Texas A&M
Science of Information Days

Abstract:

In this tutorial, we will provide an overview of various engineering approaches in genomics and molecular biology research. We will show how theories, models, and techniques from information science - such as signal processing, information theory, control theory, and statistics - can be used to analyze large-scale biological data and to study complex biological systems. The tutorial will begin with a brief review of basic molecular biology that is needed to understand the research problems that arise in biology. This will be followed by an overview of several active areas of engineering research in biology, including: (i) biological network modeling, inference, simulation, and intervention; (ii) modeling, analysis, and prediction of biological molecules and their structures; (iii) comparative analysis of biological networks and prediction of functional modules, (iv) gene expression data analysis and integrated analysis of heterogeneous measurement data.

Note that this tutorial is mainly designed for engineers and information scientists who are interested in learning about the challenges and opportunities in computational biology research and that no prior knowledge of molecular biology will be assumed.

Abstract:

Biological data sets are characterized by the fact that the number of measured features is two or three orders of magnitude higher than the number of samples for which such measurements are available. In this talk two new algorithms will be presented for sparse feature selection, one for binary classification, and the other for regression (curve-fitting). These are applied respectively to predicting the risk of lymph node metastasis in endometrial cancer patients, and the IC50 values of natural compounds applied to lung cancer cell lines.

Biography:

Mathukumalli Vidyasagar was born in Guntur, India on September 29, 1947. He received all of his degrees in electrical engineering from the University of Wisconsin, Madison, WI, including the Ph.D. in 1969. He spent the next twenty years mostly in Canada before returning to India in 1989 to set up two R&D laboratories from scratch, first under the Government of India and then in the private sector under Tata Consultancy Services, India's largest IT company. In 2009 he retired from his job in TCS and joined the University of Texas at Dallas as Cecil & Ida Green in Systems Biology Science, a position that he continues to hold. He is the author of about 140 journal articles and 12 books. He has received a number of honors in recognition of his research, including the IEEE Control Systems Award and the Fellowship of The Royal Society, the oldest scientific society in the world.

Abstract:

When designing an operator on a system, say, a classifier on a feature-label distribution, a filter on a stochastic process, or a controller on a regulatory network, there are three possibilities regarding ones state of knowledge regarding the system: full knowledge, no knowledge, or partial knowledge (incomplete or inconsistent). With full knowledge, one analytically derives the optimal operator; with no knowledge, one must rely solely on data; and with partial knowledge, one utilizes prior knowledge alone or in conjunction with data, hopefully, in an optimal manner. With complex systems, rarely are we sufficiently knowledgeable of the system to proceed with direct derivation and in many areas of application, such as genomics, rarely do we possess sufficient data to rely fully upon it. This talk discusses the limitations of data-driven design and the use of knowledge-based design for classification, filtering, and intervention in gene regulatory networks. It takes a mile-high approach so that it can focus on the unity of ideas across a wide range of applications.

Biography:

Edward R. Dougherty is a Distinguished Professor in the Department of Electrical and Computer Engineering at Texas A&M University in College Station, TX, where he holds the Robert M. Kennedy '26 Chair in Electrical Engineering and is Director of the Center for Bioinformatics and Genomic Systems Engineering. He holds a Ph.D. in mathematics from Rutgers University and an M.S. in Computer Science from Stevens Institute of Technology, and has been awarded the Doctor Honoris Causa by the Tampere University of Technology. He is a fellow of both IEEE and SPIE, has received the SPIE President's Award, and served as the editor of the SPIE/IS&T Journal of Electronic Imaging. At Texas A&M University he has received the Association of Former Students Distinguished Achievement Award in Research, been named Fellow of the Texas Engineering Experiment Station and Halliburton Professor of the Dwight Look College of Engineering. Prof. Dougherty is author of 16 books and author of more than 300 journal papers.

Abstract:

Complex networks of interacting components are ubiquitous across many scientific and engineering disciplines. By employing Leaky Markovian Networks, a broad class of complex network models, we present an information-theoretic approach for their analysis which employs a potential energy landscape perspective coupled with a macroscopic description based on statistical thermodynamics. The proposed approach reveals the emergence of avalanching, a critical behavior in Leaky Markovian Networks, which dominates system dynamics at near-critical or sub-critical network sizes. We attribute this behavior to a reallocation of the ground states of the system's stationary potential energy landscape caused by a noise-induced deformation of its topographic surface. We illustrate our approach using an epidemiological model of bacterial infection and a computational model of brain activity in neural networks. Our work demonstrates that avalanching is a spontaneous behavior brought about by a phenomenon reminiscent to that of phase transition in statistical mechanics.

Biography:

John Goutsias received his PhD from the University of Southern California, Los Angeles. Since 1986, he has been a Professor of Electrical and Computer Engineering, Biomedical Engineering, and Applied Mathematics and Statistics in the Whitaker Biomedical Engineering Institute at The Johns Hopkins University. His current research interests include the development of statistical techniques for the modeling and analysis of complex reaction networks with applications to problems related to biological networks.

Abstract:

Cancer cells acquire genetic and epigenetic alterations that often lead to dysregulation of oncogenic signal transduction pathways, which in turn alters downstream transcriptional programs and produces widespread gene expression changes. Inferring aberrant signaling pathways from gene expression data alone has proved to be challenging. Here we present a statistical method to link upstream signaling to downstream transcriptional response by exploiting reverse phase protein array (RPPA) and mRNA expression data in The Cancer Genome Atlas (TCGA) breast cancer project. Formally, we use an algorithm called affinity regression – a regularized bilinear regression approach – to learn an interaction matrix between upstream signal transduction proteins and downstream transcription factors (TFs) that explains target gene expression. The trained model can then predict the TF activity given a tumor sample's protein expression profile or infer the signaling protein activity given a tumor sample's gene expression profile. We further show how to use the model to identify dysregulated signaling pathways associated with transcriptomic subtypes, therapeutic response, and survival in breast cancer.

Biography:

Christina Leslie did her undergraduate degree in Pure and Applied Mathematics at the University of Waterloo in Canada. She won an NSERC1967 Science and Engineering Fellowship to pursue her PhD in Mathematics at the University of California, Berkeley, where she studied differential geometry and representation theory and graduated in 1998. She was then awarded an NSERC Postdoctoral Fellowship and did her postdoctoral training in the Mathematics Department at Columbia University in 1999-2000. In 2000, she took a lecturer position in the Computer Science Department at Columbia University, where she began a research transition to computational biology and machine learning. She moved in 2003 to the Center for Computational Learning Systems, a new machine learning research center in the School of Engineering at Columbia University, where she was a research scientist and principal investigator leading the Computational Biology Group. In 2007, she moved her lab to the Computational Biology program of Memorial Sloan-Kettering Cancer Center, where she is currently an Associate Member of the Sloan-Kettering Institute and an Associate Professor of Physiology, Biophysics, and Systems Biology at Cornell Weill Medical College.

Setty M, Helmy K, Khan AA, Silber J, Arvey A, Neezen F, Agius P, Huse JT, Holland EC, Leslie CS. Inferring transcriptional and microRNA-mediated regulatory programs in glioblastoma. Mol Syst Biol. 2012;8:605. doi: 10.1038/msb.2012.37. View PDF

Kundaje A, Xin X, Lan C, Lianoglou S, Zhou M, Zhang L, Leslie C. A predictive model of the oxygen and heme regulatory network in yeast. PLoS Comput Biol. 2008 Nov;4(11):e1000224. doi: 10.1371/journal.pcbi.1000224. Epub 2008 Nov 14. View PDF

Abstract:

We propose Bayesian methods for estimating the precision matrix in Gaussian graphical models. The methods lead to sparse and adaptively shrunk estimators of the precision matrix, and thus conduct model selection and estimation simultaneously. Our methods are based on selection priors leading to parsimonious parameterization of the precision (inverse covariance) matrix, which is essential in several applications involving learning relationships among the variables. We introduce a novel type of selection prior that develops a sparse structure of the precision matrix by making most of the elements exactly zero, in addition to ensuring positive definiteness. We extend our methods for network based classification problems. We apply our methods to different genomics and proteomics data.

Biography:

Bani K. Mallick is a Distinguished Professor and the Director of the Center for Statistical Bioinformatics in Department of Statistics at Texas A&M University in College Station.Dr. Mallick is well known for his contribution to the theory and practice of Bayesian Semiparametric methods and Uncertainty Quantification. Mallick is an elected fellow of American Association for the Advancement of Science,[3] American Statistical Association, Institute of Mathematical Statistics, International Statistical Institute and the Royal Statistical Society. He received the Distinguished research award from Texas A&M University and the Young Researcher award from the International Indian Statistical Association. He has coauthored or co-edited six books and more than 150 research publications. He has supervised and mentored 25 Ph.D students and 6 Post doctorate fellows. Mallick earned his undergraduate from the Presidency University, Kolkata, MS from the Calcutta University and Ph.D from the University of Connecticut.

Abstract:

Dr. Coleman will discuss recent efforts by his research group in combining flexible skin-mounted wireless electronics with multi-modal analytics algorithms that can be efficiently implemented in the cloud. This results in a suite of tools to transform "big data" into "small, relevant" data and can be passed to the clinician remotely, or transformed into decision support tools to allow the user to make better health decisions. The end result is a suite of human-computer interface applications that on the one hand, blur the distinction between man and machine, while on the other, promote humans and computers playing to their individual strengths. The development of new quantitative approaches based upon constraints due to this unique synergy of sensors, analytics, and medicine will be emphasized. Specifically, (a) the development of new signaling schemes due to energy/body-integration constraints will be shown. In addition, (b) the development of robust distributed "cloud-based" statistical inference algorithms due to dynamic multi-modal physiological, and human visualization constraints will be shown. Some examples will be provided within the context of perinatal health and chronic disease management.

Biography:

Todd P. Coleman received B.S. degrees in electrical engineering (summa cum laude), as well as computer engineering (summa cum laude) from the University of Michigan. He received M.S. and Ph.D. degrees from MIT in electrical engineering, and did postdoctoral studies at MIT in neuroscience. From 2006-2011, he was an Assistant Professor of ECE and Neuroscience at the University of Illinois. He is currently an Associate Professor in Bioengineering at UCSD, where he directs the Neural Interaction Laboratory and co-Directs the Center for Perinatal Health. His research is highly inter-disciplinary, lying at the intersection of bio-electronics, medicine, and machine learning. He is conducting research in wearable health by wedding his research group's expertise in large-scale analytics with their recent development of "epidermal electronics", featured in Science in 2011. Current applications of those synergies include perinatal health, chronic disease management, and cognitive monitoring during aging. Dr. Coleman is a science advisor for the Science & Entertainment Exchange (National Academy of Sciences) and his research has been featured on CNN, BBC, and the New York Times.

J. A. Fan, W. Yeo, Y. Su, Y. Hattori, W. Lee, S. Jung, Y. Zhang, Z. Liu, H. Cheng, L. Falgout, M. Bajema, T. Coleman, D. Gregoire, R. J. Larsen, Y. Huang, and J. A. Rogers, "Fractal design concepts for stretchable electronics", Nature Communications, February 2014 http://coleman.ucsd.edu/wp-content/uploads/2014/02/FanEtAlFractalNatureComm.pdf

C. Quinn, N. Kiyavash, and T. P. Coleman "Efficient Methods to Compute Optimal Tree Approximations of Directed Information Graphs", IEEE Transactions on Signal Processing, June 2013 http://coleman.ucsd.edu/wp-content/uploads/2013/07/QKC-TransSigProc-2013.pdf

S. Kim, R. Ma, D. Mesa, and T. P. Coleman, "Efficient Bayesian Inference Methods via Convex Optimization and Optimal Transport", IEEE International Symposium on Information Theory, July 2013. http://coleman.ucsd.edu/wp-content/uploads/2013/07/KMMC-ISIT2013.pdf

D. Kim,* N. Lu,* R. Ma,* Y. Kim, R. Kim, S. Wang, J. Wu, S. M. Won, H. Tao, A. Islam, K. J. Yu, T. Kim, R. Chowdhury, M. Ying, L. Xu, M. Li, H. Chung, H. Keum, M. McCormick, P. Liu, Y. Zhang, F. G. Omenetto, Y. Huang, T. Coleman, J. A. Rogers, "Epidermal Electronics", Science, August 12, 2011. http://coleman.ucsd.edu/wp-content/uploads/2011/08/2011EpidermalElectronics.pdf

Ulisses Braga-Neto, Texas A&M University
"Particle Filtering Approach to State Estimation in Boolean Dynamical Systems"
Authors: Ulisses Braga-Neto
Abstract: Exact optimal state estimation for discrete-time Boolean dynamical systems may become impractical computationally if system dimensionality is large. In this work, we consider a particle filtering approach to address this problem. The methodology is illustrated through application to state tracking in high-dimensional Boolean network models. The results show that the particle filter can be very accurate under a moderate number of particles. The impact of resampling on performance is also investigated.
http://www.soihub.org/group/33/SID_BragaNeto.pdf

Sunday Ogedengbe, Prairie View A&M University
"Hybrid Systems Modeling of Drug Effect on Tumor Growth"
Authors: Sunday Ogedengbe, Xiangfang (Lindsey) Li
Abstract: It is widely accepted that effective mathematical modeling would allow quantitative thinking to the development of oncology drugs and would be beneficial to increase drug development productivity. In this study, a novel hybrid system model is proposed to study drug effect from the perspective of tumor growth dynamics, and we study the practical case of periodic drug intake. Specifically, the effects of the timing (when) and the dosage (quantity) of the drug is investigated. We tend to answer such questions: whether small frequent drug intake or large infrequent drug intake is better? We propose a new concept called drug efficacy region and derive it analytically. Simulations are performed using MATLAB/SIMULINK to validate our analytical results.
http://www.soihub.org/group/33/Sunday-Ogedengbe.pdf

Xingde Jiang, Texas A&M University
"High Dimensional Bolstered Error Estimation Based on Naive Bayes"
Authors: Xingde Jiang, Ulisses Braga-Neto
Abstract: Good error estimation has been a crucial problem in classification. It is more so in Genomics research, where we have just a few dozens of samples, but thousands of features. In this setting, bolstered error estimation has been shown to have better performance than cross-validation and competitively with bootstrap. Previous research on bolstered error estimation has focused on finding an optimal spherical kernel for all features by introducing an additional correction factor. But the factor depends on classification rule, sample size, feature space, etc. In this paper, we proposed ellipsoid kernels, with a separate univariate variance parameter estimated from each feature separately based on naive Bayes. Simulation results on synthetic data and real data showed that it is more accurate in high-dimensional spaces than the original bolstering method and the other error estimation methods considered here.
http://www.soihub.org/group/33/xingde-jiang.pdf

Meng Lu, Texas A&M University
"Categorical Principle Component Analysis for Genome-Wide Association Analyses"
Authors: Meng Lu, Jianhua Huang and Xiaoning Qian
Abstract: Genome Wide Association Studies (GWAS) aim to identify genetic variants with significant role in diseases by examining the relationship between a disease and millions of SNPs across chromosomes. The high-dimensionality of SNP data is a big challenge in this association analysis. We propose a Categorical PCA to aggregate SNPs into a few optimal combined signals by dimension reduction based on an explicit modeling of SNP data. The results from simulation study and Crohn's disease analysis demonstrate the superiority of our CPCA compared with other state-of-the-art methods.
http://www.soihub.org/group/33/Meng-Lu.pdf

http://www.soihub.org/group/33/Poster_Ting.pdf; Ting Chen, Texas A&M University
"Maximum Likelihood Estimation of the Binary Coefficient of Determination"
Authors: Ting Chen and Ulisses Braga-Neto
Abstract: The Coefficient of Determination (CoD) has significant application in Genomics. Assuming a two-input stochastic logic model, we introduce a new sample CoD estimator based upon maximum likelihood (ML) estimation. Experiments have been conducted to assess how the ML CoD estimator performs in recovering predictors in a proposed logical gene regulatory setting. Performance is compared with traditional “model-free” CoD estimators such as resubstitution, leave-one-out, cross-validation and bootstrap CoD estimators. It shows that ML CoD estimator can give better recovery rate than those nonparametric approaches when prior knowledge about gene regulation is available.
http://www.soihub.org/group/33/Poster_Ting.pdf

Amin Zollanvari, Texas A&M University
"Effect of Mixing Probabilities on the Bias of Cross-validation Under Separate Sampling"
Authors: Amin Zollanvari, Ulisses M Braga-Neto, and Edward R. Dougherty
Abstract: Cross-validation is commonly used to estimate the overall error rate of a designed classifier in a small-sample expression study. The true error of the classifier is a function of the prior probabilities of the classes. With random sampling these can be estimated consistently in terms of the class sample sizes, but when sampling is separate, meaning these sample sizes are determined prior to sampling, there are no reasonable estimates from the data and the prior probabilities must be “estimated” outside the experiment. We have conducted a set of simulations to study the bias of cross-validation as a function of these “estimates.” The results show that a poor choice for estimating these probabilities can significantly increase the bias of cross-validation as an estimator of the true error.
http://www.soihub.org/group/33/Poster_ZollBragDoug.pdf

Arghavan Bahadorinejad, Texas A&M university
"Change Detection in Stochastic Boolean Dynamical Systems"
Authors: Arghavan Bahadorinejad, Ulisses Braga-Neto
Abstract: Boolean dynamical models have been established as useful models for gene regulatory networks. The problem of change detection is particularly important in the study of cancer, where it is known that the onset of different stages for disease progression is characterized by structural changes in transcriptional regulatory networks. We consider that the Boolean network characterizing the disease-free ("normal") model is known. Then two separate cases are considered: (1) the disease ("fault") model is known, in which we apply the Boolean Kalman Filter (BKF) methodology of [1]; (2) the fault model is unknown, in which case we develop an algorithm for change detection based on the BKF estimation residue.
http://www.soihub.org/group/33/Arghavan-Bahadorinejad.pdf

Mohammad Shahrokh Esfahani, Texas A&M University
"Prior Probability Construction Using Biological Signaling Pathways"
Authors: Mohammad Shahrokh Esfahani and Edward R. Dougherty
http://www.soihub.org/group/33/esfahani_poster_prior.pdf

Priyadharshini Sundararajan Venkatasubramanim, Texas A&M University
"Belief Propagation in a Probabilistic Graphical Model of the Pancreatic cancer Signaling Pathway"
Authors: Priyadharshini S.V., A. Datta
Abstract: Pancreatic cancer is among the most lethal types of cancers, with a five year survival rate of less than 6%. Cancer cells rely on their ability to rewire metabolic pathways to provide themselves with the necessary building blocks. Cancer cells can also activate autophagy, the mechanism that involves cell degradation and recycling of cellular components as internal fuel sources. Better inference models can help in cancer diagnosis and intervention. Our objective is to develop models that help in statistical inference of gene regulatory networks which at the same time retain the biological significance of the networks.
http://www.soihub.org/group/33/Priyadharshini-Venkat.pdf