Biology models are simplified representations of living systems, built so researchers can study processes that would otherwise be too complex, too slow, too small, or too ethically fraught to observe directly. They range from a dish of cultured cells on a lab bench to a supercomputer running millions of equations that simulate how a disease spreads through a population. Every model deliberately leaves things out, keeping only the features relevant to a particular question, and that selective simplification is exactly what makes models useful. The variety is enormous, and the choice of which model to use shapes every conclusion a biologist draws.
Physical and Anatomical Models
The oldest biology models are physical objects you can hold in your hands. For centuries, anatomists built replicas of the human body and its organs from whatever materials were available. These models have been constructed from metal, wood, ivory, wax, papier-mâché, plaster, and plastic, and have more recently evolved into computerized and digital representations.1PubMed. The Evolution of 3D Anatomical Models: A Brief Historical Overview A wax heart sitting on a teaching table in 1750 and a 3D-printed liver replica sitting in a surgical planning room today are doing the same fundamental job: letting someone study a structure without needing the actual organ in front of them. Physical models remain a standard part of medical education and preoperative planning, though their role has expanded dramatically with 3D printing and virtual reality.
Model Organisms
When biologists talk about “models” in everyday lab conversation, they often mean model organisms: living species chosen because they are easy to breed, genetically well-understood, and share enough biology with humans (or with the system of interest) to be informative. Fruit flies, zebrafish, roundworms, and mice are all workhorses of biology for these reasons. In fields like metabolomics, model organisms provide excellent samples for developing new methods and for building comparative studies of conserved and unique metabolic pathways across species.2PubMed Central. Considerations when choosing a genetic model organism for metabolomics studies A researcher studying a gene involved in heart development, for instance, might first investigate its equivalent in zebrafish embryos, whose transparent bodies make it possible to watch the heart form in real time under a microscope.
Model organisms are powerful, but they come with a well-known limitation: they are not human. As research has pushed into subtler biological dimensions, the differences among species have often outweighed the similarities, and these differences are thought to be one of the main reasons many human clinical trials fail.3PubMed Central. The Flaws and Human Harms of Animal Experimentation A drug that cures cancer in a mouse does not necessarily do anything useful in a person. Even when researchers identify a genuinely important biological mechanism, animal models often cannot capture the full range of variation seen in human disease, making them poor predictors of whether a treatment will actually work in patients.4PubMed Central. Why animal model studies are lost in translation This gap between animal results and human outcomes has driven a major push toward alternative model types.
Cell Cultures and In Vitro Models
Growing cells in a dish gives researchers a controlled, simplified system for studying cellular behavior, testing drug responses, and dissecting molecular pathways. Traditional two-dimensional (2D) cell cultures, where cells grow in a flat layer on the bottom of a plastic plate, have been a backbone of biomedical research for decades. They are cheap, fast, and reproducible. But 2D cultures have real drawbacks: growing cells flat disrupts the interactions between cells and their surrounding environment, changes cell shape and polarity, and alters the way cells divide. These limitations led to the development of three-dimensional (3D) culture systems that more closely mimic conditions inside the body.5PubMed Central. 2D and 3D cell cultures – a comparison of different types of cancer cell cultures
In a 3D culture, cells can form clusters, interact with a scaffold or gel that imitates the tissue around them, and behave in ways that are more biologically realistic. Cancer researchers, for example, have found that tumor cells growing in a 3D environment respond differently to drugs than the same cells spread on a flat dish. This makes 3D models a better starting point for predicting whether a treatment might work in a patient.
Organoids and Organs-on-Chips
Organoids take 3D culture a step further. These are tiny, self-organizing structures grown from stem cells or patient tissue that develop some of the architecture and function of real organs. A gut organoid, for instance, forms a hollow interior lined with the same types of cells found in the intestinal wall. In cancer research, patient-derived organoids are especially promising: because they retain the genetic, epigenetic, and phenotypic features of the original tumor, researchers hope they can predict how an individual patient will respond to a specific drug, enabling personalized therapy on a clinically useful timeline.6PubMed Central. Patient-Derived Organoids as a Model for Cancer Drug Discovery Organoids derived from liver and bile duct cancers, for example, are being used for high-throughput drug screening to identify effective treatments faster than traditional methods allow.7PubMed. Patient-derived functional organoids as a personalized approach for drug screening against hepatobiliary cancers
Organs-on-chips push the concept further still. These are microfluidic devices, usually the size of a USB stick, that contain tiny channels lined with living human cells. Fluid flows through the channels to mimic blood circulation, and the device can be designed to recreate the physical environment of a specific organ, including the mechanical stretching of lung tissue during breathing or the shear forces blood vessels experience. By recreating multicellular architectures, tissue interfaces, and vascular flow, these chips produce levels of tissue function that are not possible with conventional 2D or 3D culture systems.8PubMed Central. Microfluidic organs-on-chips The dynamic culture environment these chips provide is a critical advantage over static models, and researchers are working to eliminate remaining technical hurdles like the need for external pumps and tubing.9PubMed. Unlocking the Potential of Organ-on-Chip Models through Pumpless and Tubeless Microfluidics
Mathematical and Computational Models
Not every biology model involves living cells or physical objects. Mathematical models use equations to describe biological processes. These “in silico” models run on computers and can simulate things that would take years or be impossible to observe directly. At the molecular level, systems of coupled equations are the natural language for describing how enzymes catalyze reactions, how signals propagate inside a cell, or how gene networks respond to a stimulus.10PubMed Central. Classic and contemporary approaches to modeling biochemical reactions The goal is not to reproduce every molecule in a cell but to capture the key relationships well enough to make predictions that experiments can then test.
Agent-based models take a different approach. Instead of writing equations for the whole system, researchers define rules for individual agents, typically individual cells, and then simulate what happens when thousands of those agents interact. Each cell “decides” what to do based on local conditions: whether to move, divide, signal its neighbors, or die. The collective behavior that emerges from these simple rules can reproduce complex real-world patterns. One study used agent-based modeling to understand how bacterial cells aggregate into multicellular mounds during starvation, testing different mechanisms for biased cell movement and identifying a chemical-signaling model that reproduced the patterns seen in laboratory experiments.11PubMed Central. Agent-Based Modeling Reveals Possible Mechanisms for Observed Aggregation Cell Behaviors Agent-based modeling was originally widespread in ecology and the social sciences before gaining traction in biomedical research, where it is now used to study tissue development, wound healing, and tumor growth.12Briefings in Bioinformatics. Combining experiments with multi-cell agent-based modeling to study biological tissue patterning
Epidemiological and Ecological Models
During the COVID-19 pandemic, many people encountered biology models for the first time through news coverage of epidemic forecasting. Compartmental models divide a population into groups, with the classic version sorting everyone into susceptible, infected, and removed categories. People flow between compartments according to rates of transmission and recovery. Variants of this framework add extra categories, such as an “exposed” group for people who have been infected but are not yet contagious. These models informed public health decisions about lockdowns, hospital capacity, and vaccination rollouts worldwide.13PubMed Central. A model for the spread of infectious diseases compatible with case data
Ecological models address different questions but use a similar philosophy: simplify a complex system into tractable pieces and study how they interact. Species distribution models, for example, map where a species currently lives, characterize the environmental conditions it needs, and then project where suitable habitat will exist under future climate scenarios. These models are now critical tools for conservation planning and forest management. Recent improvements in climate data quality, modeling algorithms, and the incorporation of genomic information have sharpened their predictions.14PubMed Central. Advancements in ecological niche models for forest adaptation to climate change: a comprehensive review One study using two different climate models and two distribution algorithms to project the future ranges of 60 California landbird species found that most were projected to shrink by 2070.15PubMed Central. Niches, models, and climate change: assessing the assumptions and uncertainties Those projections feed directly into decisions about which habitats to protect and where to invest restoration resources.
Predicting future distributions generally involves two steps: first, build a model that characterizes where a species can live based on current conditions, and second, link that model to a simulation of how the species will actually move and establish in new areas over time.16PubMed. A framework for using niche models to estimate impacts of climate change on species distributions Getting the second step right is hard, because it depends on how fast organisms can disperse and whether they encounter barriers like cities, mountains, or oceans.
Machine Learning and Protein Structure Prediction
One of the most dramatic recent successes in biological modeling comes from machine learning. Predicting the three-dimensional shape a protein will fold into, based only on the sequence of its building blocks, was an unsolved problem for over 50 years. AlphaFold, a neural-network-based system, changed that by predicting protein structures with accuracy competitive with experimental methods in a majority of cases, vastly outperforming everything that came before.17PubMed Central. Highly accurate protein structure prediction with AlphaFold The practical implications have been enormous. Researchers studying drug targets, enzyme function, or the mechanisms of genetic disease can now get a reliable structural model in hours rather than spending months or years on laboratory crystallography. AlphaFold’s database now includes predicted structures for hundreds of millions of proteins, essentially covering most known protein sequences on Earth.
Evolutionary Models
Biology does not just study how living systems work right now; it also asks how they got here. Evolutionary models use the branching tree of species relationships to study how traits change over time. Phylogenetic comparative methods account for the fact that closely related species share a common ancestor and therefore are not truly independent data points. Researchers can translate hypotheses about adaptation in different environments into explicit mathematical models, test those models against observed trait data, and infer details about the evolutionary process that produced the diversity we see today.18PubMed. Phylogenetic Comparative Analysis: A Modeling Approach for Adaptive Evolution Newer approaches incorporate information about gene-tree discordance, the fact that different genes within a genome can have different evolutionary histories, to produce more accurate inferences about ancestral traits and shifts in evolutionary rate along specific lineages.19PubMed Central. Phylogenomic comparative methods: Accurate evolutionary inferences in the presence of gene tree discordance
Multiscale Models and the Integration Challenge
One of the defining challenges in modern biology is connecting processes that happen at very different scales. A mutation in a single gene changes a protein, which changes how a cell behaves, which changes a tissue, which changes an organ, which changes a whole organism’s health. Multiscale computational models try to bridge these gaps by linking simulations at the molecular level to simulations at the tissue or organ level. Powerful computing platforms and high-throughput experimental data have allowed these models to expand as a way to investigate biological phenomena across scales in experimentally relevant ways.20PubMed Central. Multiscale computational models of complex biological systems Building such models is deeply data-intensive, relying on structural biology, microscopy, and medical imaging to construct high-resolution data sets at each level, while also relying on physics to constrain how molecular and cellular processes scale up to produce organ-level physiology.21Biophysical Journal. Systems Biophysics: Multiscale Biophysical Modeling of Organ Systems
Why All Models Are Wrong, and Why That Is Fine
A phrase that circulates widely in the modeling community captures the essential tension: all models are wrong, but some are useful. Because every model is a deliberate simplification, no model perfectly reproduces reality. The value lies not in being right about everything but in being right about the specific question you are asking. The best models start with a well-defined biological question, include only the level of detail needed to address it, and aim not just to explain existing data but to make predictions that experiments can test.22PubMed Central. “Essentially, all models are wrong, but some are useful”-a cross-disciplinary agenda for building useful models in cell biology and biophysics A model that predicts something surprising, and turns out to be right, teaches you far more than one that merely re-describes data you already had.
Validation, the process of checking a model’s predictions against independent data, is where many modeling efforts succeed or fail. A model of disease spread that accurately predicted hospitalizations during one wave may collapse during the next if a new variant changes the biology. A cell culture model that captures drug sensitivity for one tumor type may be useless for another. Researchers need to be transparent about what their models were trained on, where they have been tested, and where the edges of their reliability lie.
Ethics and the 3Rs
The push toward non-animal models is not only scientific but ethical. The 3Rs framework, which stands for Replacement, Reduction, and Refinement, has shaped research ethics for decades and is now incorporated into legislation worldwide.23PubMed Central. The 3Rs and Humane Experimental Technique: Implementing Change Replacement means using a non-animal alternative whenever possible. Reduction means minimizing the number of animals used. Refinement means minimizing suffering when animal use is unavoidable. Advances in organoids, organ-on-chip devices, and computational modeling have given researchers more options for replacing animal experiments than ever before, and this shift has changed the perspective of regulators, who increasingly accept data from more human-relevant in vitro approaches for safety testing.24PubMed. The new paradigm in animal testing – “3Rs alternatives”
Synthetic biology has added yet another dimension. Researchers are working to build minimal cells, stripping a living cell down to only the genes and components it absolutely needs to survive and reproduce. This proceeds in two directions: “top-down,” by deleting genes from existing bacteria until only the essentials remain, and “bottom-up,” by assembling DNA, RNA, proteins, and membranes from scratch in the lab.25PubMed Central. Update on designing and building minimal cells These minimal cells serve as models for understanding what life fundamentally requires, and they may eventually serve as programmable chassis for applications in medicine and manufacturing.
Sharing Models and Making Them Reproducible
A model that only its creator can run is not very useful to the broader scientific community. Reproducibility has been a persistent headache in computational biology, and the field has responded by developing standardized languages for encoding models and simulations. The Systems Biology Markup Language (SBML) lets different software tools operate on an identical representation of a model, removing opportunities for translation errors and ensuring a common starting point for analysis.26PubMed Central. The Systems Biology Markup Language (SBML): Language Specification for Level 3 Version 1 Core A companion standard, SED-ML, describes the simulation experiment itself: what settings were used, how long the simulation ran, and what outputs were recorded. With SED-ML, other scientists can reproduce published results in different software tools, and experiments covering models from different fields can be accurately described and combined.27PubMed Central. Reproducible computational biology experiments with SED-ML–the Simulation Experiment Description Markup Language
Public repositories where researchers deposit their models, along with the data used to build and validate them, have become an expected part of the publication process in many subfields. This infrastructure matters because science is cumulative. A model of gene regulation built in one lab might be exactly the starting point another lab needs for a study of drug resistance. Without a shared format and a shared place to store models, that kind of building-on-prior-work stalls out, and researchers end up reinventing the wheel instead of extending it.