MFBM-MS06

Mathematical and computational methods to augment the reliability of biological models for better decision-making

Tuesday, June 15 at 04:15am (PDT)
Tuesday, June 15 at 12:15pm (BST)
Tuesday, June 15 08:15pm (KST)

SMB2021 SMB2021 Follow Monday (Tuesday) during the "MS06" time block.
Share this

Organizers:

Vincent Lemaire (Genentech, CA, USA, United States), Khamir Mehta (Amgen, Inc, United States), Malidi Ahamadi (Amgen, CA, United States)

Description:

Mechanistic quantitative systems pharmacology (QSP) models are often limited in their utility by the lack of data for complete characterization of the model parameters and the associated parameter uncertainty. Application of such models in drug development will benefit from robust qualification of the model in terms of parameter estimates and also creation of a representative ‘virtual population’ that will likely reflect the observed variability in biological outcomes. Various methodologies have been employed to achieve this goal; including virtual populations-based methods, model reduction to enable direct estimation of parameters, its variability and uncertainty, as well as model averaging to enable spanning across different model functions for individual model components. Recently, data analytics and machine learning methods have also been employed in conjunction with QSP models to expand the utility of complex models, but also to help in the model building process. This session will highlight the key strengths and weaknesses of the above methods as applicable to specific scenarios in the drug development process. Importantly, it will also reflect on the outstanding challenges in the area and will pave the way to further discussions and pollination of ideas and techniques to overcome them.



Chris Rackauckas

(MIT and Pumas AI, MA, USA, United States)
"Accelerating Quantitative Systems Pharmacology with Machine Learning"
Scientific machine learning (SciML) is the burgeoning field combining scientific knowledge with machine learning for data-efficient predictive modeling. We will introduce the Julia SciML ecosystem by describing some of its recent advances, showing how the GPU-accelerated differential equation solvers gave 175x acceleration on Pfizer's internal C-based QSP models and the 15,000x acceleration seen by the NASA Launch Services upon switching from Simulink to ModelingToolkit.jl. After describing the advances in differential equation solvers and automated model discovery, we will describe the JuliaSim simulation ecosystem and its ability to use continuous-time echo state networks (CTESNs) for automatically generating surrogates of highly stiff QSP models. This technique is shown to be validated on a wide variety of models by using CellML and SBML imports to automate the surrogate training process on ~1000 models. Using the Robertson chemical reaction network as an example case, we will see how multi-layer perceptrons (MLPs), recurrent neural networks (RNNs), Long short term memory networks (LSTMs), and physics-informed neural networks (PINNs) all fail to adequately train while only the CTESN succeeds in building a stable surrogate. Examples of accelerating simulations by over 560x over the Dymola Modelica compiler will showcase the scalability of the technique. The will showcase how JuliaSim composes with tools like Pumas to bridge QSP into clinical pharmacology. We will end by describing new adjoint techniques which are required to build neural ODE surrogates on stiff ODE models. Together this showcases the practical changes users of the JuliaSim ecosystem are seeing through scientific simulation


Oleg Demin Jr

(InSysBio, Russia)
"Implementation of variability or uncertainty in parameter values to validate QSP models."
Validation is an important step to test the reliability of the mathematical models including quantitative systems pharmacology (QSP) models. Clinical endpoints for the population of patients are usually used to validate QSP models. For example, percent of responders or mean +/- SD of the particular biomarker. Variability or uncertainty in parameter values should be implemented to describe these endpoints. There are various approaches to extract and implement variability or uncertainty in parameters in model predictions. These methods and cases of their implementation in mechanistic and QSP models will be discussed in the framework of this presentation.


Gianluca Selvaggio

(Fondazione The Microsoft Research - University of Trento Centre for Computational and Systems Biology (COSBI), Italy)
"Parameter free approaches in QSP: modelling the cytokine release following bispecific T-cell engager therapy"
Bispecific T-cell Engaging therapy is a promising treatment that leverages patient’s own immune system to eliminate cancerous cells. To realize the full potential of therapy, it is necessary to mitigate the adverse effects of cytokine release from the immune activation, which eventually lead to adverse effect of cytokine release syndrome (CRS). Computational approaches can be instrumental to explore, systematically, the effects of combined therapies on the tumor killing efficacy and CRS. However, to be fully characterized and validated, quantitative models (such as ODEs) require a priori information, that may be poorly available. An alternative parameter free approach is to use the logical formalism to provide a qualitative representation of the processes. This modelling approach can overcome the data/knowledge gap and the sparsity of clinical data by leveraging on several types of information and integrating both qualitative and quantitative information into computable networks. The presentation will demonstrate a logical QSP model that was used to investigate, through systematic sensitivity analysis, the system behavior and then applied to understand strategies to hamper the inflammatory response without impairing the tumor killing capacity. Our analysis suggests that IFN-γ may be a good mechanism to control CRS risk in patients. Furthermore, it entails the existence of a time window to administrate anti-PDL1 therapy and mitigate inflammation without compromising tumor clearance.


Sietse Braakman

(AbbVie Inc., Quantitative Translational Modeling Group, United States)
"A framework for the evaluation of QSP models, with a focus on verification, validation and uncertainty quantification (VVUQ) methods"
Quantitative systems pharmacology (QSP) and other mechanistic mathematical models are increasingly used to support decisions in drug research and development, as well as regulatory decisions (Nijsen et al., 2018; Zineh, 2019). However, despite their demonstrated value, QSP models are not as widely used as they could be (Leil and Bertz, 2014). Reasons for this include the complexity of these models, a lack of consensus on standards for the evaluation of systems models, and short project timelines that are incompatible with the development of complex models. To work towards a consensus on evaluation standards, we introduce a framework for the evaluation of QSP models (Braakman et al., 2021). The framework is designed to accommodate the wide variety of risk and application settings common for QSP models, by applying certain quantitative and qualitative methods to a model. We include verification, validation, and uncertainty quantification (VVUQ) methods such as global sensitivity analysis, identifiability analysis, confidence and profile likelihood intervals, and model validation with hold-out or external data. Nijsen MJMA, et al., Preclinical QSP Modeling in the Pharmaceutical Industry: An IQ Consortium Survey Examining the Current Landscape. Clinical Pharmacology and Therapeutics: Pharmacometrics and Systems Pharmacology, 2018 7(3): 135-146. https://doi.org/10.1002/psp4.12282 Zineh I, Quantitative Systems Pharmacology: A Regulatory Perspective on Translation. Clinical Pharmacology and Therapeutics: Pharmacometrics and Systems Pharmacology, 2019 8(6): 336-339. https://doi.org/10.1002/psp4.12403 Leil TA and Bertz R, Quantitative Systems Pharmacology can reduce attrition and improve productivity in pharmaceutical research and development. Frontiers in Pharmacology 2014 5:247. https://doi.org/10.3389/fphar.2014.00247 Braakman S, Pathmanathan P, Moore H, Evaluation Framework for Systems Models. Under review 2021.




SMB2021
Hosted by SMB2021 Follow
Virtual conference of the Society for Mathematical Biology, 2021.