Engineering & Technology · Biomedical, Bioprocess & Biological Engineering

Living-system models, biological variation and engineering assumptions

Two biological samples receive the same nominal input and produce different responses. Repeating the sensor reading may reduce doubt about the reading while leaving the biological difference intact. Engineering living systems starts by asking what varies, what is measured and which prediction is actually needed.

Use Process Balances & Design, Material Structure & Properties and Instrumentation & Embedded Integration for balances, mechanical models and measurement chains. Biology’s chemistry of life and Chemistry’s enzymes and metabolism provide further preparation. These lessons teach engineering principles, not cultivation protocols, patient-device settings or permission for human research.

Back to the chapter

After reading, you can

Define a useful living-system model and distinguish biological variability, measurement uncertainty and missing model structure.

A model answers a specified question

A living system exchanges matter and energy, changes over time and can respond to its history. Choose a boundary and state variables before writing equations. A cell-population model might track viable cell number; a metabolic model might track chemical amounts; a mechanical model might track tissue deformation. These describe different aspects of the same system and need different evidence.

A parameter is a quantity supplied or estimated within the chosen model, such as a growth coefficient. An assumption restricts how the model represents the world: complete mixing, constant conditions or one uniform population, for example. An omitted mechanism, such as changing nutrient availability, is not repaired merely by reporting the chosen parameter to more decimal places.

Two original growth predictions

Consider an imaginary population with initial count N₀=10 in arbitrary count units and intrinsic coefficient r=0.5 per hour. The exponential model says dN/dt=rN, giving N(t)=10 exp(0.5t). It assumes the same fractional net growth rate throughout the interval; it does not include resource limitation or a changing environment.

For comparison, supply a hypothetical limiting level K=100 and use dN/dt=rN(1−N/K). With the same initial count, its solution is N(t)=100/[1+9 exp(−0.5t)]. The extra factor reduces fractional growth as N approaches K. At four hours the predictions are about 73.9 and 45.1; at eight hours they are about 546 and 85.8. The figure displays the growing disagreement rather than choosing a model by appearance.

N₀ 10 count units and r 0.5 per hour. Exponential N at 4 h is 73.9 and at 8 h 546. Logistic with K 100 gives 45.1 and 85.8. Curves plotted on common time 0–8 h and count 0–550 axes; dashed 100 line is the supplied limiting level. Counts are continuous model values, not literal fractional-cell observations.

Variation is part of the object being studied

Different donors, developmental states, histories or production batches may have different responses. That biological variation differs from uncertainty about an instrument’s calibration or an estimated parameter. Both differ from a structural error shared by all predictions. A model of a mean response does not automatically predict an individual response or the spread around the mean.

Keep biological sample identities and technical repeats linked. Five readings of one sample help describe the measurement process; they do not become five independent biological samples. A reproducible record states the sample source, relevant conditions, method, exclusions and uncertainty, so another reader can see what population and boundary the result represents.

Check the initial rate before extrapolating

At N=10, exponential net growth is 0.5×10=5 count units/hour. Logistic net growth is 0.5×10×(1−10/100)=4.5 count units/hour. Equal r values therefore do not make the initial slopes identical. The exponential model is only an approximation to this logistic model when N/K is sufficiently small for the question.

Neither set of numbers describes an actual organism, experiment or cultivation plan. To select a model, compare predictions with independent observations across the intended range and investigate systematic discrepancies. Even a good fit within one interval does not establish a constant K, a universal growth coefficient or a uniquely identified biological mechanism.

Try the reasoning

Do ten repeated sensor readings establish ten independent biological responses?
Show the answer and reasoning
No. Technical repeats of one sample share its biological identity and do not supply independent biological replication.
Why do the two models differ even at the initial point’s slope?
Show the answer and reasoning
The logistic rate includes the factor 1−N/K, which is 0.9 initially, while the exponential rate does not.
Does fitting the logistic curve prove that nutrients alone limit growth?
Show the answer and reasoning
No. Different mechanisms can produce similar aggregate curves; mechanistic interpretation requires additional evidence.

What to carry forward

State the biological boundary and prediction, preserve variation and replication, and test assumptions before extending a fitted relationship.

References

Teaching resources and research supporting this lesson’s explanations and examples.