Dec. 6, Karen Becerra, MPH, DDS - Clinical Project Lead, West Health Instutite, La Jolla, CA
Wireless Remote Monitoring of High Risk Pregnancies
Dr. Becerra is the lead research for the Sense4baby project at West Health. Sense4baby is a small portable fetal monitoring system paired with a smart phone that allows fetal monitoring and management of high risk pregnancies anywhere cellular or wireless internet is available. Dr. Becerra will share the development, clinical validation, and system deployment of Sense4baby.
Nov. 22, Susan Little, MD - Professor, Department of Medicine, Division of Infectious Diseases, UCSD
Strategies for HIV Epidemic Control
Antiretroviral treatment (ART) markedly reduces the risk of HIV transmission between stable HIV discordant heterosexual partners and increasing ART use can effect “population-level” reductions in HIV incidence. However, current community cluster randomized trials to determine if widespread ART can significantly reduce HIV transmission rates in affected communities (i.e., “treatment as prevention” [TasP]) require immense resources and sample sizes ranging from 12,000 – 60,000 persons. Additionally, population-level strategies may not provide an assessment of individual HIV transmission risk. Alternative, less resource-prohibitive approaches are needed to test the efficacy of TasP and other prevention interventions. Ideally these methods would also provide real-time insight into HIV transmission dynamics and be easily modified to address unique regional and epidemiological factors. Network-focused prevention interventions are explored as a means to more efficiently reduce network incidence. These data highlight the dynamics of HIV spread and importance of high degree nodes within this network for prevention intervention efforts.
Nov. 15, Zia Agha, MD, MS - Associate Clinical Professor, Department of Medicine, UCSD; Director of Health Services Research at VA San Diego
EHR usability and clinical work - time to put the horse in front of the cart
EHRs have potential to improve patient safety and clinical efficiency but some studies suggest that EMRs integrate poorly in clinical workflow and have a number of usability issues. In particular, EHRs must support efficient documentation and information retrieval of patient information at the point of care. Heath and Luff argued that “the practices through which the document is written, read and used within consultation have been largely ignored.” We propose to address this knowledge gap by studying how clinicians use EHR functions (eg CPRS Notes) for documentation and retrieval of clinical information, how they organize their notes (eg, SOAP; by condition) and in particular to understand quantitatively the degree and sources of redundancy in patients progress notes over time. Findings from our research (PACE study) suggests that approximately half of VA primary care providers’ EHR workflow is related to the CPRS Notes function (reflecting both documentation and information retrieval tasks), making this documentation component a good candidate for redesign. We propose to explore the notion of an interactive progress note interface (aNotes) that leverages the advantages in terms of searchability, indexing, and cross-validation of modern EMRs, but introduces a novel interface based on the notion of the old paper-based progress note. aNotes are electronic progress notes that are integrated into a modern EMR, but keep the concept of the patient history and the structure of a note as the central way to enter, access and retrieve information. Physicians never leave a note when accessing or entering patient data, and always use the note to retrieve old data, test results, appointments, consultation, vital signs, medications, etc In summary we want to give to physicians a novel interface to the EMR that is simple and easy to work, enable fast and intuitive access to data.
Nov 8, Mary Devereaux, PhD - Director, Biomedical Ethics Seminar & Assistant Director, Research Ethics Program & San Diego Research Ethics Consortium, UCSD
The Use of Patient Records (EHR) for Research
The growing availability of electronic medical records, and large databases of health information such as EPIC, afford researchers a range of opportunities. Computerized searches of large troves of de-identified patient medical information promise a better understanding of disease patterns, treatment efficacy, and the contribution of genetic and other factors in health. These possibilities, however, raise a number of ethical, legal, and regulatory challenges. What does de-identification mean? Can it be done, and with what assurance? Do patients own their medical information/their charts — with the right to refuse consent for its use — or is such information a shared social resource available to all? What are the responsibilities of biomedical researchers using EHRs and what does the IRB require? Download slides
Oct 25, Sergei Pond, PhD - Associate Professor & Director, Bioinformatics Core, Center for AIDS Research UCSD
Promises and Challenges of Next Generation Sequencing for HIV
Next generation sequencing (NGS) technologies are rapidly reshaping the fields of medical diagnostics, molecular virology, and sequence analysis. The ultimate transformative potential of NGS in the field of HIV/AIDS research is critically dependent on the bioinformatics tools, analytical methods, and the availability of clinically relevant interpretations. I'll describe several key research applications of NGS in HIV research, including drug resistance, vaccine research, and molecular epidemiology and explain some of challenges faced by bioinformaticians who analyze and interpret these data.
Oct 18, Claudiu Farcas, PhD - Assistant Research Scientist, California Institute for Telecommunication and Information Technology (Calit2), UCSD
Scaling Science through Software Engineering
The discipline of biomedical computing and informatics has made colossal leaps in transforming healthcare through science and informatics in the past ten years. Yet, we witness widespread increase of health-care costs, limited availability of vital and critical data during emergency care, slow penetration of IT within the “last-mile” doctor-patient relationship, and severe lack of integrated information around each individual. We will discuss the role of software engineering principles and techniques to address these challenges and open up a new era of advancements through the adoption of massive computing and big-data solutions to major scientific problems, along with the corresponding shift from single purpose, monolithic systems to large-scale, often high performance, systems-of-systems.
Oct 11, Danielle L. Mowery, MS - PhD candidate, University of Pittsburgh
Developing Natural Language Processing Resources for Stakeholders: Use Cases for Clinical Researchers and Care Providers
Stakeholders such as clinical researchers, care providers, and ailing patients are becoming increasingly interested in information locked in clinical free-texts stored within electronic medical records. For each stakeholder, natural language processing (NLP) systems can unlock and provide utility of this rich clinical information in various ways. For clinical researchers, NLP systems can be used to redact Protected Health Information (PHI) according to HIPAA and encode texts with study variables to answer population-level research questions. For care providers, NLP systems can automatically identify problems and their statuses to generate patient problem lists. For ailing patients, NLP systems can be leveraged to encode unfamiliar medical concepts and abbreviations with more consumer-friendly terms to promote easier readability and understanding of clinical texts.
In this presentation, I will review two collaborative projects aimed toward facilitating the development of natural language processing resources - datasets and tools – for use cases that can aid clinical researchers and care providers. For the first project, I will describe the development and evaluation of an annotated, de-identified corpus of clinical texts that can be leveraged to develop, train, or test NLP de-identification systems. For the second project, I will describe the development and evaluation of a clinical information model that can be leveraged to generate patient problem lists from clinical texts.
Oct 4, Chun-Nan Hsu, PhD - Associate Professor, Division of Biomedical Informatics, UCSD
Scaling up learning of a semantic distance metric for phenotype variable mapping
Determining semantic similarity is an important shared task in Biomedical NLP. For example, it is essential when we need to map an input text to standardized terminology, coding, or classification, match query terms to data for data retrieval, or link terms together for data integration. These are important functions for killer apps of BioNLP. Semantic similarity can be learned from training examples of semantically similar and dissimilar text. However, this usually requires a very high dimensional representation. In this talk, I present some effective techniques, including random projection, to scale up learning of semantic similarity from very high dimensional data. Our result shows that it is now possible to learn accurate semantic similarity metric with nearly 20K dimensions from nearly 100K pairs of semantically similarity pairs of text with a laptop computer in 15 minutes while accomplishing very high area under curve scores. The same learning task will run out of memory with no attempt of scaling up.
Sep 27, Staal A. Vinterbo, PhD - Associate Professor, Division of Biomedical Informatics, UCSD
Privacy and information release
Breaches of individual privacy are unfortunately a common occurrence, mainly caused by loss or theft of storage equipment containing unencrypted personal information. A different and more subtle form of breach occurs when personal information can be inferred from data and information that has been deemed safe and consequently disclosed. Examples of such are the Netflix prize rating data set where ratings could be matched with IMDB data, the AOL search log data set from which a New York Times reporter identified Thelma Arnold, and the successful matching of individuals in complex mixtures of genotyping microarray data by Homer et al. Traditionally, this inferential type of breach is sought eliminated by requiring data to be anonymous, i.e., disallowing the association of data with a unique identifier. This is evident in the definitions of personal data in the EU data protection directive and protected health information in the US HIPAA Privacy Rule. It is trivial to be perfectly safe by not disclosing anything. However, real life does require the use of human subjects data.
In this talk, I will discuss quantitative analysis of privacy risk, and how traditional approaches based on anonymity and definition of privacy that is based on properties of data, so-calledsyntactic definitions, make this difficult. As an alternative I will present Differential Privacy which is a measure of privacy risk based on properties of the process by which information is extracted from data. Finally, I will present recent results of mine in this context.
June 7, Beverly Morris, RN-NP, MBA - Department of Nursing Education, Development & Research, UCSD Medical Center
100% Change or Rain
May 31, Li-Fan Lu, PhD - Division of Biological Science, UCSD
miRNA in immune regulation
Abstract: Recent studies investigating the role of a subset of small regulatory RNA species – microRNAs – suggested that they can act as important molecular regulators in the immune system. Here, through employing genetic, biochemical, immunological approaches and whole animal experimentation, we aimed to explore the cellular and molecular mechanisms by which such regulatory RNA species control our immune responses with the goal of developing new therapeutic means for treating human immunological diseases. Unpublished work - contact the speaker for more information.
May 24, Olivier Harismendy, PhD - Department of Pediatrics, UCSD
Diving Deeper into DNA: Genomic Approaches for Cancer Research and Clinical Care
Abstract: Recent technological advances allow the affordable sequencing of a large number of genes, an entire human exome or a whole genome. I will introduce these advances and how they are applied to transform cancer research, diagnostics and care. I will describe the challenges of both technical and clinical implementation and present the practical initiatives chosen to facilitate the wider adoption of these strategies in clinical research and patient care.Unpublished work - contact the speaker for more information.
May 17, Alexander Zambon, PhD - Department of Pharmacology and Medicine, UCSD
Use of Comparative Genomics and Pathway Analysis to Extract Testable Hypoitheses from -Omics Data
Abstract: High content -omics methodologies often result in "mountains" of data that are difficult to analyze for basic scientists. These datasets often require sophisticated computational and bioinformatic methods to efficiently extract meaningful information for the development of testable hypotheses. We have developed several bioinformatics tools (GO-Elite) and pipelines (Whole Genome RVista, and InSilico ChIP) to facilitate analysis of -omics datasets geared towards basic scientists. These tools will be discussed in the context of their application to a significant unanswered question in cardiovascular biology: What signaling pathways mediate perinatal cardiac myocyte cell cycle arrest? Cardiac developmental malformation and disease are leading causes of death and disability worldwide and can be attributed, in large part, to the inability of adult cardiac myocytes to divide (proliferate) in response to loss or damage. Using tools that we have developed we have identified new transcriptional signaling circuits that regulate this important process during development. Unpublished work - contact the speaker for more information.
May 10, Marcio von Muhlen, PhD - Director of Product Management, Doximity
Doximity: Growing a Physician Professional Network to 20% of US Docs
Abstract: After postdoc'ing with DBMI at UCSD in 2010, and studying the emerging research area of social media use by physicians, I joined an internet startup called Doximity vying to become the "LinkedIn for doctors". Doximity is unique in limiting membership to verified healthcare providers and providing free HIPAA-secure messaging and faxing tools through web and mobile applications. Recently Doximity has grown to include over 20% of US physicians. I will share some of our learnings growing the network and my personal experience in a startup environment.
May 3, Mary McNamara - Department of Bioengineering, UCLA
Patient Portals and the Potential for Personalized Supporting Health Information
Abstract: Patients searching for health information to complement their knowledge of their health and diagnoses have a variety of sources at their disposal. However not all information is relevant to an individual patient and some of the most quality sources of information are written for professionals. Although patients may be accessing information, they are not necessarily viewing content applicable to their health or understanding information within. Using practitioner guidelines, consumer health content, and patient information needs and preferences, we designed an information model to provided patients access to their personal health information with the ability to link to individualized relevant supporting information to a patient record content.
April 26, Christopher Benner, PhD - Director of Integrative Genomics and Bioinformatics Core, Salk Institute
Leveraging Next-Generation Sequencing to Transform Our Understanding of Transcriptional Networks and Regulatory DNA
Abstract: Recent efforts to find genetic variants associated with disease have revealed that over 90% of these variants are found in non-coding regions of the genome. Many of these regions encode regulatory information responsible for specifying the spatial-temporal expression of genes throughout the body, making it imperative that we gain a deep understanding of biochemical mechanisms and the DNA codes that specify their action. Advances in sequencing technology and creativity in its use have revealed features of regulatory elements at unprecedented scope and detail. I will describe how we have used next-gen sequencing methods to map chromatin states, transcription factors, chromatin structure and nascent transcription to build a comprehensive understanding of how the cellular machinery interprets the genome to drive gene expression. I will also describe how we used 5’ GRO-Seq, a method we developed to assay the transcriptional initiation of nascent transcripts, to reveal tens of thousands of promoters for non-coding RNAs that are undetectable with traditional methods. Together these data highlight a close relationship between transcriptional activity, DNA regulatory elements, and genomic structure, and provide a foundation to understand how changes in regulatory DNA may influence disease. Unpublished work - contact the speaker for more information.
April 19, Xiaoqian Jiang, PhD - Post-doctoral fellow, Division of Biomedical Informatics, UCSD
A SCANNER Affiliated Project: Grid Logistic REgression (GLORE)
Abstract: Traditional approaches to data sharing have limitation when sensitive clinical data are involved. To address the challenge in data analysis across distributions without undermining the utility, we proposed a novel framework to share model without sharing data. As an instance of the SCANNER framework, Grid Logistic Regression (GLORE) was developed to build a global logistic regression model from distributed data repositories. Leveraging a distributed Newton-Ralphson algorithm, we combined summary statistics from individual data repositories iteratively to construct an accurate logistic regression model as if it is trained using data collected in one central repository. Experimental results showed accurate prediction and reduced computational time as a result of applying the distributed algorithm.
April 12, Trey Ideker, PhD - Professor and Chief, Division of Medical Genetics, School of Medicine, UCSD
Network Guided Stratification of Tumor Genomes
Abstract: Many types of cancer are stratified into multiple subtypes with differing molecular causes. Somatic tumor genomes provide a rich new source of data for identifying these subtypes, but they have proven difficult to compare as two tumors rarely share the same mutations. Here, we introduce a method called ‘network-based stratification’ which integrates somatic tumor genomes with gene networks. This approach allows for stratification of cancer into informative subtypes by clustering together patients who have mutations within similar network regions. We show that network-based stratification of ovarian tumor exomes results in subtypes that are predictive of patient survival and response to therapy, in contrast to subtypes based on other types of molecular profiles. The most aggressive subtype associates specifically with mutations in networks related to the Fibroblast Growth Factor Receptor, the nucleoskeleton, caspases, and protein transport. Subtypes are used to train an mRNA expression signature which provides similar information in the absence of DNA sequence. Unpublished work - contact the speaker for more information
April 5, Andrew Su, PhD - Associated Professor, Department of Molecular and Experimental Medicine, Scripps Research Institute
Crowd-sourcing Biology: the Gene Wiki, BioGPS and GeneGames.org
Comprehensively annotating the function of human genes is a formidable challenge for the biomedical research community. Current efforts to organize biological knowledge are driven by a few centralized teams of curators and developers. Here, we describe several efforts to engage the entire biomedical research community in addressing this challenge. The Gene Wiki focuses on building a gene-specific review article for every human gene. BioGPS aims to build a community-maintained gene annotation portal. And a suite of games at GeneGames.org addresses a variety of biomedical research goals using biological games.
1/11/2013 Marios Gavrielides, PhD - Food and Drug Administration
Quantitative assessment of change in lung tumor size
Estimated changes in lung nodule size are clinically used to confirm diagnosis and to monitor the response of lung cancer patients to treatment. There are a number of factors affecting the uncertainty of lung nodule size estimation including imaging parameters, nodule characteristics, the choice of size metrics, and the performance of software tools. In my talk, I will describe the work we are currently conducting at the FDA and in collaboration with the Quantitative Alliance for Imaging Biomarkers (QIBA) to quantify these factors. The answers will help us to determine the bounds of performance where we can confidently distinguish between a true tumor response and measurement error. At the same time, we aim to optimize methodologies and protocols for size estimation so that tumor response can be detected earlier.
1/18/2013 Jane Georges, PhD - University of San Diego
The Epistemology of Suffering: Application to Biomedical Informatics
Epistemology, understood as the study of knowledge, explores what knowledge is, how it is acquired, and the extent to which a subject may be known. Recent advances in the sharing of biomedical information have shifted the boundaries regarding the type and quality of biomedical knowledge possessed by patients, families, and health care professionals. Social processes surrounding health care, including the Health Insurance Portability and Accountability Act (HIPAA) of 1996, were designed ostensibly to protect the privacy of individually identifiable health information. However, the unintended consequences of such social processes may include the actual amplification of suffering in patients and families now faced with fear of disclosure of specific health information to selected entities. In addition, the ethical dilemmas associated with navigating who really “knows” what about a specific patient has become part of the landscape of contemporary U.S. health care for health care professionals, including physicians, nurses, and social workers. This presentation utilizes exemplar narratives of suffering experienced by patients, family members, and health care professionals in relationship to the knowledge of biomedical information, with accompanying ethical analyses.
1/25/2013 Jonathon Mack, PhD - University of San Diego and West Health Institute
Implementation of a Remote Monitoring and Diagnostic System for High Risk Pregnancies
An overview of West Health Institute’s steps in identifying unique opportunities to implement wireless remote monitors in various clinical disease states. Presentation will focus on the development of a prototype wireless fetal monitoring device which allows obstetric specialists to remotely monitor and manage high risk pregnancies via cellular technology. Summary of the development of the device, smart phone application, data management, web portal, and deployment plan will be included.
2/1/2013 Hauke Bartsch, PhD - University of California, San Diego
Why visualization is not enough
Systems for data driven discovery of correlations between imaging, genetics, and neurological and behavioral functions require the effective integration of diverse sources of information. In order to be effective such systems also have to support collaborations of large distributed groups of experts from the fields of informatics, statistics, biology, and psychology. Ideally such a system needs to capture the knowledge of each domain expert, hide technical details and create an abstraction layer that allows each investigator to derive non-trivial correlations in the process of collaborative data driven exploration.
2/8/2013 Neda Alipanah, PhD - University of California, San Diego
Ontological Approach for Phenotype Information Retrieval in Existing GWAS Studies
We present the results of our effort to create such a system as a rich client, HTML5 based web-portal for the Pediatric Imaging, Neurocognition and Genetics (PING) study which collected information from over 1,400 children between the ages of 3 and 20 years to link genetic variations and developing patterns of brain connectivity. The web-portal allows for the simultaneous exploration of roughly 1,000 morphological, behavioral and genetical measures obtained for each study participant. Investigators can define and execute online statistical models used for hypothesis testing. This makes it possible to discover and explore trends in the data with the flexibility to correct for known covariates using a rigorous statistical framework. For a given model the portal supports the exploration of multi-modal image data generated offline. The displayed data includes structural MRI, DTI, atlas tracking and surface reconstructions with vertex based measures for a large number of brain regions. The web-portal therefore combines the elements of study level hypothesis testing with the capabilities of personalized study participant specific exploration of key developmental features.
2/15/2013 Jeff Johnson, MPH - San Diego County, Health and Human Service’s Agency
Biosurveillance and Public Health Informatics within San Diego County
The presentation will highlight several public health functions around biosurveillance, public health informatics, information systems, technology opportunities and challenges. San Diego County Public Health Services is actively working on several local and national initiatives involving information technology. Several of these activities and systems will be discussed.
2/22/2013 Peter J. Park, M.D. - Balboa Naval Hospital
3/1/2013 Amil Gentili, M.D. - University of California, San Diego
3/8/2013 Alison Marsden - University of California, San Diego
Simulation-based treatment planning in pediatric cardiology
Hemodynamics plays an essential role in the progression and treatment of cardiovascular disease. This is particularly true in pediatric cardiology, due to the wide variation in anatomy observed in congenital heart disease patients. While medical imaging provides increasingly detailed anatomical information, clinicians currently have limited knowledge of important fluid mechanical parameters. Treatment decisions are therefore often made using anatomical information alone, despite the known links between fluid mechanics and disease progression. Patient-specific simulations now offer the means to provide this missing information, and, more importantly, to perform in-silico testing of new surgical designs at no risk to the patient. In this talk, we will outline the current state of the art in methods for cardiovascular blood flow simulation and virtual surgery. We will then present new methodology for coupling optimization with simulation and uncertainty quantification to customize treatments for individual patients. We will present two examples in pediatric cardiology that illustrate the potential impact of these tools in the clinical setting: a novel Y-graft for the Fontan surgery, and simulations of flow in coronary artery aneurysms in patients with Kawasaki disease.
3/15/2013 Jim Schultz, M.D. - Neighborhood Healthcare