Session Description:
- The session will highlight currently funded multi-scale model projects since 2020 to show the breath and depth of currently funded work under the consortium.
- Five podium presentations (10 minutes each wtih Q&A)
- Please post questions for speakers in the comments section below!
Charge to Speakers: What are the present day questions that MSM is helping you to address? what are the scales? what is the next step to improve creation and analysis of multi-scale models?
Speaker Bios & Presentation Materials:
Gary An, "Doing what is hard: The case for complex models to address complex problems"
Biography: Dr. Gary An is a Professor of Surgery and Vice-Chairman for Surgical Research in the Department of Surgery at the University of Vermont Larner College of Medicine. He is a clinically active trauma/critical care surgeon who has worked on the application of complex systems analysis, agent-based modeling and in silico trials to study sepsis, inflammation, wound healing, host-pathogen interactions and cancer since 1999. He is one of the co-founders of Translational Systems Biology, a discipline that promotes the use of multi-scale mechanistic simulation models to cross the Valley of Death of Drug Development. He asserts that the biggest bottleneck in drug development/repurposing is the inability to effectively predict the effect of a molecular manipulation of cellular behavior (e.g. a drug) demonstrated to be effective in pre-clinical studies or with existing clinical usage when it is then applied in a novel clinical context. His work consists of the development of multiscale, cell-based computer simulations and the integration of machine learning and artificial intelligence with such models to represent the individual diversity within clinical populations (e.g. populations of medical digital twins for in silico trials) and for discovery and development of therapeutic control modalities.
Reinhard Laubenbacher, "Modular design of multiscale models, with an application to the innate immune response to fungal respiratory pathogens".
Biography: Dr. Laubenbacher joined the University of Florida in May 2020 as a professor in the Department of Medicine, Division of Pulmonary, Critical Care, and Sleep Medicine. He is the director of the Laboratory for Systems Medicine. Prior to joining UF, he served as director of the Center for Quantitative Medicine and Professor in the Department of Cell Biology in the University of Connecticut School of Medicine. Concurrently, he held a faculty appointment at the Jackson Laboratory for Genomic Medicine. He is a fellow of AAAS, the Society for Mathematical Biology, and the American Mathematical Society. Dr. Laubenbacher is a mathematician by training, and his broad research interests lie in computational and mathematical systems medicine, with a special focus on applications to lung immunology.
Veronika Zarnitsyna, "Multi-scale modeling for waning of vaccine-induced protection"
Biography: Dr. Zarnitsyna is an Assistant Professor in the Department of Microbiology and Immunology at Emory University School of Medicine and a faculty of the Emory Alliance for Vaccine Epidemiology. She is an Associate Member of the American Association of Immunologists. Originally trained as a physicist with a master's degree from the Moscow Institute of Physics and Technology, she developed a deep fascination for the complexity and immense potential of the immune system. Since joining Emory University, she has been doing integrative work, using data from human studies and mouse immunity experiments to model humoral and cellular adaptive immunity to viral infections. One of her recent research areas involves estimating the waning of vaccine-induced protection on both immunological and epidemiological levels.
Aleksander Popel, "Multiscale quantitative systems pharmacology models to conduct virtual clinical trials in cancer immunotherapy"
Biography: Aleksander S. Popel, Ph.D., is a Professor of Biomedical Engineering at the Johns Hopkins University School of Medicine, also a Professor of Oncology and Medicine and a member of the Sidney Kimmel Comprehensive Cancer Center. His areas of expertise are Systems Biology, Quantitative Systems Pharmacology, immuno-oncology, and development of therapeutic peptides. He has published over 350 scientific papers in these areas. Among his honors, he received the Eugene M. Landis Award from the Microcirculatory Society; keynote addresses for The Virtual Physiological Human (VPH) European Union Physiome Project, and Next-Generation Integrated Simulation of Living Matter Project in Japan; C. Forbes Dewey Distinguished Lecturer in Biological Engineering at MIT, A.C. Suhren Lecture at Tulane University, Robert M. and Mary Haythornthwaite Distinguished Lecturer at Temple University, and Kawasaki Medical Society Lecturer in Japan. He is an elected Fellow of AIMBE, AHA, APS, and ASME, and an Inaugural Fellow of BMES. He mentored over 80 postdoctoral fellows and graduate students, many of whom are leaders in the field of bioengineering and systems biology in academia and pharmaceutical industry. He has served in an advisory role to biotech and pharmaceutical companies.
Qing Nie, "Multiscale spatiotemporal reconstruction of single-cell genomics data"
Biography: Dr. Qing Nie is a Chancellor’s Professor of Mathematics and Developmental and Cell Biology at University of California, Irvine. Dr. Nie is the director of the NSF-Simons Center for Multiscale Cell Fate Research jointly funded by NSF and the Simons Foundation – one of the four national centers on mathematics of complex biological systems. In research, he uses multiscale modeling and data-driven methods to study complex biological systems with focuses on single-cell analysis, cellular plasticity, stem cells, embryonic development, and their applications to regeneration, aging, and diseases. Dr. Nie has published more than 200 research articles. In training, Dr. Nie has supervised more than 50 postdoctoral fellows and PhD students, with many of them working in academic institutions. Dr. Nie is a fellow of the American Association for the Advancement of Science, a fellow of American Physical Society, and a fellow of Society for Industrial and Applied Mathematics.
Moderator Bios:
Denise Kirschner (MSM)-My research for the past 25 years has focused on building multi-scale models (MSMs) of the immune response to host-pathogen interactions during infection. My main focus has been to study the immune response to persistent infections (e.g. Helicobacter pylori and Mycobacterium tuberculosis (Mtb) and HIV-1). Such pathogens have evolved strategies to evade or circumvent the host-immune response and my goal is to understand the complex dynamics involved, together with how perturbations to this interaction (via treatment with chemotherapies or vaccines) can lead to prolonged or permanent health. My research focus has pioneered models of the host immune response to HIV-1 and Mtb at multiple spatial and time scales and in multiple physiological compartments including lung, lymph nodes and blood. Our models are continually updated with the latest data from our wetlab collaborators, using primarily non-human primate models of TB. We also have focused on developing tools for analyzing MSMs. We also perform PK/PD drug studies, asking questions about drug efficacy, drug distributions within heterogenous microenvironments, drug resistance, and drug-drug interactions. I have served as president of the Society for Mathematical Biology and also Editor in Chief of the Journal of Theoretical Biology for 20 years.
Reed Shabman is a Program Officer in the Office of Genomics and Advanced Technologies at the NIAID and currently serves as the Program officer for the Systems Biology for Infectious Diseases. The systems biology program consists of a community that integrates experimental biology, computational tools and modeling across temporal and spatial scales to develop strategies that predict and alleviate disease severity across multiple human pathogens. https://www.niaid.nih.gov/research/systems-biology-consortium
Comment
Has MSM supported projects on molecular simulations with physics-based approaches on multiscales, such as molecular dynamics simulations, of biomolecular interactions?
The "valley of death" is actually several steps AFTER the failure. Like Wyle Coyote running off the edge but not falling for several seconds. The flaw is far left on the slide, the chasm is Wyle finally actually falling., the gound fell away long before he actaully fell.
Yes, the failure is predetermined at time of candidate identification. A substantial problem is a general unwillingness to do the "hard" work ahead of molecular entity identification.
To Gary An: One thing I didn't quite get was the relationship in your talk between mechanistic and AI/ML techniques. One of the final slides showed an AI/ML treating a patient but in the talk you mentioned many times the utility of mechanistic models. Where do they fit into your picture.
In this case the AI is trained on a multi scale mechanistic model. The MSM is necessary to because there is no way to capture the counter factuals needed to do DRL to train the AI. One major issue our community deals with is that we are open about the need to utilize methods across the board (ie experimentalists, data scientists, ML/AI, etc); unfortunately many in these areas deny the importance of what we do.
In this case the AI is trained on a multi scale mechanistic model. The MSM is necessary to because there is no way to capture the counter factuals needed to do DRL to train the AI. One major issue our community deals with is that we are open about the need to utilize methods across the board (ie experimentalists, data scientists, ML/AI, etc); unfortunately many in these areas deny the importance of what we do.
regarding adaptability how much real world data would you expect needed to have a starting model? How often would you think "updating" would need to be built in?
Anything that might have clinical use would require a substantial amount of offline testing to see if it were trustworthy. My own opinion is that with the MRM paradigm and nonfalsifiability this shifts what you want of your data: rather than looking for statistical significance, you would rather want a great of "coverage" of what is biologically possible. In this case, multiple data sources (eg hospitals) would be beneficial, particularly if this was used as synthetic data to train AIs: this approach would mitigate the intrinsic limitation of AI systems in terms of generalizability and data drift. This also is a means of overcoming inequities in health data from underrepresented populations.
Anything that might have clinical use would require a substantial amount of offline testing to see if it were trustworthy. My own opinion is that with the MRM paradigm and nonfalsifiability this shifts what you want of your data: rather than looking for statistical significance, you would rather want a great of "coverage" of what is biologically possible. In this case, multiple data sources (eg hospitals) would be beneficial, particularly if this was used as synthetic data to train AIs: this approach would mitigate the intrinsic limitation of AI systems in terms of generalizability and data drift. This also is a means of overcoming inequities in health data from underrepresented populations.
This is a tough question to answer because the starting model has to begin with a theoretical foundation of how the system works; this is typically mined from the accepted literature which can be based on decades of experimental and clinical.
In terms of ‘updating,’ I think that depends on system dynamics. To instantiate your computational model into a personalized digital twin with a reasonable ability to forecast, you need multiple variables measured longitudinally to establish parameter ranges that are valid in your model for the individual. Updating then occurs at fixed time horizons, based on forecasted probability cones, or in circumstances in which the in vivo system evolves outside the computationally predicted probability cones.
For Dr. Popel: do you account (or have any insight) into whether the virtual patient cohort accounts for multiple socioeconomic backgrounds? Is there any link between immuno-onocological data and environment (e.g., chronic stress)? If so, can you account for this difference in the modeling framework?
once question is how close are we to using your results to influecne clinical trials in your area?
To any of the presenters who consider parameter inference / inverse problems, could you please comment on how you address this in the setting of multiscale modeling? How should be handle calibrating our models to mutlimodal and multiscale data from multiple and unique experiments?