Integrating machine learning and multiscale modeling—perspectives, challenges, and opportunities in the biological, biomedical, and behavioral sciences

Abstract

Fueled by breakthrough technology developments, the biological, biomedical, and behavioral sciences are now collecting more data than ever before. There is a critical need for time- and cost-efficient strategies to analyze and interpret these data to advance human health. The recent rise of machine learning as a powerful technique to integrate multimodality, multifidelity data, and reveal correlations between intertwined phenomena presents a special opportunity in this regard. However, machine learning alone ignores the fundamental laws of physics and can result in ill-posed problems or non-physical solutions. Multiscale modeling is a successful strategy to integrate multiscale, multiphysics data and uncover mechanisms that explain the emergence of function. However, multiscale modeling alone often fails to efficiently combine large datasets from different sources and different levels of resolution. Here we demonstrate that machine learning and multiscale modeling can naturally complement each other to create robust predictive models that integrate the underlying physics to manage ill-posed problems and explore massive design spaces. We review the current literature, highlight applications and opportunities, address open questions, and discuss potential challenges and limitations in four overarching topical areas: ordinary differential equations, partial differential equations, data-driven approaches, and theory-driven approaches. Towards these goals, we leverage expertise in applied mathematics, computer science, computational biology, biophysics, biomechanics, engineering mechanics, experimentation, and medicine. Our multidisciplinary perspective suggests that integrating machine learning and multiscale modeling can provide new insights into disease mechanisms, help identify new targets and treatment strategies, and inform decision making for the benefit of human health.

The reproduction number of COVID-19 and its correlation with public health interventions

“…Public health interventions. The classical SEIR model (1) assumes that the disease develops freely and that the contact rate β , latency rate α, and infectious rate γ are constant throughout the course of the outbreak. It is obvious that the contact rate β will change in response to community mitigation strategies and political actions, e.g., local lockdown and global travel restrictions [15].…”

“…We introduce a dynamic SEIR model with a time-varying contact rate β (t) that transitions smoothly from the initial contact rate β 0 at the beginning of the outbreak to the current contact rate β t under global travel restrictions and local lockdown. We express the time-varying contact rate β (t) = R(t)/C as a function of the effective reproduction number R(t) and use machine learning [1] to learn the evolution of the reproduction number for each country of the European Union from its individual outbreak history [12]. Our model allows us to precisely quantify the initial basic reproduction number, the reduced current effective reproduction number, and the time to achieve this reduction, which are important quantitative metrics of the effectiveness of national public health intervention.…”

Outbreak dynamics of COVID-19 in Europe and the effect of travel restrictions

“…However-just like any infectious disease model-our model will naturally face limitations associated with data uncertainties from differences in testing, inconsistent diagnostics, incomplete counting, and delayed reporting across all countries. Additional limitations arise from model uncertainties including ill-defined initial conditions, mapping reported populations into model compartments, and choosing the model itself (Alber et al 2019). However, as more data become available, we are confident that we will learn from uncertainty quantification, become more confident in our model predictions, and learn how to quickly extract important trends.…”

Multiscale Modeling Meets Machine Learning: What Can We Learn?

“…This article seeks to answer the question of how multiscale models can benefit from machine learning [3]. It is the introduction to a special issue on 'Uncertainty Quantification, Machine Learning, and Data-Driven Modeling of Biological Systems' that seeks to place this theme within the broader field of computational mechanics.…”

“…Finally, natural questions that machine learning can help us answer focus on sensitivity analysis [62] and uncertainty quantification [61,146]. We have structured this introduction to the special issue around four methodological areas, ordinary and partial differential equations, and data and theory driven machine learning [3]. For each area, we discuss the state of the art, identify applications and opportunities, raise open questions, and address potential challenges and limitations in view of specific examples from the life sciences.…”