Seminars

Abstract

Recent years have witnessed a shift to more objective and biologically-driven methods for central nervous system (CNS) tumor classification. The recent world health organization (WHO) classification update ("blue book") introduced molecular diagnostic criteria and methylation profiling into the definitions of specific entities as a response to the plethora of evidence that key molecular alterations define distinct tumor types and are clinically meaningful. While in the past such diagnostic alterations included specific mutations, copy number changes, or gene fusions, the emergence of DNA methylation arrays in recent years has similarly resulted in improved diagnostic precision, increased reliability, and has provided an effective framework for the discovery of new tumor types. In many instances, there is an intimate relationship between these mutations/fusions and DNA methylation signatures. The adoption of methylation data into neuro-oncology nosology has been greatly aided by the availability of technology compatible with clinical diagnostics, along with the development of a freely accessible machine learning-based classifier. In my remarks, I will highlight the utility of DNA methylation profiling in CNS tumor classification with a focus on recently described novel and rare tumor types, as well as its contribution to refining existing types. In addition, I will discuss how methylation-based classification may be useful for diagnostics of tumors of additional organ systems, including hematopoietic and renal neoplasms.

Bio

 Kenneth Aldape is a neuropathologist and Chief of the Laboratory of Pathology at the National Cancer Institute at the National Institutes of Health in Bethesda, Maryland.  He focusses on genomics and molecular diagnostics of primary brain tumors, especially gliomas, focusing on the integration of biologic tumor signatures into clinical use for brain tumor classification.  He has approximately 500 peer-reviewed publications, several grants and is the recipient of several awards.  He has contributed to the WHO working group on the classification of central nervous system neoplasm and is a member of the c-IMPACT group involved in real-time updates for the classification of brain tumors.  Currently he is actively involved in technologies such as methylation profiling that can improve how we understand and diagnose these tumors from a biological basis.

Abstract

The management of cancer hinges on discovering and interrupting critical events during its course of development, a challenging task due to cancer’s high heterogeneous and dynamic nature.  Current technologies enable observing cancer specimens at single-cell resolution with cellular spatial context, providing an unprecedented opportunity to characterize molecular heterogeneity and identify predictive signatures.  In this talk, I will summarize our efforts in developing novel data science tools for multimodal data integration, transformation, feature engineering and functional interpretation, and applying tools for discovering cellular-molecular signatures associated with tumor and microenvironment coevolution and immunotherapy responses.

Bio

Dr. Chen obtained B. Eng. from Tsinghua University (Beijing) and Ph.D. in Electrical and Computer Engineering from University of Illinois at Urbana-Champaign. He is currently a full professor in the department of Bioinformatics and Computational Biology at the University of Texas MD Anderson cancer center. His primary interest is to develop computational methods to analyze and interpret high-throughput human genetic and clinical data towards understanding the evolution of cancer as a consequence of genetics and environment and identifying molecular targets useful for cancer diagnosis and therapeutics. Among the computational tools he developed, BreakDancer, VarScan and Monovar have been widely used for characterizing genomes and transcriptomes of tumor tissues and single cells.

Abstract

Whole Genome/Exome Sequencing (WGS/WES) data and Electronic Health Records (EHRs), such as large scale national and institutional biobanks, have emerged rapidly worldwide.  In this talk, I will discuss the analytic tools and resources for scalable analysis of large scale biobank- and population-based Whole Genome Sequencing (WGS) association studies of common and rare variants by integrating WGS data with multi-faceted functional annotation data. Discussions include fitting mixed models for continuous and discrete and survival phenotypes using sparse GRM in population and biobank based studies, and rare variant association tests and meta-analysis by incorporating multi-faceted variant functional annotations including single-cell-based cell-specific annotations using individual level data and WGS summary statistics. I will also provide a demo of FAVOR (favor.genohub.org), a variant functional annotation online portal and resource that provides multi-faceted functional annotations of genome-wide 9 billion variants, and FAVORAnnotator, a tool to functionally annotate any WGS/WES studies. Cloud-based platforms for these resources will be discussed. The presentation will be illustrated using ongoing large scale population-based whole genome sequencing studies and biobanks of quantitative, case-control, and time-to-event phenotypes, including the Genome Sequencing Program (GSP) of the National Human Genome Research Institute and the Trans-Omics Precision Medicine Program (TOPMed) from the National Heart, Lung and Blood Institute, and the UK Biobank and FinnGen, which have collectively sequenced about 1 million genomes.

Bio

Xihong Lin, PhD is Professor and former Chair of the Department of Biostatistics, Coordinating Director of the Program in Quantitative Genomics at the Harvard T. H. Chan School of Public Health, and Professor of the Department of Statistics at the Faculty of Arts and Sciences of Harvard University, and Associate Member of the Broad Institute of MIT and Harvard. Dr. Lin’s research interests lie in development and application of scalable statistical and machine learning methods for analysis of massive high-throughput data from genome, exposome and phenome, as well as complex epidemiological, biobank and health data.  Dr. Lin received the MERIT Award (R37) (2007-2015) and the Outstanding Investigator Award (OIA) (R35) (2015-2029) from the National Cancer Institute (NCI). She is the contact PI of the Harvard Analysis Center of the NHGRI Genome Sequencing Program, and the multiple PI of one of the Predictive Modeling Centers of the NHGRI Impact of Genomic Variation on Function (IGVF) program.  Dr. Lin is an elected member of the National Academy of Medicine. She has received several prestigious awards, including the 2002 Mortimer Spiegelman Award from the American Public Health Association, the 2006 Presidents’ Award of the Committee of Presidents of Statistical Societies (COPSS), and the 2022 Marvin Zelen Leadership in Statistical Science Award. She is an elected fellow of American Statistical Association, Institute of Mathematical Statistics, and International Statistical Institute. Dr. Lin is the former Chair of the COPSS (2010-2012) and a former member of the Committee of Applied and Theoretical Statistics of the National Academy of Science. She is the founding chair of the US Biostatistics Department Chair Group, and the founding co-chair of the Young Researcher Workshop of East-North American Region (ENAR) of International Biometric Society. She is the former Coordinating Editor of Biometrics and the founding co-editor of Statistics in Biosciences.  She has served on a large number of committees of many statistical societies, and numerous NIH and NSF review panels.

Abstract

Social media such as Twitter and Facebook provide a rich, if imperfect, portal into people's lives.  We analyze tens of millions of people Facebook posts and billions of tweets to study variation in language use with mental and physical well-being.  Word clouds provide insights into depression, empathy, and stress, and correlations between language use and county-level health data suggest connections between health and happiness, including potential psychological causes of heart disease. Recent advances in generative models such as GPT3 offer promise and peril in supporting mental health treatment.

Bio

Lyle Ungar is a Professor of Computer and Information Science at the University of Pennsylvania, where he also holds appointments in Psychology, Bioengineering, Genomics and Computational Biology, and Operations, Information and Decisions.  He has published over 300 articles, supervised two dozen Ph.D. students, and is co-inventor on ten patents. His current research focuses on using natural language processing and explainable AI for psychological research, including analyzing social media and cell phone sensor data to better understand the drivers of physical and mental well-being.

Abstract

Alzheimer disease (AD) is a progressive neurodegenerative disease with complicated underlying disease mechanisms. The clinical and neuropathological heterogeneity of AD may be explained by specific disease mechanisms at a brain cell-level. We postulate that the path to effective treatment of AD depends on identifying distinctive subtypes linking to specific disease mechanisms at brain cellular level. We applied polygenic risk scores defined by co-regulated cellular networks to stratify individuals into low and high-risk groups (genetic subtypes), while genes in each network were nominated as drug targets. We applied for a graph-based ranking algorithm to prioritize drug targets, while clustering algorithms using compound structures from PubChem and drug perturbation databases were used to identify existing drugs/compounds for repositioning in AD. We identified top-ranked genetic subtypes from astrocytes and oligodendrocytes with prioritized genes as drug targets for the genetic subtypes. We validated the genetic subtypes linking to previously defined AD related subgroups using domain specific cognitive scores and neuroimaging phenotypes. We identified approved drugs targeting the astrocyte subtype, all of them previously modulated expression of prioritized genes in the drug perturbation database. Clustering algorithm identified multiple compounds targeting the prioritized genes in an astrocyte subtype with at least 80% similarity in structure. In future, we will apply for advanced artificial intelligence or machine learning methods linking genetic subtypes to imaging clusters novel graph neural network approaches how genetic subtypes can predict clinical and imaging subtypes in collaboration with members from the AI4AD consortium.   

BIo

Gyungah Jun is an Associate Professor in Department of Medicine, Boston University School of Medicine. She is a genetic epidemiologist who focuses on genetic data driven precision medicine for Alzheimer’s disease (AD). Previously, she was the Director of Neurogenetics and Integrated Genomics at Eisai Inc. developing genome-guided drug discovery projects for AD. Dr. Jun has been involved in the Alzheimer’s Disease Genetics Consortium (ADGC), Alzheimer’s Disease Sequencing Project (ADSP), and the International Genomics of Alzheimer’s Project (IGAP) to identify AD risk genes within APOE genotype subgroups and across multi-ethnic populations. She is the leader of the Genome Guided Drug Discovery (GGDD) Core at the AI4AD consortium. She is a founding member of the Asian Cohort for Alzheimer’s Disease (ACAD) Consortium and a Steering Committee member at the Framingham Heart Study Brain Aging Program (FHS-BAP) and BU Alzheimer’s Disease Research Center.

Abstract

How can A.I.-based methods inform social listening applications during public health crises? The COVID-19 pandemic has uprooted the mode and method of human communication and interaction. The magnitude of the pandemic has led to an ‘infodemic’ along with considerable increase in stress and anxiety across communities. At the same time, the use of social media has increased dramatically as individuals sheltered in place. In this talk, I will introduce how our group is using big data from social media sources for contributing to social good. I will discuss the application of machine learning and natural language processing techniques to obtain insights on the heterogeneity in vaccine acceptance and mental health measurement across communities in the United States and outside.

Bio

Sharath Guntuku is an Assistant Professor in the research-track in the Department of Computer and Information Science at the University of Pennsylvania and a Senior Fellow at the Leonard Davis Institute of Health Economics. His research focuses on building predictive models for and uncovering insight into health outcomes and psychological states of individuals and communities. The goal of this research is to supplement clinical diagnoses and facilitate early and personalized interventions for improving treatments and well-being. His team develops computational models utilizing large-scale user-generated text, image, and mobile sensor data to answer questions pertaining to geospatial, cross-cultural, and temporal aspects of human behavior. His research is supported by the National Institutes of Health, Penn Global, and the World Bank Group and has been covered by the American Psychological Association, WIRED, Canadian Broadcasting Company, The Atlantic, US News, and other venues. 

Abstract

Deep Learning (DL) has the potential to optimize machine learning in both the scientific and clinical communities. However, greater expertise is required to develop DL algorithms, and the variability of implementations hinders their reproducibility, translation, and deployment. Here we present the community-driven Generally Nuanced Deep Learning Framework (GaNDLF), with the goal of lowering these barriers. GaNDLF makes the mechanism of DL development, training, and inference more stable, reproducible, interpretable, and scalable, without requiring an extensive technical background. GaNDLF aims to provide an end-to-end solution for all DL-related tasks in computational precision medicine. We demonstrate the ability of GaNDLF to analyze both radiology and histology images, with built-in support for k-fold cross-validation, data augmentation, multiple modalities and output classes.  Our quantitative performance evaluation on numerous use cases, anatomies, and computational tasks supports GaNDLF as a robust application framework for deployment in clinical workflows.

Bio

Sarthak Pati is a Researcher and Software Architect at the Perelman School of Medicine at the University of Pennsylvania, and is the MLCommons Technical Lead for GaNDLF. He holds a Master's in Biomedical Computing (TUM), and works on AI and clinical workflow management, with a focus on MIC, ML/DL, and privacy-protected algorithms for healthcare. He has additionally worked towards designing tutorials, training, managing, setting up coding guidelines, etc. for researchers to get a jump start on writing good code using academically proven scientific tools. You can find his academic profile here: https://www.med.upenn.edu/cbica/aibil/spati.html.

Abstract 
Electronic and photonic chips are essential for many systems from cell phone and computers to airplanes and medical devices. Combining electronic-photonic chips under electronic-photonic co-design concept can profoundly impact both fields resulting in advances in several areas such as energy efficient portable and implantable devices to hand-held imagers.

In this talk, I will present examples of our work on all electronic implantable devices such wireless somatosensory feedback system using human body communication, artificial mechanoreceptor, and neural signal recorder and processor with unsupervised analog classifier for spike sorting. Electronic in-vitro circulating tumor cell detection chip will also be presented. Example of photonic chips for medical applications such as photonic assisted microwave imager and optical phased arrays for optogenetics will be discussed. Finally, our work on Brain inspired electronic-photonic AI systems will be presented.

Bio

Firooz Aflatouni received the Ph.D. degree in Electrical Engineering from the University of Southern California, Los Angeles, in 2011. He was a post-doctoral scholar in the Department of Electrical Engineering at the California Institute of Technology before joining the University of Pennsylvania in 2014 where he is an Associate Professor in the Department of Electrical and Systems Engineering. His research interests include electronic-photonic co-design and low power RF and mm-wave integrated circuits for applications from communication and computation to healthcare. In 1999, he co-founded Pardis Bargh Company where he served as the CTO for five years working on design and manufacturing of inclined-orbit satellite tracking systems.

Firooz received the Bell Labs Prize in 2020, the Young Investigator Program (YIP) Award from the Office of Naval Research in 2019, the NASA Early Stage Innovation Award in 2019, and the 2015 IEEE Benjamin Franklin Key Award. He is a Distinguished Lecturer of the Solid-State Circuit Society and has served on several IEEE program committees (ISSCC, CICC, and IMS). He is an Associate Editor of the IEEE Open Journal of the Solid-State Circuits Society and currently serves as the chair of IEEE Solid State Circuits Society (SSCS) Philadelphia chapter.