Model-based Physiomarkers (Vasilis Marmarelis lead)
- The driving idea is to employ data-based models that quantify the dynamics of physiological systems of interest in order to compute reliable "physiomarkers" with clinical diagnostic utility (i.e. model-based indices measuring specific physiological attributes of each given patient that assist differential diagnosis). These physiomarkers seek to capture "macroscopic" characteristics of the relevant patient physiology in contrast to "microscopic" (molecular or cellular) characteristics relevant to "biomarkers".
A physiomarker can: (1) improve clinical diagnosis by making it more quantitative, data/evidence-based, reliable and patient-specific; (2) advance medical and physiological understanding of the underlying pathology and disease process, enabling novel treatments; (3) rationalize and streamline clinical procedures for patient care in a manner that engenders disease management improvements; (4) achieve and maintain a higher, uniform standard of patient care while reducing cost and patient discomfort.
As our ability to collect time-series data in a clinical setting expands, our modeling capabilities/methodologies become potentially applicable and enable the computation of model-based physiomarkers with clinical utility. The purpose of this theme is to discuss how the MSM Consortium can help this idea move forward and guide its proper implementation.
The tentative outline of the one-hour Theme presentations session (September 3rd, 11 am to noon) is given below.
Intro (4 min) Motivation and potential clinical utility of model-based Physiomarkers. Speaker 1 (14 min): Case study from a cardiovascular disease (Andrew McCulloch). Speaker 2 (14 min): Case study from a immune-response disease (Gary An). Speaker 3 (14 min) Case study from a host-pathogen infection system (Denise Kirschner). Speaker 4 (14 min): Case study from a neurodegenerative disease (Vasilis Marmarelis).
There will be a Breakout session for group discussion on the next day, September 4th, from 11:30 to 12:30 in Room H of Building 45. A panel of Discussants will stimulate the group discussion with comments on the Theme and the presented case studies (5 min each). Among them (so far) are: Jim Bassingthwaighte (overall Theme) , Dan Beard (cardiovascular disease) and Bill Lytton(neurodegenerative disease). You may indicate to the Theme-Lead (V. Marmarelis, firstname.lastname@example.org) your interest to be on the panel.
Below are brief bios of the speakers and Abstracts of their talks:
Case Study 1: Deriving Novel Physiomarkers of Therapeutic Outcomes from Patient-Specific Models of the Failing Heart. Andrew D. McCulloch PhD, Adarsh Krishnamurthy, Christopher Villongco, David Krummen, Roy C.P. Kerckhoffs. University of California, San Diego
Dyssynchronous heart failure (DHF) is congestive heart failure complicated by an electrical conduction disorder that prolongs the QRS complex such as left bundle branch block (LBBB). Heart failure has a 5-year mortality of 50%, and the prognosis for DHF is worse. However, DHF patients are frequently indicated for cardiac resynchronization therapy (CRT), which uses biventricular pacing to restore ventricular electromechanical synchrony with the goal of improving ventricular function and clinical outcomes. Unfortunately, over 30% of CRT candidates are clinical non-responders and efforts at identifying clinical biomarkers of CRT outcomes have been disappointing.
In ten patient volunteers undergoing CRT at the San Diego VA Medical Center who gave informed consent, we created patient-specific multi-scale models using data from CT imaging, echocardiography, cardiac catheterization, electrocardiography and endocardial electroanatomic mapping. We used these models to test hypotheses about DHF etiology and CRT responses. In particular, we had previously shown with multi-scale models that the severity of DHF is significantly dependent on the interaction between electrical dyssynchrony and ventricular dilatation, and that the coefficient of variation of regional myocardial work (COVRW) computed with the model is a good physiomarker of this interaction. Here we found a significant correlation between CRT outcomes and COVRW and we use the model result to explore the underlying physiological basis of this finding.
Andrew McCulloch, Ph.D. Engineering Science (1986) is Distinguished Professor of Bioengineering and Medicine and Jacobs School Distinguished Scholar at the University of California San Diego. He is Director of the UCSD Cardiac Biomedical Science and Engineering Center, a Senior Fellow of the San Diego Supercomputer Center, and a Principal Investigator of the National Biomedical Computation Resource. He served as Chair of the Bioengineering Department at UCSD from 2005 to 2008 and currently directs the HHMI-NIBIB Interfaces Graduate Training Program in Multi-Scale Biology. Dr. McCulloch is a Fellow of AIMBE and the American Physiological Society, Associate Editor of PLoS Computational Biology and Editor-in-Chief of Drug Discovery Today: Disease Models. He is also chair of the Physiome and Systems Biology Committee of the International Union of Physiological Sciences. He has published over 170 articles on experimental and computational modeling of the normal and diseased heart from basic molecular mechanisms to clinical applications.
Case Study 2: Translational Systems Biology of Systemic Inflammation: Addressing the Translational Dilemma by generating clinically relevant phenotypes with multi-scale dynamic modeling Gary An, MD, Department of Surgery, University of Chicago
The Translational Dilemma, the discrepancy between the gains in basic biological knowledge and the ability to effectively translate that knowledge into clinically effective therapeutics, is the greatest challenge facing the biomedical research community. While advances in correlative methods show promise in the diagnosis, prognosis and patient substratification of various diseases, the fact is that the pipeline to refill the therapeutic toolbox has run dry. Translational Systems Biology is the application of multi-scale dynamic computational models with an expressed goal of affecting clinical practice, and was developed with a focus on systemic inflammation as a prototypical multi-scale systemic disorder. Presented herein will be a survey of over a decade of the use of dynamic multi-scale modeling of inflammation to generate basic insights that have anticipated clinically relevant phenotypic characterization of sepsis, trauma and multi-organ failure. The benefits of qualitative dynamic knowledge representation will be placed into the context of a general strategy aimed at augmenting the throughput of mechanistic hypothesis evaluation necessary to address the Translational Dilemma.
Dr. Gary An is an Associate Professor of Surgery and the co-Director of the Surgical Intensive Care Unit at the University of Chicago. In addition to being an active clinician he is a Senior Fellow of the Computation Institute at the University of Chicago. He is a graduate of the University of Miami, Florida School of Medicine, and did his surgical residency at Cook County Hospital/University of Illinois, Chicago. He was previously a Trauma Surgeon at Cook County Hospital from 1997 to 2003, the Director of the Burn Intensive Care Unit at Cook County Hospital from 2003-2006 and a Trauma Surgeon at Northwestern Memorial Hospital in Chicago, IL 2006 to 2010. He is a founding member of the Society of Complexity in Acute Illness (SCAI) and is the current president of the Swarm Development Group, one of the original organizations promoting the use of agent-based modeling for scientific investigation. He is the founder and director of the Fellowship in Translational Systems Biology at the University of Chicago. He is member of multiple medical and computer science societies, and serves on the editorial board of several journals.
Case Study 3: A two-pronged data-mining approach to physiomarker discovery in tuberculosis. Denise Kirschner PhD, Department of Microbiology and Immunology, and co-Director of Center for Systems Biology, University of Michigan
Identifying and assessing physiomarkers are ongoing challenges in developing immunological correlates for new treatments and vaccines. To effectively predict physiomarkers for infection progression, large amounts of data are required to reach statistical significance and make accurate predictions. Additional tools would be useful for use in this pursuit, and we propose a method using computational modeling as a viable approach. Since reproducible and viable data from humans are mostly collected in the blood, we must include the following compartments: sites of infection and blood. We apply both classical data mining techniques as well as new classification algorithms on in vivo and in vitro data. A major benefit of using machine-learning techniques on in silico data is that the data is exhaustive in both time and density. We apply our technique to study the immune response to M. tuberculosis infection and are able to predict physiomarkers using this method that have otherwise gone unidentified.
Dr. Denise Kirschner, has focused her research on the host immune response to M. tuberculosis at multiple spatial and time scales and in multikple physiological compartments including lung, lymph nodes and blood. To date we have worked and collaborated with experiemntalists generating data on TB with in vitro, mouse, non-human primate and human studies. We have spent a considerable effort to study unique structures, granulomas, that are involved in the immune response to M. tuberculosis. Dr. Kirschner currently serves (and has for the past decade) as co- Editor-in-Chief of the Journal of Theoretical Biology the oldest and the top theoretical biology journal. She has served as both a member of and as chair for many study section review panels at the National Institutes of Health. This gives her solid knowledge and experience of the field and broad expertise in many areas of computational and mathematical biology. She serves as the founding co-director of The Center for Systems Biology at the University of Michigan, a new interdisciplinary center at the University of Michigan aimed to facilitate research and training between wetlab and theoretical scientists. She has published over 100 peer-reviewed papers in the area of host-pathogen systems.
Case Study 4: Model-based physiomarkers of vasomotor reactivity for diagnosis of early-stage Alzheimer's disease. Vasilis Marmarelis PhD, Department of Biomedical Engineering & Biomedical Simulations Resource Center, University of Southern California
Data-based dynamic nonlinear models of cerebral hemodynamics can be obtained in a practical context using 5-6 min of beat-to-beat time-series measurements of cerebral blood flow velocity, arterial blood pressure and end-tidal CO2. This is achievable by using the novel modeling concept of Principal Dynamic Modes (PDMs). These PDM-based predictive models are extracted from the data in an assumption-free context and describe the dynamics of cerebral flow autoregulation and CO2 vasomotor reactivity in each subject. They can be used to compute indices quantifying these physiological attributes in a subject-specific manner (physiomarkers). It is posited that such model-based physiomarkers may assist clinical diagnosis. Case in point is our recent finding that early-stage Alzheimer's patients exhibit impaired vasomotor reactivity, as quantified by such a model-based index/physiomarker. Although confirmed so far in a small sample of Alzheimer's patients, this finding is corroborated by previous qualitative observations of impaired vasomotor reactivity in neurodegenerative diseases with a vascular component and charts a novel pathway for exploring the physiological mechanisms that underpin such diseases. In addition to their potential clinical utility for diagnosis and treatment monitoring when confirmed with a large sample of patients, such model-based physiomarkers may advance our understanding of the underlying physiological mechanisms and allow the formulation of testable hypotheses that may elucidate our scientific understanding of the process of pathogenesis in such diseases.
Vasilis Marmarelis, PhD in Engineering Science, Caltech (1976), Professor of Biomedical Engineering at USC and co-Director of the Biomedical Simulations Resource, a research center funded by NIH since 1985. He served as Department Chairman from 1990 to 1996. His key research interests are: (1) dynamic nonlinear modeling of biomedical systems; (2) neural information processing; (3) modeling of physiological autoregulation; (4) multimodal ultrasound tomography (MUT) for diagnostic imaging; (5) model-based diagnostic physiomarkers. The key application domains of interest are: cognitive systems and neurostimulation, cerebral hemodynamics and neurodegenerative disease, endocrine-metabolic regulation and diabetes, and non-invasive lesion differentiation via MUT diagnostic imaging. Co-author of the seminal book: "Analysis of Physiological System: The White Noise Approach"(1978; Russian, 1981; Chinese, 1990) and author of the 2004 monograph: "Nonlinear Dynamic Modeling of Physiological Systems". He has published more than 150 journal papers and book chapters. In 2000, he invented the MUT diagnostic imaging system for early detection of breast cancer, which is currently clinically evaluated in Europe. He is a Fellow of the IEEE and the AIMBE.
Some Thoughts and Questions in Brain Physiomarkers. German Cavelier - NIMH
I am interested in getting suggestions about models of intracellular signaling pathways in neurons as possible physiomarkers. Some initial knowledge is being generated about contributions of some neurotransmitters (for example dopamine), some ion channels (the NMDA receptor), and some ions (intracellular Ca++) to mental health diseases. Would a model be able to provide additional insight about these signaling pathways, and to direct to some Physiomarker for mental health diseases?
Some Thoughts and Questions in Liver Physiomarkers. Jim Sluka - Indiana University
Liver toxicity is a significant human (and in some cases non-human) problem for both pharmaceuticals and environmental chemicals. Clinically, a standard panel of liver function markers in the blood are often the only clinical data available upon which clinicians must base their treatment plan. Computational models of liver toxicity will perhaps be most useful to clinicians if the models can predict the time course and level of these standard liver makers. This may require development of PBPK type models for the distribution and clearance of these (typically) proteinaceous blood markers and coupling those models to cell-scale models of liver cell damage and death. In addition, there may be an opportunity to suggests more specific and/or more sensitive makers of liver toxicity based on the increasing amount of data on toxin induced changes, including zone dependent changes, in proteins expressed in the liver.