Abstract

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) causes coronavirus disease 2019 (COVID-19). Imaging tests such as chest X-ray (CXR) and computed tomography (CT) can provide useful information to clinical staff for facilitating a diagnosis of COVID-19 in a more efficient and comprehensive manner. As a breakthrough of artificial intelligence (AI), deep learning has been applied to perform COVID-19 infection region segmentation and disease classification by analyzing CXR and CT data. However, prediction uncertainty of deep learning models for these tasks, which is very important to safety-critical applications like medical image processing, has not been comprehensively investigated. In this work, we propose a novel ensemble deep learning model through integrating bagging deep learning and model calibration to not only enhance segmentation performance, but also reduce prediction uncertainty. The proposed method has been validated on a large dataset that is associated with CXR image segmentation. Experimental results demonstrate that the proposed method can improve the segmentation performance and decrease prediction uncertainties simultaneously.

Speaker Bio

Lucy Nwosu (Presenter) received B.S. degree in electronic engineering, a M.S. degree in computer engineering, and is a Ph.D. student, pursuing multidisciplinary research in biomedical data analysis and machine learning, in CRI Center for Computational Systems Biology at Department of Electrical and Computer Engineering at Prairie View A&M University (PVAMU), supervised by Dr. Xishuang Dong. She is also a Graduate Research Assistant with the center, performing her research in biomedical image processing, deep learning, and computational systems biology.

Xishuang Dong is a member of CRI Center for Computational Systems Biology and Assistant Professor at Department of Electrical and Computer Engineering at Prairie View A&M University (PVAMU). He received B.S. degree in computer science and technique at Harbin University of Science and Technology, M.S. degree in computer software and theory at Harbin Engineering University, and Ph.D. in computer application at Harbin Institute of Technology. His research interests include: (1) machine learning based computational systems biology; (2) biomedical information processing; (3) deep learning for big data analysis; (4) natural language processing.

Abstract

Many years and billions spent for research has not yet produced an effective answer to cancer. Not only each human, but even each cancer nodule in the same tumor, has unique transcriptome topology. The differences go beyond the expression level to the expression control and networking of individual genes. Moreover, the transcriptome topology has random dynamics due to the stochastic fluctuations of the microenvironment and external stimuli. The unrepeatable heterogeneous transcriptomic organization among humans makes senseless the meta-analysis of large populations and the quest for universal biomarkers and “fit-for-all” treatments. We present a bioinformatics procedure to identify for each patient his/her unique triplet of cancer Gene Master Regulators (GMRs) and predict consequences of their experimental manipulation from high throughput gene expression profiles. The procedure is based on the Genomic Fabric Paradigm (GFP), which characterizes each individual gene by the independent expression level, variability and coordination with each other gene, a four order of magnitude increase of the information provided by the transcriptomic experiments. GFP can identify the GMRs whose controlled alteration would selectively kill the cancer cells with little consequence on the normal tissue. So far, the method was applied to microarray data generated in the IacobasLab by profiling surgically removed prostates, kidneys and thyroids from people with metastatic cancers, and standard cell cultures of human cancers of prostate, thyroid, lung and blood. The applications verified that each cancer case is unique and predicted the consequences of the GMRs’ manipulation. The experiments were performed in the academic laboratories of former trainees from: Australia, Brazil, Georgia, Israel, New Jersey, New York, Romania and Texas.

Speaker Bio

Dr. Dumitru A. Iacobas, Research Professor and Director of the CCSB Personalized Genomics Laboratory since 2018, is an expert of both experimental and computational genomics. Trained as a biophysicist (PhD of the University of Bucharest, Romania), he was on faculty positions at medical schools from Romania (1981-2001) and NY (Albert Einstein College of Medicine-Neuroscience 2001-2013, New York Medical College-Pathology 2013-2017). At NYMC he founded and directed the Systems Biology Core and at Einstein he was the Associate Director of the Neurogenomics and Biometry Core Facilities. Of the 323 Iacobas’ publications (3 patents, 37 books/editions, 21 book chapters, 105 articles, 27 conference proceedings, 86 genomic dbases, 2 bioprotocols, 7 bioprojects, 7 nucleotides, 24 proteins etc.) 74 are as single author, 157 as first author and 32 as last (senior) author. The cancer-related publications include 3 book chapters and 10 papers in: Biological Theory, Cancers, Cancer and Oncology Research, Cells, Current Issues in Molecular Biology, Genes, J Cancer Immunology, Oncotarget, and World J Clinical Oncology. Moreover, he is the Guest Editor of the MDPI Special Issue “Molecules at Play in Cancer” https://www.mdpi.com/journal/cimb/special_issues/molecules_cancer (12 articles so far).

Abstract

Microsoft Windows supports hundreds of millions of customers across numerous countries. In this talk, we will cover how data science and machine learning supports improvements to customer experiences with a focus on an end-to-end example on Windows updates. We will address the complexity of the space, challenges, illustrate techniques used and deployment of the techniques.

Speaker Bio

Archana Ramesh is a Principal Data Science Manager within the Windows + Devices organization. She has more than a decade of experience in working with data, conceptualizing, building and shipping ML solutions at scale. She has worked with multiple industries, gaining exposure to different types of data including genomics, healthcare, advertising and telemetry. Her academic background is a Ph.D. in Computer Science from Arizona State University where she focused on knowledge integration in large-scale biological networks. Her current focus area is towards improving the experience of Windows updates.

Abstract

Light as an energy source is essential for plants to synthesize their food from water and carbon dioxide through photosynthesis. Therefore, plants continuously monitor light intensity, direction, and quality to adjust their growth and development to the light conditions in their microenvironments. In particular, they are sensitive to shading by their neighbors because it could be determinantal to their survival. Plants can detect their proximity to neighbor plants in their surroundings before they are shaded, and activate growth and developmental responses known as the shade avoidance syndrome (SAS). The SAS includes shoot elongation, reduced branching, and early flowering, which are also important determinants of plant shoot architecture and yield of crops. Therefore, understanding how shade signals regulate shoot elongation and branching could help to improving crop production and productivity. In this seminar, I will present our research progress on how shade signals promote shoot elongation in sorghum.

Speaker Bio

Dr. Tesfamichael Kebrom is a research scientist with a joint appointment at the Center for Computational Systems Biology (College of Engineering) and the Cooperative Agricultural Research Center (College of Agriculture and Human Sciences) at Prairie View A&M University (PVAMU). He received B.Sc. in Plant Sciences from the University of Asmara in Eritrea, M.Phil. (Master of Philosophy) in Crop Physiology from the University of Reading in England, and Ph.D. in Molecular & Environmental Plant Sciences from Texas A&M University. His research focuses on identifying molecular pathways and gene regulatory networks controlling shoot elongation and branching in response to environmental and developmental signals using plant systems biology methods.

Dr. Lealon L Martin joins the Center for Computational Systems Biology to strengthen his research and further his collaboration with the center. Dr. Martin is a faculty member in the Chemical Engineering Department at Prairie View A&M University and also serves as Associate Dean (interim) of the Roy G. Perry College of Engineering. As an aside, Dr. Martin is also a licensed attorney and a member of the State Bar of Texas.

Dr. Martin, in collaboration with Dr. Seungchan Kim, was awarded a Sandia START collaborative grant (Jan 1, 2022 - Sept. 30, 2024) to study “Engineering Methanotrophy for Carbon Capture and Utilization”. The project is a collaboration with Sandia National Laboratory’s Bioresource & Environmental Security Team to identify key microbial and environmental factors for upregulating methanotrophy in typically methanogenic systems in agriculture and waste management, including landfills, waste-water, and rice cultivation, among others. In particular, Dr. Martin is looking to synthesis and reconstruction methane pathways in methanotrophic and methanogenic microorganisms, and provide guidance on how to maintain stable consortia of these microorganisms to convert greenhouse gases into biomass.

Project Title: Engineering Methanotrophy for Carbon Capture and Utilization

Principal Investigator: Seungchan Kim, Ph.D.

Purpose and Objective The purpose of the project is to identify key microbial and environmental factors for upregulating methanotrophy in typically methanogenic systems in agriculture and waste management, including landfills, waste-water, and rice cultivation, among others.

The Subcontractor shall use publicly available and Sandia National Laboratories (SNL) generated data to computationally identify interacting metabolic and microbial networks in methanogenic, methanotrophic, and perturbed consortia, for application in Agent Based Modelling (ABM) and metabolic Flux Balance Analysis (FBA).

Tasks

  1. The Subcontractor shall provide the SNL Principal Investigator (PI) with preliminary analysis, based on literature review, for determination of experimental data requirements for statistically significant biochemical and microbial network analysis results from a Design-Build-Test-Learn (DBTL) cycle incorporating Design of Experiments (DoE) based on methanogenic and methanotrophic microbial consortia.

  2. The Subcontractor shall apply network analysis using SNL provided multi-omics data, including microbiomes from 16S rRNA-amplicon sequencing, phenotypic analysis, and flux of major metabolytes from testbed cultures. Major microbial and biochemical nodes that are identified shall provide the SNL R&D team insights into engineering targets for the intended applications.

  3. The Subcontractor shall deliver reduced-order cohort model scenarios composed of 3-10 agents from all available data for implementation in ABM. The cohort model scenarios shall include key interacting microbial species and associated biochemical pathways relevant to maximizing and stabilizing methanotrophy in the presence of methanogenic agents, and include sensitivity estimates and environmental bounds (temperature, pH, oxygen, redox potential, conductivity, nutrients, and gross substrate flux balance) for each cohort.