comparison of entropy measures
Mohamed Yusuf Hassan
UAE University, UAE
Title: Integrated quadrbly reduced additive weighted mixture poisson distribution
Biography:
Mohamed Yusuf Hassan has completed his PhD from University of California, Riverside. He is the first Author of bivarivarite Mixture Transition Distribution and the Bimodal Skew-Symmetric Normal
Abstract:
We propose a new family of probability distributions derived from the mixtures of weighted Poisson probability distributions. This family of distributions will overcome some of the potential limitations suffered by the existing dominant distributions including lack of modeling over-dispersion, under-dispersion and bimodality. The family is flexible and could be applied to a variety of problems from different disciplines like business, finance, medicine and reliability in engineering. To construct this family, a baseline Poisson distribution and a number of reasonable weights are chosen to get weighted versions of the baseline distribution. These distributions are combined into a mixture format that sums up to a single component probability distribution in a closed form. The resulted density function will be parsimonious and flexible and will be able to capture the nature of many count data patterns. Real count data will be used and goodness-of-fit statistical techniques will be developed to compare its performance with the other existing competing models
- Video Presentation
Session Introduction
Aluko O S
University of KwaZulu-Natal, South Africa
Title: Statistical methodologies for handling ordinal longitudinal responses with monotone dropout patterns using multiple imputation
Biography:
Aluko O S is pursuing his PhD at the University of KwaZulu-Natal, South Africa. Three of his papers are in review under reputable journals and the fourth paper is about to be sent to another journal for publication. As a matter of fact, he use and write codes in both R and SAS softwares conveniently
Abstract:
Missing data are common challenge in any longitudinal clinical trials. Multiple imputation is one of the modern methods of handling incomplete data. This approach is applicable to different missing data patterns but sometimes faced with complexity of the type of variables to be imputed and the mechanism underlying the missing values. In this study, we compare the performance of three methods under multiple imputation, namely expectation maximization, fully conditional specification and multivariate normal imputation in the presence of ordinal responses with monotone dropout. We proposed and demonstrated the usefulness of the ordinal negative binomial distribution for ordinal data generation through simulation studies and implementation. However, the real dataset application and simulation studies reveal that the three methods perform equivalently well, thus any of the methods can be recommended for use