Selected Abstracts
Talks & Panel discussions: NatSci 128 (Keynotes, Technical/Lightning talks, and WiDS Career Panel)
Lunch: NatSci 150
Posters: NatSci 145
PhD student working with the Krishnan Lab in CMSE at MSU. Interested in machine learning and natural language processing in human genomics.
Predicting Alzheimer’s disease-associated protein biomarkers from multiple evidence sources
Anna Yannakopoulos*, Alison Bernstein, Irving Vega, Arjun Krishnan
Complex diseases such as Alzheimer’s disease (ALZ) are caused by several perturbed cellular mechanisms associated with hundreds of interacting genes. To identify novel disease-gene linkages, we use machine learning (ML) to build upon known disease-associated genes to identify others based on hundreds of thousands of human genetic and molecular data that is publicly available. We specifically focus on the fact that known disease-gene associations are often sourced from a variety of studies and study types of varying quality. To that end, we examine methods to generate predictions that take into account both the data itself as well the reliability of its source in the context of predicting ALZ-associated protein biomarkers. In addition to proteins quantified directly using mass spectroscopy, we evaluate multiple study types including microarray and genome-/epigenome-wide association studies on their ability to recapitulate ALZ biomarkers. We then incorporate these concordance measures within weighted ML methods to produce an integrated genome-wide ranking of ALZ-associated protein biomarkers. Our results show that, compared to only including known ALZ protein biomarkers, naively including all available data without regard to quality lessens predictive performance while incorporating data weighted with respect to its quality improves predictive performance. Experimental evaluation of our top predictions implicates several novel candidates as bona-fide ALZ-associated biomarkers.
Melissa earned her undergraduate and Master's degrees at Montana State University. She graduated Magna Cum Laude from the honors college with a BS in Computer Science, BA in Modern Languages, and a minor in Physics. She currently is working on a PhD at Michigan State University in Computer Science with a focus on Machine Learning, Computer Vision, and Medical Imaging.
Leveraging Auto-Tuned Machine Learning Models in a Fusion Framework for Stem Cell Detection
Melissa Dale, Arun Ross, Erik Shapiro.
MSU Computer Science-College of Engineering, MSU Department of Radiology.
Stem cells have the potential to play a critical role in the treatment of many medical conditions. The ability to detect the presence of stem cells is an important aspect in understanding this potential but detection is a challenging task that requires trained medical professionals to laboriously inspect MRI scans. This process is time intensive, expensive, and subject to human error.
To study automating the process, Afridi et. al [1] created a dataset of MRI scans of rat brains injected with stem cells. In addition to the direct MRI samples, we generate additional feature-sets using descriptors such as histogram of oriented gradients (HOG) and local binary pattern (LBP) to describe the local textures of the MRI scans surrounding stem cells.
We then obtain several machine learning models from this data with an automated machine learning tool, over 30 of which outperform previous hand-crafted models created by domain knowledge experts with AUC scores near 94%.
We push AUC scores to over 99% using score fusion. To obtain scores, we consider prediction probabilities for model classifiers such as linear regression and multi-layer perceptrons, euclidean distance for K Nearest Neighbors classifier, and majority voting for Extra Trees and Random Forest classifiers.
[1] Afridi, Muhammad Jamal, et al. "Intelligent and automatic in vivo detection and quantification of transplanted cells in MRI." Magnetic resonance in medicine 78.5 (2017): 1991-2002.
Morgan Oneka is a first year PhD student in the Department of Computational Medicine and Bioinformatics at the University of Michigan. Her recent work focuses on quantifying the spatial arrangement of tumor micro-environments in various types of cancer.
Characterizing Tumor Microenvironments Using a Multidisciplinary Approach.
Morgan Oneka (University of Michigan Department of Computational Medicine and Bioinformatics), Jenny Lazarus (University of Michigan Department of Surgery), Souptik Barua (University of Michigan Department of Computational Medicine and Bioinformatics), Arvind Rao (University of Michigan Department of Computational Medicine and Bioinformatics), Timothy L. Frankel (University of Michigan Department of Surgery)
Cytotoxic T-cells, or CTLs, are immune cells that kill cancer cells. Utilizing spatial G-function analysis, a method commonly used in ecology, we have shown that the infiltration of CTLs into cancer cells is a useful tool in predicting patient prognosis. Current ideas being investigated following this discovery include the adaptation of social network analysis and graph theory to identify clustering of T-cells, how to characterize these clusters, and whether clustering of CTLs holds clinical significance.
I am an Academic Specialist at Michigan State University (MSU) and the Research Evaluation and Data Manager of MSU’s AgBioResearch, the research arm of the College of Agriculture and Natural Resources (CANR). I work closely with MSU AgBioResearch leadership to formulate and implement the monitoring and evaluation (M&E) and research evaluation goals for MSU AgBioResearch and CANR. I develop appropriate metrics and tools for evaluation, collect statistical data, and conduct research and evaluation studies. I also liaise with CANR and other units in the university to build comprehensive M&E system and deploy research intelligence techniques to assist in the planning, monitoring and impact assessment of MSU AgBioResearch and CANR activities. My academic and policy-related publications (http://orcid.org/0000-0003-1105-5851) focus on scientometrics, innovation management, technology transfer and intellectual property management (IP) of knowledge products in agriculture. My interests include: internationalization of science, data-driven analytics and research evaluation, IP management for development countries, econometric studies on international development, and metrics design and development. Prior to joining MSU, I worked at King Abdullah University of Science and Technology in Saudi Arabia, Washington State University, and Philippine Rice Research Institute in the Philippines. I also have adjunct faculty appointment with Washington State University and a core faculty of MSU’s Center for Gender in Global Context, MSU’s Africa Studies Center, and MSU’s Institute for Public Policy Research. I received my Doctor of Philosophy at Washington State University, Pullman, WA; and undergraduate and Master’s degrees in the Philippines: Bachelor of Science in Biology (Cum laude) from Central Luzon State University and Master in Technology Management (outstanding graduate) from the University of the Philippines Diliman.
Judging Value and Returns on Investment in Africa-Focused Research and Partnerships through Bibliometrics
U.S. global engagement in various areas including global health, education, environment and agriculture has grown in the last 15 years. However, there is limited literature that acknowledges the return on investment of U.S.-global research collaboration. This research showcases the use of bibliometric approaches to judge the value and assess return on investment in global-focused research and global collaboration using select research metrics. Specifically, this research focused on measuring the research output, describe the collaboration patterns and research trends, and assess the impact of Michigan State University's research engagement on Africa for the last ten years (2006-2015). It also attempts to determine the gender dimension of MSU’s research focus on Africa. Our analysis provides some evidence of scholarly success, research focus and diversification, collaboration, and impact in MSU’s engagement on Africa, especially in STEM (Science, Technology, Engineering and Math) related disciplines. It also provides some evidence of greater collaboration of MSU with non-African countries on African-focused research. Finally, it provides some evidence of increasing involvement of women in advancing knowledge and global innovation to address Africa’s socio-economic development challenges. This aims to establish a quantitative measure related to the impacts of R&D investment in Africa, and results are expected to inform future research and collaboration strategy between MSU and African institutions.
Kayla is a dual Ph.D. student in the departments of Biochemistry and Molecular Biology and Computational Mathematics, Science, and Engineering at Michigan State University. She is currently working in Arjun Krishnan's group in the computational biology field.
Building Robust Gene Co-expression Networks from RNA-seq Data
Kayla Johnson, Arjun Krishnan
Departments of Biochemistry and Molecular Biology and Computational Mathematics, Science, and Engineering, Michigan State University.
As the cost of RNA sequencing has continued to fall, the amount of publicly available RNA-seq data has continued to grow; currently, there are over 80,000 publicly-available human RNA-seq samples. A predominant method for studying gene function in specific biological contexts is to construct a gene co-expression network using transcriptomes from those contexts. Although many studies have focussed on best preprocessing procedures for use of RNA-seq data for analysis of differential expression, not enough attention has been given to best practices for processing RNA-Seq data for calculating gene co-expression. Constructing an accurate co-expression network depends on several factors including expression quantification from read count and the presence of experimental and technical artifacts, which introduce non-biological variation into the data. In this research, we leverage thousands of uniformly aligned RNA-seq samples from various experiments that span diverse tissues, diseases, and conditions to investigate these factors. We construct gene co-expression networks using different within-sample and between-sample normalizations and network transformation methods and then evaluate the resulting networks based on their ability to recover documented tissue-naive and tissue-specific gene functional relationships. This comprehensive benchmarking provides insight to the best procedures for deriving a robust gene co-expression network from an RNA-seq dataset.
PhD Candidate & Graduate Assistant. Statistics and Probability, Natural Science
Application of REML model in Big Data
Raka Mandal, Gustavo De Los Campos, Taps Maiti, Alexander Grueneberg
Restricted maximum likelihood (REML) methods have been classically used for estimating variance components in linear mixed models. In this work, we have applied REML model in UKBiobank data (interim release) and empirically shown that the REML estimate an be seriously biased.
Sarah Tymochko is a PhD student at Michigan State University in the Department of Computational Mathematics, Science, and Engineering. Her primary research interests include topological data analysis and machine learning.
Quantifying Periodic Circular Structures in Hurricane Felix
Sarah Tymochko* (Dept. of CMSE), Elizabeth Munch (Dept. of CMSE and Dept. of Mathematics), Jason Dunion (Cooperative Institute for Marine and Atmospheric Studies University of Miami and Hurricane Research Division NOAA/Atlantic Oceanographic and Meteorological Laboratory), Kristen Corbosiero (University at Albany Dept. of Atmospheric and Environmental Sciences), Ryan Torn (University at Albany Dept. of Atmospheric and Environmental Sciences)
The diurnal cycle of tropical cyclones (TCs) is a daily cycle in clouds that can be seen in infrared (IR) satellite imagery and may have implications for the structure and intensity of the TC. This pattern is seen as cyclic pulses in the cloud field that propagate radially outward from the center of nearly all Atlantic-basin TCs. These pulses, a distinguishing feature of the diurnal cycle, appear as a ring-like region propagating outward from the storm’s inner core. The state of the art TC diurnal cycle measurement has a limited ability to analyze the behavior beyond qualitative observations. We present a method for quantifying the TC diurnal cycle using tools from Topological Data Analysis. Using Geostationary Operational Environmental Satellite IR imagery data from Hurricane Felix (2007), our method is able to detect an approximate daily cycle.
Data Scientist, Analytics & Data Solutions
Data Science Toolkit
Morgan Patterson1, Ezra Brooks.
Analytics & Data Solutions.
The Analytics and Data Solutions Data Science team here at MSU has and continues to work on both analytical and interactive data visualization projects utilizing R and R Shiny. The objective of this project is to create a Data Science Toolkit through an R Shiny application that allows users to input various types of data and receive suggestions on appropriate analysis methods and provide data summaries and visualizations. Initial work of this project includes a statistical modeling section and a text/sentiment analysis section. Currently the modeling section allows a user to select desired variables for a linear regression and would receive both graphical and numerical output from that method. However, we want this toolkit to also be a guide and recommendation tool for various types of data. The user also has the option to upload text data in which we want the application to provide graphical summaries of the text and sentiment analyses. Moving forward this application is sought to be a comprehensive toolkit for users to more efficiently visualize their data and to be able to appropriately apply statistical methodology.
Jessie Micallef is a third-year graduate student at Michigan State University getting a dual PhD in Physics and Computational Mathematics, Science, and Engineering. She works on the particle physics experiment IceCube under Dr. Tyce DeYoung studying neutrino oscillations, using computational methods to improve the experiment’s current software. Jessie has won awards for the NSF Graduate Research Fellowship Program and for the ACM SIGHPC/Intel Computational & Data Science Fellowship.
Using Neural Networks to Reconstruct Sparse Neutrino Events in IceCube.
The IceCube Neutrino Observatory is a particle physics experiment buried under the ice in the South Pole. It aims to detect astrophysical and atmospheric neutrinos to discover cosmic sources of neutrinos and to better constrain fundamental neutrino parameters. IceCube uses a 3D hexagonal array of 5160 photomultiplier tubes (PMTs) that detect light from neutrino interactions in the ice. The detected photons can be used to reconstruct the neutrino path, energy, interaction time, and interaction position. Atmospheric neutrinos (typically of energies 1-60 GeV) leave a very sparse signal, with their photons only hitting a fraction of the total PMTs (less than 5%). Using the incident photon time and charge recorded by the array of PMTs, my goal is to see if it is possible to train neural networks to reconstruct atmospheric neutrino events. The first trial is to use a convolutional neural network to reconstruct the energy of the neutrino. Some challenges being faced, which would benefit from peer discussion, are dealing with sparse data in a neural network and implementing a convolutional network on a hexagonal structure.
Monique Noel is a 3rd year PhD Student in the Dept of Chemistry. She received her B.S. from Florida Agricultural and Mechanical University. She enjoys educating and giving scholarly advice to underrepresented youth in STEM. She plans to work in a Industry/Military environment before following the professorial path.
Performance of Continuum Solvent Models to compute Solvation Energetics.
Modern quantum chemical tools allow quite accurate modeling of the structures and reactions of molecules in isolation. However, most organic reactions and essentially all biological processes occur in solutions, condensed phases in which the reacting species are in direct contact with the surrounding medium. This medium is usually a key component in directing the chemical paths followed, which typically differ from those expected in the gas phase. However, explicit inclusion of even a few solvent molecules, with their many degrees of freedom, quickly makes computations overwhelmingly expensive. For these reasons, various Continuum Solvent Models (CSMs) have been developed, in which the species of interest are embedded in a polarizable continuum which represents the solvent. Different CSMs incorporate different electrostatic and non-electrostatic interactions to capture effects such as the continuum's dielectric polarization, cavity formation, and dispersion forces between substrate and medium to yield a free energy of solvation. Such quantum calculations can in principle predict vapor-liquid phase equilibria, basic thermodynamic properties of common solvents. An in-depth study of these equilibria using several Solvent Models is hereby proposed to both test their value and to familiarize the public with Quantum Chemical methods that practically and usefully predict real chemical behavior.
PhD Candidate, MSU Department Of Statistics and Probability
Mean-Reverting Processes with Gumbel Noise, and Modelling of Sea-Level Rise Risk
Fatimah Alshahrani, Prof. Frederi Viens
Michigan State University
Extreme Value Theory (EVT) deals with studying ”rare events” by taking either the maxima or minima of observations. We applied EVT to the annual maximum for sea level data provided by the Actuaries Climate Index, (ACI). Our analysis shows that the annual maxima follow the Type I Generalized Extreme Value Distribution, which is known as the Gumbel distribution. For the whole time series of monthly records, as a random sample, ACI claims it follows the Gumbel distribution. We know it is a time series, and our initial investigation shows the detrended data has some of the features of the Ornstein Uhlenbeck (OU) process, i.e. a Gaussian mean reverting process. But unlike the OU process, our data is presumably not Gaussian. Using the relationship between stochastic differential equations and discrete time processes, we are building a reliable model using tools from stochastic analysis including the Malliavin Calculus, and objects such as the AR(1) process and the OU process, while drawing on the so-called Gumbel noise, which is an object used largely in Machine Learning.
I am a doctoral student in Measurement and Quantitative Methods program at MSU. I am currently working on my dissertation on test validity via investigating gender and performance level factors. My other research interests are: computerized adaptive tests, item response theory, structural equation modeling, multilevel modeling, multivariate statistics, test fairness, STEM education, SES and gender in education, social-cognitive learning theories, student and teacher motivation and well-being. I am an R user and working on improving my programming skills at present.
Effects of Cognitive and Social Mindsets on Women Perseverance in STEM
Nazli Uygun
MSU College of Education, Dept. of Counseling and Ed. Psychology, Measurement and Quantitative Methods
This study presents a brief literature review of the cognitive and social factors affecting women perseverance in STEM and provides suggestions for teachers and educators to reduce gender gap in the lieu of existing research.
Evidence from decades of research presents that youth with growth mindset, i.e., (believe that intelligence can be developed by practice and persistence) are more likely to persist in the face of challenge; while individuals with fixed mindset; (i.e. intelligence is innate) are susceptible to give up in the face of challenge or fear of failure (Blackwell, Trzesniewski, & Dweck 2007; Yeager and Dweck 2012; as cited in Wang, 2008). Dweck et al. (2007) concluded that girls cope not as much as boys when confronting challenging math problems which cause reduced math performance compared to boys. To sum up, it is obvious that fixed mindsets can be problematic which may lead to performance decrease and career choice avoidance in challenging STEAM fields (Wang, 2008).
Luigi-Guiso (2008) investigated PISA 2015 data to explore gender gap in math performance across the world. Girls from all of the participating countries have 10.5 lower mean score than boys. However, this score varies base on gender equity of the country. Being more gender-neutral countries, Norway and Sweden has no gender gap, while in Turkey; gender gap is greater where boys outperform 22.6 points by average. Iceland has a reversed gender difference: girls outperform boys 14.5 points. Finally, researchers concluded gender gap would be eliminated if counties treated women as equal as men.
Lessons from literature are: 1) teachers and educators should distribute growth or malleable mindset of intelligence to promote women success cognitively in STEM, 2) gender-neutral countries should be role-modeled while treating women in classroom and society to achieve gender equity in STEM fields.
Graduate Research Assistant, University of Michigan, Dept. of Mechanical Engineering.
Identifying turbulence closures using sparse regression with embedded invariance.
S. Beetham, J. Capecelatro
University of Michigan.
In this work, we present a data-driven framework for model closure of the Reynolds Average Navier–Stokes (RANS) equations. It is well understood that direct numerical and large-eddy simulations of turbulent flows are inaccessible for industrial and large-scale problems. For this reason, RANS models remain the most widely used tool for informing engineering designs and decisions. In recent years, the scientific community has turned to machine learning techniques to improve the models the RANS equations require for closure. While the body of work in this area has focused on various areas of the closure problem and built upon a diversity of machine learning techniques, we leverage a sparse regression framework to develop new closure models due to two key properties: (1) The resulting model takes on a closed, algebraic form, allowing for direct physical inferences to be drawn and direct integration into existing CFD solvers and (2) Galilean invariance can be guaranteed by thoughtful tailoring of the feature space. We find preliminary success for an exemplary case: free shear turbulence.
Sarah Goodwin, Ph.D., is a postdoctoral researcher in Human Development and Family Studies at Michigan State University. In their Early Language and Literacy Investigations Laboratory, she conducts psychometric analyses of assessments of phonological awareness and manages lab data. Sarah's research interests include language assessment, specifically Rasch measurement approaches, and language acquisition. Her work can be found in Language Assessment Quarterly, Assessing Writing, and the TESOL Encyclopedia of English Language Teaching.
An Item Level Examination of Monolingual and Multilingual Children's English Phonological Awareness.
Sarah Goodwin, Lori Skibbe, Gary Troia, Ryan Bowles
Michigan State University
Phonological awareness, the understanding of the sound structure of language, is crucial to literacy skill development (Foy & Mann, 2006). Yet for bilingual or multilingual children, it is unclear whether multilingual children may develop stronger early literacy skills due to greater exposure to various phonological forms, or have weaknesses resulting from irregularities between languages. The goal of the poster is to examine phonological awareness at a more fine-grained level to consider differences between monolingual and multilingual children at the item and subskill level.
The present study assessed 3- to 7-year-olds’ phonological awareness skills (n=102). Parents reported that multilingual children used another home language and understood or spoke little English. Monolingual children were chosen from a larger group (n=966) and matched with the multilingual participants to the extent possible by age, gender, ethnicity, socioeconomic status, and vocabulary level.
Children responded to 60 rhyming, blending, or segmenting items: hearing a recorded prompt and selecting the picture best representing the correct answer (“Dol… phin. What is the whole word?” [birdhouse, dolphin, and highchair pictures]). The findings suggest that multilingual children may be slightly more sensitive to removing sounds or syllables from words. Similar to prior work with older participants (Loizou & Stuart, 2003; Marinova-Todd et al., 2010), evidence points to multilingual children having different phonological awareness strengths when compared to monolingual peers.
Elizabeth LaPensée, Ph.D., is an award-winning Anishinaabe and Métis game developer and researcher. She is an Assistant Professor of Media & Information and Writing, Rhetoric & American Cultures at Michigan State University and a 2018 Guggenheim Foundation Fellow.
Generative Generations: Game Development as a Pathway to Engage Indigenous Youth in Science.
Elizabeth LaPensee (Media & Information and Writing, Rhetoric, & American Cultures), Christie M. Poitra (MSU Native American Institute), Estrella Torrez (Residential College for Arts & Humanities), Angela Kolonich (MSU CREATE for STEM Institute)
For Indigenous youth from elementary through to post-secondary levels, there are many challenges to seeing themselves in Science, Technology, Education, Arts, and Math (STEAM) education and careers due to lack of representation and culturally relevant connections. In the hopes of addressing this concern, a circle of Indigenous and Latinx women community-oriented scholars collaborated to conduct pilot research on Generative Generations, a series of game development curriculum kits aimed at integrating Indigenous pedagogical approaches and culturally relevant identity development in Science, Technology, Education, Arts, and Math (STEAM) for Indigenous youth. The hands-on kits were designed with culturally rooted science ideas from across all four domains including Life Science, Earth and Space Science, Physical Science, and Engineering in collaboration with Indigenous community partners for Indigenous youth ages 9-16 to be implemented in informal STEAM learning contexts, such as community workshops or after-school programs. The pilot study held with the Indigenous Youth Empowerment Program (IYEP) reflects how interweaving Indigenous ways of knowing with game design for culturally relevant Indigenous identity development can lead to STEAM learning and self-efficacy.
Undergraduate Research Assistant, Department of Pathobiology and Diagnostic Investigations.
Identifying Diagnostic Targets for Pathogenic Nontuberculous Mycobacteria
Lauren Sosinski, Philip Calhoun, Samuel Chen, Janani Ravi
Pathobiology and Diagnostic Investigation, Michigan State University.
Nontuberculous mycobacterial (NTM) infections are incredibly detrimental to livestock in the agriculture community and cause secondary infections in humans. Currently, there are no effective vaccines against NTM, and diagnosis techniques include, but are not limited to, testing clinically healthy cattle fecal matter and lymph nodes at slaughter. However, the bacterial diagnoses between the lymph nodes and fecal matter drastically differ. The only way to get an accurate diagnosis is after the infection has spread or the animal is near death. Secondary infections by NTM that occur in humans require long-term treatment and can become a chronic, and expensive, issue. Therefore, there is a critical need for improving the methods for diagnosing, treating, and vaccinating against NTM. In this project, I plan to identify protein targets for diagnosis in NTM by first determining known virulence factors and diagnostic targets in other mycobacteria, such as M. tuberculosis, and identifying the homologs in NTM using the BLAST database. I will use molecular evolutionary approaches (sequence-structure-function & phylogenetic) to determine the domains and genomic neighborhoods of these homologs. These analyses will show how these genes have evolved, what their putative functions are and whether they can be used as diagnostic targets. Finally, I will compare the genomes of pathogenic NTM to both Mycobacterium tuberculosis and nonpathogenic NTM to ensure specificity and sensitivity of the identified targets. This will allow me to find unique proteins in pathogenic NTM to use in diagnosis.
DO/PhD candidate in the Physiology department
The role of E2F5 in breast cancer progression and metastasis.
Briana To (MSU Physiology, College of Osteopathic Medicine), Dr. Eran Andrechek (MSU Physiology, College of Human Medicine)
Despite advances in cancer genomics, the molecular drivers of triple negative breast cancer (TNBC) remains highly elusive. Utilizing a genomic approach, we examined genetic differences across mouse models of breast cancer and identified a gene, E2F5, that was consistently deleted in TNBC mouse models. In human breast cancer, E2F5 activity and expression is associated with better overall and relapse-free survival. To study the role of E2F5 in breast cancer, E2F5 was deleted in the mammary gland of FVB mice (E2F5cKO) through the use of a Cre-lox system driven by the MMTV promoter. E2F5cKO mice were aged and observed for tumor development. This revealed mammary tumor formation and metastatic lesions. To begin to understand the mechanism of tumorigenesis in E2F5cKO mice, differential gene expression was performed on data derived from E2F5 or GFP overexpressed Human Mammary Epithelial Cells. One of the leading candidates that arose from this analysis was the oncogene KRAS. The expression of KRAS is negatively correlated with E2F5 expression. Using a bioinformatic approach, we demonstrated that this inverse relationship between KRAS and E2F5 appears to be conserved in human breast cancer. Based on these findings, we postulate that E2F5 deletion in the mammary gland leads to increased KRAS activity. Consistent with our hypothesis, western blot analysis revealed high activity in two major KRAS signaling pathways, AKT and ERK, in majority of E2F5cKO tumors. Together, this preliminary data suggest that E2F5 may behave as a tumor suppressor by negatively regulating the KRAS.
Student, MSU Department of Food Science and Human Nutrition.
Nutrition Content analysis of #typeonediabetes on Instagram
Phoebe Tuyishime, Deanne Kelleher, MS, RDN
The current study presents a content analysis of Type 1 Diabetes(T1D) nutrition messages on Instagram as it is a popular tool used to acquire T1D nutrition information. Social media has been widely used as a medium to share information, including nutrition information, to a wide variety of audiences. However, the quality of information on social media has been questioned by many health and nutrition professionals. Research demonstrates that there is a growing influence and use of healthcare and nutrition information obtained online (The Harris Poll, 2011). For instance, patients with chronic conditions like T1D use social media frequently to communicate with other patients whom they might be having similar health conditions to gain deeper knowledge and support by sharing ehealth information. We will carry out a content analysis of the information using quantitave and qualitative methods to assess the nutrient density of nutrition related T1D posts on Instagram. A search of Instagram posts will be conducted using popular T1D hashtags. We will carry out a content analysis of the information qualitatively using a coding system where posts with the same themes will be analyzed. We will also evaluate T1D nutrition information on Instagram quantitatively using the 2015-2020 Dietary Guidelines for Americans.
Phd student, MSU Department of Computer Science & Engineering
PHiMM: Fast and Accurate Introgression Detection using Statistical Phylogenomic Inference and resampling
Qiqige Wuyun (Computer Science & Engineering, Michigan State University), Kevin Liu (Computer Science & Engineering, Michigan State University)
Introgression is the movement of genes from one species to the gene pool of another by recurrent backcrossing of hybrid. Introgression is thought to play an important role in genome evolution throughout the Tree of Life, the evolutionary history of all life on Earth. To quantitatively investigate this hypothesis, a variety of state-of-the-art techniques have been developed for detecting introgression from genomic sequence data. However, no existing method is capable of fast and accurate introgression detection on datasets with many dozens of genomic sequences. In this work, we develop an improved introgression detection approach which enables scalable analysis of large-scale datasets. Our approach combines the multi-species network coalescent model with hidden Markov models (HMMs) to tease apart the effects of incomplete lineage sorting (ILS) from those of introgression. Using simulated and empirical data, we perform a large-scale comparative assessment of our new method with other state-of-the-art introgression detection methods such as PhyloNet-HMM and CoalHMM.. Our experiments explore a wide range of factors that may influence the performance of introgression detection, including the number of taxa, the number of alleles sampled from each taxa, sequence length, migration probability, and starting time of migration.