Table of Contents

Definition

Modeling and simulation (M&S), referred to in the biomedical literature as computational modeling and simulation (CM&S), is the construction of mathematical and computational representations of biological systems and their numerical execution in order to reproduce, analyze, and predict physiological and pathophysiological behavior over time¹. It is an in silico methodology: rather than physically culturing cells or tissues, it encodes biological structure and cellular mechanics as formal mathematical objects, including systems of ordinary and partial differential equations, stochastic processes, constraint-based networks, and discrete rule-based agents, which are solved numerically to generate dynamic system behavior².

The biological basis of M&S is the mechanistic decomposition of living systems into quantifiable components and interactions. Depending on the scale of interest, a model may encode metabolic reaction kinetics, gene regulatory logic, signal transduction cascades, membrane transport, cell proliferation and death, tissue biomechanics, or whole-organ anatomy, perfusion, and electrophysiology². What the methodology models, therefore, is the emergent behavior of biological networks: growth, homeostasis, disease progression, and response to chemical or physical perturbation. The cellular and structural mechanics are represented explicitly through reaction stoichiometries, rate laws, interaction rules, and physical constraints, so that simulated phenotypes arise from simulated genotypes and environmental inputs².

Whole-cell modeling provides the most complete illustration of this biological basis. The E-CELL software environment demonstrated that a virtual cell, in which genes, transcripts, proteins, and metabolites interact according to defined kinetic rules, can be constructed and simulated on a computer, including substance transport across the membrane, cell growth, and division³. Subsequently, a whole-cell computational model of the human pathogen Mycoplasma genitalium was constructed that includes all of the molecular components and interactions of the cell, accounts for all 525 annotated genes, and simulates the complete life cycle of the organism².

Purpose

The primary value proposition of modeling and simulation is quantitative prediction of system behavior under conditions that are experimentally difficult, expensive, slow, or ethically contentious to test. The paradigmatic expression of this value is the in silico clinical trial, defined as the use of individualized computer simulation in the development or regulatory evaluation of a medicinal product, medical device, or medical intervention⁴. By simulating interventions in populations of virtual patients, M&S can reduce, refine, and partially replace both animal experimentation and clinical testing across the development and assessment cycle⁴.

Modeling and simulation improves upon legacy in vivo models in three principal respects. First, it displaces animal use: physiologically based kinetic (PBK) models constructed from in vitro and in silico data have the demonstrated potential to reduce animal testing by providing a methodology capable of incorporating in vitro human data to facilitate in vitro to in vivo extrapolation of hazard information⁵. Second, it provides human relevance without cross-species extrapolation: because model parameters can be derived directly from human in vitro measurements and human physiological data, predictions concern the human organism rather than an animal proxy, which avoids the well-documented limitations of translating animal findings to humans⁵. Third, it renders assumptions explicit and testable: a computational model is a formal, auditable representation of biological knowledge, in which every mechanistic claim is stated quantitatively and can be interrogated, falsified, and refined by simulation². Computational prediction models such as (quantitative) structure-activity relationship models further extend this replacement value to regulatory toxicology, where they permit endpoint prediction without conducting new experimental tests⁶.

Methodology

The construction of a biomedical simulation proceeds through a defined sequence. The investigator first specifies the question and the context of use, namely the specific role and scope of the model in informing a decision⁷. Data are then assembled from curated literature, omics repositories, and in vitro assays; for example, the Mycoplasma genitalium whole-cell model was parameterized from more than 1,900 experimentally observed parameters curated from over 900 primary sources, reviews, books, and databases².

The formal representation is selected according to the biological question. Continuous deterministic formulations, such as ordinary differential equation systems, describe bulk reaction kinetics and are the standard substrate for kinetic and pharmacodynamic models. Agent-based formulations represent individual cells as discrete agents that sense local information and act according to predefined rules, which makes them well suited to spatially structured tissue phenomena and immune dynamics⁸. Molecular dynamics simulation represents biomolecules at full atomic detail, with empirical force fields describing the potential energy of the system and numerical integration of the equations of motion yielding trajectories of conformational change and ligand binding⁹. Quantitative systems pharmacology combines bottom-up systems biology knowledge with top-down observed clinical data, permitting simulation of drug activity as a perturbation of a biological network¹⁰.

Because M&S is computational rather than physical, its substrates are mathematical structures, software, and data corpora rather than cells, media, or growth factors; human in vitro measurements, such as metabolic clearance determined in human hepatocyte systems, supply the kinetic parameters that physical assays would otherwise provide in vivo⁵. Model credibility is established through verification, which confirms that the equations and code are solved correctly, and validation, which compares model output against independent experimental data, with the rigor of these activities scaled to the risk of the decision being informed, as codified in the risk-informed ASME V&V 40 framework⁷. Reproducibility is maintained through documented software practices, including modular code, version control, automated testing, and structured documentation of data provenance¹¹.

Performance

Empirical performance is best documented for mechanistically complete models. The Mycoplasma genitalium whole-cell model was validated against a broad range of independent data, accurately reproduced measured cellular behaviors, and predicted previously unobserved phenomena, including in vivo rates of protein-DNA association; experimental analysis directed by the model’s predictions subsequently identified previously undetected kinetic parameters and biological functions, demonstrating that comprehensive models can be accurate enough to drive biological discovery².

Reproducibility and credibility are supported by maturing frameworks rather than by biological assay metrics. The ASME V&V 40 standard provides a risk-informed framework in which the credibility of a computational model is assessed as the trust, obtained through collected evidence, in its predictive capability for a stated context of use, with evidentiary requirements commensurate with the consequence of the decision being wrong⁷. Computational reproducibility depends on disciplined software practice, since scientific code is frequently written by investigators without formal software engineering training, and reliability is achieved through modularity, testing, and version control¹¹.

Current technical boundaries remain substantive. Parameter uncertainty is the principal constraint, particularly for models built largely without in vivo data, where the assessment of validity and credibility of PBK models constructed from in vitro and in silico sources alone remains an open methodological challenge⁵. Atomistic molecular dynamics simulations are bounded by computational cost, which restricts the accessible timescales relative to many biological events of interest⁹. Finally, clinical uptake is inconsistent: a survey of 163 clinicians with an 83 percent response rate found low awareness of in silico clinical trials and digital twins among practitioners, with barriers to adoption including restricted access to computing resources and the perception that simulation is slow¹.

Applications

In research, M&S is applied to mechanistic investigation and hypothesis generation across scales, including whole-cell physiology and synthetic biology, where comprehensive models can facilitate biological discovery and computer-aided rational design of microorganisms²; immune and tissue dynamics through agent-based representations⁸; and molecular recognition and drug-target interactions through atomistic simulation⁹. In quantitative systems pharmacology, models are used to make go and no-go decisions, optimize dose and dosing schedules for mono- and combination therapies, and inform development strategy across phases¹⁰.

In industrial and regulatory practice, model-informed drug development (MIDD) is the standardized context of use for pharmaceuticals. The United States Food and Drug Administration established a Model-Informed Drug Development Paired Meeting Program, which advanced from a pilot under PDUFA VI to a continuing pathway for the application of exposure-based, biological, and statistical models in development and regulatory review¹². This context has been harmonized internationally through the ICH M15 guideline on general principles for model-informed drug development, which provides a harmonized framework for planning, model evaluation, and documentation of model-derived evidence in regulatory submissions¹³. In toxicology and chemical safety, PBK models are applied in support of regulatory decision making, including animal-free risk assessment based on in vitro human data⁵, and (Q)SAR models are accepted for the regulatory assessment of chemical endpoints under the OECD validation principles⁶. In medical devices, in silico evidence is submitted under the ASME V&V 40 credibility framework, which regulators regard as the reference standard for risk-based model assessment⁷. In the clinic, digital twins that couple mechanistic models with patient data are being developed to enable precision cardiology, and in silico clinical trials are proposed for refining and partially replacing experimental trials^{14 4}.

Deployment stage and readiness are context dependent. MIDD, PBK modeling, and (Q)SAR prediction are mature and routinely incorporated into regulatory submissions under defined guidance^{12 13 6}. In silico evidence for medical devices is at an advanced stage, supported by an operative international standard⁷. Digital twins and in silico clinical trials remain at an earlier deployment stage, with clinical adoption limited by awareness, computational access, and the absence of a fully established evaluation infrastructure^{14 1}.

Variations

• Physiologically Based Kinetic (PBK/PBPK) Modeling: Mechanistic models of absorption, distribution, metabolism, and excretion that predict internal exposure in humans and are increasingly parameterized solely from in vitro and in silico data to support animal-free risk assessment⁵.

• Quantitative Systems Pharmacology (QSP): Models that integrate disease pathophysiology, systems biology, and pharmacology to quantify drug response in biological networks, supporting dose selection and development decisions across phases¹⁰.

• Agent-Based Modeling (ABM): Discrete, rule-based simulation of individual cells or agents whose local interactions generate emergent tissue and immune behavior, implemented in environments such as NetLogo⁸.

• Molecular Dynamics (MD) Simulation: Atomistic simulation of biomolecular motion using empirical force fields, capturing protein conformational change and ligand binding at fine temporal resolution for drug discovery⁹.

• Whole-Cell Modeling: Comprehensive models that account for all annotated gene functions and molecular components of a cell, pioneered by the E-CELL environment and the Mycoplasma genitalium model^{3 2}.

• In Silico Clinical Trials: The use of individualized computer simulation in the development or regulatory evaluation of medicinal products and medical devices, evaluated on populations of virtual patients⁴.

• (Quantitative) Structure-Activity Relationship ((Q)SAR) Models: Statistical and machine learning models that predict chemical hazard endpoints from molecular structure, accepted for regulatory use under the OECD validation principles⁶.

• Digital Twins: Patient-specific mechanistic models coupled to longitudinal clinical data that project disease trajectories and treatment response, most advanced in precision cardiology¹⁴.