2025 Biomedical Informatics Seminars
April 11, 2025
Vikram Agarwal, PhD, Head of mRNA Platform Design Data Science | mRNA Center of Excellence, Sanofi
Abstract: One of the core principles of post-transcriptional gene regulation occurs at the step of translation, during which the cell tunes the number of proteins produced per mRNA molecule. The degree to which translational control is specified by mRNA sequence is poorly understood. Here, we collected a compendium of 2,894 mammalian ribosomal profiling datasets, distilling them into a transcriptome-wide atlas of translation efficiency measurements representing 78 human and 68 mouse cell types. We developed RiboNN, a multi-task deep convolutional neural network, and classic machine learning models to demonstrate that sequence-encoded mRNA features were sufficient to strongly predict translation efficiency. We further interpreted RiboNN to give mechanistic insight into how mRNA sequence features influence translation rates.
Bio: Vikram Agarwal completed his Ph.D. in Dr. David Bartel's lab at MIT, and his post-doc in Dr. Jay Shendure's lab at the University of Washington. There, he applied deep learning methods and massively parallel reporter assays to investigate the mechanisms of transcriptional gene regulation, further building upon these approaches at Calico Life Sciences. He is currently the Head of mRNA Platform Design Data Science at the mRNA Center of Excellence at Sanofi, where he is applying machine learning and deep learning methods towards the design of enhanced mRNA therapeutics.
Abstract: One of the core principles of post-transcriptional gene regulation occurs at the step of translation, during which the cell tunes the number of proteins produced per mRNA molecule. The degree to which translational control is specified by mRNA sequence is poorly understood. Here, we collected a compendium of 2,894 mammalian ribosomal profiling datasets, distilling them into a transcriptome-wide atlas of translation efficiency measurements representing 78 human and 68 mouse cell types. We developed RiboNN, a multi-task deep convolutional neural network, and classic machine learning models to demonstrate that sequence-encoded mRNA features were sufficient to strongly predict translation efficiency. We further interpreted RiboNN to give mechanistic insight into how mRNA sequence features influence translation rates.
Bio: Vikram Agarwal completed his Ph.D. in Dr. David Bartel's lab at MIT, and his post-doc in Dr. Jay Shendure's lab at the University of Washington. There, he applied deep learning methods and massively parallel reporter assays to investigate the mechanisms of transcriptional gene regulation, further building upon these approaches at Calico Life Sciences. He is currently the Head of mRNA Platform Design Data Science at the mRNA Center of Excellence at Sanofi, where he is applying machine learning and deep learning methods towards the design of enhanced mRNA therapeutics.
April 4, 2025
Jonathan Chen, MD, PhD, Assistant Professor of Medicine and Biomedical Data Science | Faculty Director for Medical Education in AI |Stanford Center for Biomedical Informatics Research, Stanford University
Abstract: Pandora’s box has opened in the form of publicly available generative AI systems for every imaginable (and many unintended) purposes. With a global scarcity of medical expertise against the unlimited demand of people in need, AI's potential to democratize healthcare knowledge, access, and to recover efficiencies is desperately needed. The implications are vast as we converge upon a point in history where human vs. computer generated content can no longer be reliably distinguished. This session will review the attention and intention required for AI applications in the high-stakes world of healthcare as we distinguish real magic from convincing illusions.
Bio: Jonathan H. Chen MD, PhD leads a research group to empower individuals with the collective experience of the many, combining human and artificial intelligence approaches to deliver better care than either alone. Dr. Chen continues to practice medicine for the concrete rewards of caring for real people and to inspire this research focused on discovering and distributing the latent knowledge embedded in clinical data.
Jonathan Chen, MD, PhD, Assistant Professor of Medicine and Biomedical Data Science | Faculty Director for Medical Education in AI |Stanford Center for Biomedical Informatics Research, Stanford University
Abstract: Pandora’s box has opened in the form of publicly available generative AI systems for every imaginable (and many unintended) purposes. With a global scarcity of medical expertise against the unlimited demand of people in need, AI's potential to democratize healthcare knowledge, access, and to recover efficiencies is desperately needed. The implications are vast as we converge upon a point in history where human vs. computer generated content can no longer be reliably distinguished. This session will review the attention and intention required for AI applications in the high-stakes world of healthcare as we distinguish real magic from convincing illusions.
Bio: Jonathan H. Chen MD, PhD leads a research group to empower individuals with the collective experience of the many, combining human and artificial intelligence approaches to deliver better care than either alone. Dr. Chen continues to practice medicine for the concrete rewards of caring for real people and to inspire this research focused on discovering and distributing the latent knowledge embedded in clinical data.
Before his medical training, Chen co-founded a company to translate his Computer Science graduate work into an expert system with applications from drug discovery to an education tool for students around the world. His expertise is regularly featured in popular press outlets with over 100 publications in leading clinical and informatics venues and awards from the NIH, National Library of Medicine, American Medical Informatics Association, International Brotherhood of Magicians and more.
In the face of ever escalating complexity in medicine, informatics solutions are the only credible approach to systematically address challenges in healthcare. Tapping into real-world clinical data like electronic medical records with machine learning and data analytics will reveal the community's latent knowledge in a reproducible form. By delivering this back to clinicians, patients, and healthcare systems as clinical decision support, he aims to uniquely close the loop on a continuously learning health system.
March 14, 2025
Monica Agrawal, PhD, Assistant Professor of Biostatistics & Bioinformatics | Assistant Professor of Computer Science | Assistant Professor of Biomedical Engineering | Division of Translational Biomedical Informatics, Department of Biostatistics & Bioinformatics, Duke University School of Medicine.
Abstract: Language is ubiquitous in healthcare from clinical notes to medical literature to online health information. However, clinical natural language processing (NLP) often faces challenges because notes are written in their own jargon-heavy dialect, patient histories can contain hundreds of notes, and there is often minimal labeled data available. In this talk, I will discuss scalable natural language processing (NLP) solutions to overcome these technical challenges. These include the development of novel techniques for leveraging large language models, a new paradigm for EHR documentation that incentivizes the creation of high-quality data at the point-of-care, and a new framework for thinking about health search engines. I will end by discussing open challenges and opportunities for NLP to impact a variety of healthcare workflows.
Bio: Dr. Monica Agrawal is an assistant professor at Duke University, jointly appointed between the Department of Biostatistics and Bioinformatics and the Department of Computer Science. Her research tackles diverse challenges including scalable clinical information extraction, smarter electronic health records, and human-AI interaction. She has been named a Duke Whitehead Scholar, a Rising Star in EECS, and a finalist for the AMIA Doctoral Dissertation award. Dr. Agrawal earned her PhD in Computer Science at MIT in 2023 and is also a co-founder of Layer Health.
February 28, 2025
Andrew Wong, MD, Research Fellow | National Clinician Scholars Program, Institute for Healthcare Policy and Innovation, Clinical Instructor | Department of Internal Medicine, University of Michigan Health System. "Governance of Clinical Prediction Models in the Era of AI."
Bio: Dr. Wong's research applies artificial intelligence (AI) and machine learning to solve problems in clinical practice, hospital operations, and medical education. His expertise in the development and validation of provider-facing AI tools to aid in clinical diagnosis and prognostication has been applied to sepsis prediction, delirium prevention, antibiotic stewardship, and more. His focus in post-implementation AI governance has sought to safely integrate clinical AI tools in an effective and equitable manner. His work has been nationally recognized and was recently cited in the White House Blueprint for an AI Bill of Rights.
February 21, 2025
Biren Kamdar, MD, MBA, MHS, Associate Professor, Division of Pulmonary, Critical Care, and Sleep Medicine, UC San Diego Health. "Intensive Care after COVID: Challenges and Opportunities."
Bio: Biren Kamdar, MD, MBA, MS, MHS, is a board-certified pulmonologist and critical care physician. He cares for adult patients in the Medical Intensive Care Unit (MICU) and those with general lung conditions.
As an associate professor in the Department of Medicine, Dr. Kamdar trains medical students, residents and fellows at the UC San Diego School of Medicine. His NIH/NIA-funded research focuses on sleep and circadian rhythms in the ICU; in particular, methods to evaluate sleep in critically ill patients and the effect of interventions to improve sleep-wake cycles on delirium and other important outcomes. Dr. Kamdar has presented on this topic at national and international conferences, and has published in various medical journals and textbooks.
Prior to joining UC San Diego Health, Dr. Kamdar was an assistant professor in the Division of Pulmonary and Critical Care Medicine at the David Geffen School of Medicine at UCLA. Dr. Kamdar completed a fellowship in pulmonary and critical care medicine at Johns Hopkins Hospital (Johns Hopkins School of Medicine) in Baltimore. During his fellowship, he received an NIH/Kirschstein NRSA Award and a Master in Health Science (MHS) degree from the Graduate Training Program in Clinical Investigation at the Johns Hopkins Bloomberg School of Public Health. He completed his internal medicine residency at Vanderbilt University Medical Center. He received a joint MD/MBA at the Vanderbilt University School of Medicine and the Vanderbilt Owen Graduate School of Management in Nashville.
Biren Kamdar, MD, MBA, MHS, Associate Professor, Division of Pulmonary, Critical Care, and Sleep Medicine, UC San Diego Health. "Intensive Care after COVID: Challenges and Opportunities."
Bio: Biren Kamdar, MD, MBA, MS, MHS, is a board-certified pulmonologist and critical care physician. He cares for adult patients in the Medical Intensive Care Unit (MICU) and those with general lung conditions.
As an associate professor in the Department of Medicine, Dr. Kamdar trains medical students, residents and fellows at the UC San Diego School of Medicine. His NIH/NIA-funded research focuses on sleep and circadian rhythms in the ICU; in particular, methods to evaluate sleep in critically ill patients and the effect of interventions to improve sleep-wake cycles on delirium and other important outcomes. Dr. Kamdar has presented on this topic at national and international conferences, and has published in various medical journals and textbooks.
Prior to joining UC San Diego Health, Dr. Kamdar was an assistant professor in the Division of Pulmonary and Critical Care Medicine at the David Geffen School of Medicine at UCLA. Dr. Kamdar completed a fellowship in pulmonary and critical care medicine at Johns Hopkins Hospital (Johns Hopkins School of Medicine) in Baltimore. During his fellowship, he received an NIH/Kirschstein NRSA Award and a Master in Health Science (MHS) degree from the Graduate Training Program in Clinical Investigation at the Johns Hopkins Bloomberg School of Public Health. He completed his internal medicine residency at Vanderbilt University Medical Center. He received a joint MD/MBA at the Vanderbilt University School of Medicine and the Vanderbilt Owen Graduate School of Management in Nashville.
February 7, 2025
Karandeep Singh, MD, MMSc, Joan and Irwin Jacobs Chancellor’s Endowed Chair, Associate Professor of Medicine in Biomedical Informatics, Chief Health AI Officer and Associate CMIO for Inpatient Care, UC San Diego Health, "Tidier.jl: Lessons from building a data analysis ecosystem in Julia."
Abstract: Every biomedical informatics research project starts with the selection of data analysis tools. Common tools for working with tabular data include SQL, R’s tidyverse and data.table packages, and Python’s pandas and polars packages. While we often choose tools based on convenience, there are substantial differences in performance, syntax, and philosophies for each of these tools. Additionally, all of these tools suffer from the two-language problem, whereby R and Python packages wrap C, C++, and Rust code to create data analysis ecosystems that are fast and friendly to use but difficult to extend without specialized knowledge. Julia is a relatively new programming language that aims to solve this problem, making it possible to write code that is both simple and fast. In 2023, I launched an effort to recreate R’s tidyverse ecosystem entirely in Julia under the name of Tidier.jl. In the past two years, Tidier.jl has accumulated >500 stars on GitHub. In this talk, I’ll talk about the pros and cons of different data analysis ecosystems, and share lessons learned from translating an ecosystem from one programming language to another.
Bio: Karandeep Singh, MD, MMSc is the Joan and Irwin Jacobs Chancellor’s Endowed Chair in Digital Health Innovation and Associate Professor in Biomedical Informatics at UC San Diego, where he also serves as Chief Health AI Officer for UC San Diego Health. In these roles, Dr. Singh leads AI initiatives within the Jacobs Center for Health Innovation and oversees AI strategy and governance for the health system.
He completed his internal medicine residency at UCLA Medical Center, where he served as chief resident, and a nephrology fellowship in the combined Brigham and Women’s Hospital and Massachusetts General Hospital program. He completed his medical education at the University of Michigan Medical School and holds a master’s degree in medical sciences in Biomedical Informatics from Harvard Medical School.
Karandeep Singh, MD, MMSc, Joan and Irwin Jacobs Chancellor’s Endowed Chair, Associate Professor of Medicine in Biomedical Informatics, Chief Health AI Officer and Associate CMIO for Inpatient Care, UC San Diego Health, "Tidier.jl: Lessons from building a data analysis ecosystem in Julia."
Abstract: Every biomedical informatics research project starts with the selection of data analysis tools. Common tools for working with tabular data include SQL, R’s tidyverse and data.table packages, and Python’s pandas and polars packages. While we often choose tools based on convenience, there are substantial differences in performance, syntax, and philosophies for each of these tools. Additionally, all of these tools suffer from the two-language problem, whereby R and Python packages wrap C, C++, and Rust code to create data analysis ecosystems that are fast and friendly to use but difficult to extend without specialized knowledge. Julia is a relatively new programming language that aims to solve this problem, making it possible to write code that is both simple and fast. In 2023, I launched an effort to recreate R’s tidyverse ecosystem entirely in Julia under the name of Tidier.jl. In the past two years, Tidier.jl has accumulated >500 stars on GitHub. In this talk, I’ll talk about the pros and cons of different data analysis ecosystems, and share lessons learned from translating an ecosystem from one programming language to another.
Bio: Karandeep Singh, MD, MMSc is the Joan and Irwin Jacobs Chancellor’s Endowed Chair in Digital Health Innovation and Associate Professor in Biomedical Informatics at UC San Diego, where he also serves as Chief Health AI Officer for UC San Diego Health. In these roles, Dr. Singh leads AI initiatives within the Jacobs Center for Health Innovation and oversees AI strategy and governance for the health system.
He completed his internal medicine residency at UCLA Medical Center, where he served as chief resident, and a nephrology fellowship in the combined Brigham and Women’s Hospital and Massachusetts General Hospital program. He completed his medical education at the University of Michigan Medical School and holds a master’s degree in medical sciences in Biomedical Informatics from Harvard Medical School.
January 31, 2025
Mohammad Ghassemi, Phd, Assistant Professor, Computer Science and Engineering, Adjunct Assistant Professor of Medicine, Division of Health Data Science, Michigan State University, Bridging the Gap: Enhancing LLM Performance in Low-Resource Languages with New Benchmarks, Fine-Tuning, and Cultural Adjustments.
Abstract: Large Language Models (LLMs) have shown remarkable performance across various tasks, yet significant disparities remain for non-English languages, and especially native African languages. In this talk, we will discuss our work to addresses these disparities by creating approximately 1 million human-translated words of new benchmark data in 8 low-resource African languages, covering a population of over 160 million speakers of: Amharic, Bambara, Igbo, Sepedi (Northern Sotho), Shona, Sesotho (Southern Sotho), Setswana, and Tsonga. Our benchmarks are translations of Winogrande and three sections of MMLU: college medicine, clinical knowledge, and virology. Using the translated benchmarks, we report previously unknown performance gaps between state-of-the-art (SOTA) LLMs in English and African languages. Finally, using results from over 400 fine-tuned models, we explore several methods to reduce the LLM performance gap, including high-quality dataset fine-tuning (using an LLM-as-an-Annotator), cross-lingual transfer, and cultural appropriateness adjustments. The publicly available benchmarks, translations, and code from this study support further research and development aimed at creating more inclusive and effective language technologies.
Bio: Dr. Mohammad Ghassemi is a scientist, entrepreneur, and educator with a Ph.D. from MIT in computer science, specializing in AI. As a founding partner at Ghamut Corporation and a former director of data science at S&P Global and strategic consultant with BCG, he has over 15 years of experience leading AI initiatives for global organizations. He is a Professor of Computer Science at Michigan State University, where he directs the Human Augmentation and Artificial Intelligence Lab, advancing tools that integrate human and machine intelligence. A National Scholar for Data and Technology Advancement at NIH and recipient of numerous accolades, including AIMed's AI Champion award, Dr. Ghassemi is an inventor on multiple U.S. patents and author of over 70 peer-reviewed papers cited more than 10,000 times. His work has been featured in outlets like BBC, NPR, and The Wall Street Journal.
January 17, 2025
Brian Clay, MD, Professor of Medicine, Associate Chief Medical Officer for Inpatient Care, UC San Diego Health, Leveraging Informatics and the Electronic Medical Record to Support Hospital Patient Flow.
Abstract: UC San Diego Health has experience substantial growth over the last five years, including a level of demand for hospital beds that has exceeded our hospital capacity. The health system has established a Mission Control center to assist with both strategically and tactically managing hospital capacity, with the goals of (1) increasing efficiency of hospital throughput for inpatients and (2) distributing hospital patients efficiently among our several hospital campuses. This presentation will review the work of Mission Control over the last year in pursuit of these goals, with a specific emphasis on the Epic electronic medical record and our Tableau enterprise business intelligence application.
Bio: Dr. Brian Clay is a hospital medicine physician at UC San Diego Health since 2003. He has previously served for 7 years as an associate director for the Internal Medicine residency program and the Medicine 401 core clerkship; he is currently an associate director of the UCSD clinical informatics fellowship. Administratively, Dr. Clay was the Chief Medical Information Officer at UCSD Health from 2012 to 2022, during which time he oversaw multiple electronic medical record implementations and optimization projects. He currently serves as the Associate Chief Medical Officer for Inpatient Care and the physician director of Mission Control; his current work is focused on hospital throughput and capacity management.
January 10, 2025
Greg Appelbaum, PhD, Professor, Department of Psychiatry and Behavioral Sciences, University of California, San Diego, Head: Human Performance Optimization Lab (OptiLab), Director of Research, UCSD Interventional Psychiatry Program, Building a psychiatric clinical decision support system for Interventional Psychiatry.
Abstract: Interventional psychiatry is a subspecialty that uses rapid-acting neuromodulation and pharmaceutical interventions, such as transcranial magnetic stimulation (TMS), electroconvulsive therapy (ECT), and ketamine therapy, to treat psychiatric conditions, particularly those resistant to traditional therapies. While each intervention is individually successful, there is considerable heterogeneity across individual patient’s responses and therefore research has worked towards identifying predictive and prescriptive measures that can be used to guide clinical decision-making. This talk explores the creation of a psychiatric clinical decision support system (CDSS) designed to personalize and enhance interventional treatments through the integration of clinical, neuroimaging, and biomarker data. We will discuss its development, validation, and the transformative potential for improving care pathways and outcomes in disorders like depression and bipolar disorder, paving the way for precision mental health care.
Bio: Greg Appelbaum is a Professor in the Department of Psychiatry at the University of California, San Diego and Research Director of the UCSD Interventional Psychiatry Program. He received a B.A in Psychology from Emory University in 1995, and a Masters and Ph.D. from the Department of Cognitive Science at the University of California, Irvine in 2002 and 2004, respectively.
Abstract: Interventional psychiatry is a subspecialty that uses rapid-acting neuromodulation and pharmaceutical interventions, such as transcranial magnetic stimulation (TMS), electroconvulsive therapy (ECT), and ketamine therapy, to treat psychiatric conditions, particularly those resistant to traditional therapies. While each intervention is individually successful, there is considerable heterogeneity across individual patient’s responses and therefore research has worked towards identifying predictive and prescriptive measures that can be used to guide clinical decision-making. This talk explores the creation of a psychiatric clinical decision support system (CDSS) designed to personalize and enhance interventional treatments through the integration of clinical, neuroimaging, and biomarker data. We will discuss its development, validation, and the transformative potential for improving care pathways and outcomes in disorders like depression and bipolar disorder, paving the way for precision mental health care.
Bio: Greg Appelbaum is a Professor in the Department of Psychiatry at the University of California, San Diego and Research Director of the UCSD Interventional Psychiatry Program. He received a B.A in Psychology from Emory University in 1995, and a Masters and Ph.D. from the Department of Cognitive Science at the University of California, Irvine in 2002 and 2004, respectively.
Professor Appelbaum’s research is focused on understanding the psychological and neural mechanisms that support human cognition and understanding how these change with experience, rehabilitation, and training. This research utilizes behavioral psychometrics and multiple complimentary human neuroimaging (e.g. fMRI, EEG, fNIRS) and neurostimulation (e.g. TMS, ECT) techniques to understand the the processes and mechanisms enabling perception, attention, memory, emotion, and motor control. As a faculty, his research has been continually funded by grant awards from DARPA, NIH, and the Army Research Office, leading to over 110 published articles and book sections.
Dr. Appelbaum mentors a wide assortment of junior faculty, research scientists, postdoctoral fellows, PhD students from the UCSD Neuroscience Graduate program and the UCSD/SDSU Joint Doctoral Program in Clinical Psychology, research coordinators, and undergraduate students. He was awarded the 2023 UCSD, Graduate and Professional Student Association, Outstanding Faculty Mentor Award.
Dr. Appelbaum mentors a wide assortment of junior faculty, research scientists, postdoctoral fellows, PhD students from the UCSD Neuroscience Graduate program and the UCSD/SDSU Joint Doctoral Program in Clinical Psychology, research coordinators, and undergraduate students. He was awarded the 2023 UCSD, Graduate and Professional Student Association, Outstanding Faculty Mentor Award.