Plant systems biology studies how interacting genes, molecules, cells, organs and environmental conditions produce a plant’s behaviour. Its gold standard is a model that makes the connections explicit, predicts a measurable response and survives a fair test when the conditions change. A colourful network or a large collection of measurements becomes useful when it helps explain a specific outcome: where a root bends, when a leaf uses reserves, or why a response differs between tissues.
This guide teaches that reasoning through original, simplified examples. You will connect molecular measurements with transport and growth, distinguish a clock shift from a change in abundance, balance competing resource demands, and decide which experiment can separate two plausible explanations. The numerical cases are teaching constructions, not measurements from the research papers linked below or recommendations for managing crops.
For the general language of models and networks, begin with The Gold Standard of Systems Biology. Here the focus is the extra work required by plants: spatially organised tissues, changing geometry, light–dark cycles, transport between organs and responses that depend on developmental context. The learning aim follows The Gold Standard of Education: understanding that remains usable when a familiar question changes.
THE GOLD STANDARD · PLANT SYSTEMS BIOLOGY
Connect plant mechanisms across space and time
Choose a route through transport, growth, clocks, resource allocation and testable plant models.
ROUTE 1 · CHAPTERS 1–3
Define the plant system
Connect a question, boundary and measurement layers.
ROUTE 2 · CHAPTERS 4–6
Follow networks into growth
Trace evidence, transport and changing geometry.
ROUTE 3 · CHAPTERS 7–10
Connect time and resources
Test rhythms, sampling and day–night constraints.
ROUTE 4 · CHAPTERS 11–14
Challenge the model
Examine feedback, parameters, predictions and scale.
ROUTE 5 · CHAPTERS 15–17
Transfer and continue
Solve the integrated case and choose the next learning step.
Full chapter index · Try the transfer task · Gold Standard subject hub
Chapter index
Define the plant system · Chapters 1–3
Follow networks into growth · Chapters 4–6
Connect time and resources · Chapters 7–10
Challenge the model · Chapters 11–14
CHAPTER 1 OF 17 · Define the plant system
1. Choose a behaviour that needs an explanation
“Understand the plant” is too broad to determine what should be measured. “Explain why two seedlings reach the same final angle through different growth trajectories” is a tractable beginning. It names a response, a comparison and a reason to collect observations through time. The model can then represent the processes needed for that question without pretending to contain everything about the organism.
A useful question has four parts: the plant material, the changed condition, the measured response and the interval. For example, a study might concern young shoots of one genotype, a change in light direction, curvature along a marked region and observations over several hours. A second study of mature roots under water limitation asks a different question even if both use the word response.
Separate description, prediction and explanation
A descriptive rule might say that curvature rises after a light change. A predictive rule estimates the angle at a later time. A mechanistic explanation connects perception, regulation, spatially unequal growth and geometry. One account can achieve more than one of these jobs, but success at one does not automatically establish the others.
Start a notebook with the sentence, “If this explanation is right, the following quantity should change under the following comparison.” That sentence exposes whether a diagram has testable consequences. If every possible result can be described as consistent with the explanation, the proposed test is not yet informative.
The practical benefit is restraint. A question about a short growth response may need a few state variables and careful imaging. A question about resource allocation over a whole day may need organ exchange and storage. Model size should follow the scientific job.
CHAPTER 2 OF 17 · Define the plant system
2. Draw the boundary before adding the arrows
A leaf, a cell and an entire plant do not share the same bookkeeping boundary. Material exported from a leaf is a loss from the leaf’s account but can remain inside the whole plant. Material moving from cytoplasm to an organelle changes local amounts while leaving the cell total unchanged, provided no other process occurs.
Consider an original two-compartment model containing 12 units of a substance in compartment A and 8 in B. Three units move from A to B. The new amounts are 9 and 11, and the total remains 20. An equation that subtracts three from A without adding it to B would destroy material unless the missing material is explicitly assigned to a third destination.
Geometry changes the interpretation further. If the compartments have volumes of 3 and 1 volume units, their initial concentrations are 4 and 8 amount units per volume unit. After transfer they are 3 and 11. Equal amounts would not mean equal concentrations, and a statement about concentration must identify the volume used.
Growth can change a concentration without removing material
Suppose a growing model cell contains 10 amount units while its volume increases from 2 to 4. With no synthesis, removal or exchange, concentration falls from 5 to 2.5. Calling that a 50% loss of the substance would be wrong: amount is unchanged. A concentration model of growing tissue needs to account for dilution.
Keep a boundary sketch beside every equation. Label storage, exchange and growth separately. This gives the reader a way to find an accounting error before numerical output makes it harder to see.
CHAPTER 3 OF 17 · Define the plant system
3. Connect omics layers without treating them as interchangeable
A genome describes sequence information. Transcript measurements concern RNA. Proteomic measurements concern proteins or their detectable peptides. Metabolomic measurements concern chemical features or identified compounds. These layers can constrain a shared biological explanation, but they do not report one common quantity in four different colours.
An original example records a twofold RNA increase for a model enzyme, no detected change in protein amount and a smaller downstream metabolite pool. This is not automatically a contradictory dataset. RNA and protein can have different response times; protein activity can change without a change in amount; and a metabolite pool reflects both production and consumption. The observations narrow questions rather than complete the causal chain.
Ask whether the layers were measured in matching tissue, at matching times and on comparable experimental units. RNA from a young leaf and protein from a whole mature shoot cannot be joined as if they were two readings from the same cell. A shared gene label does not repair a mismatch in biological context.
The existing guides provide the measurement foundations: Plant Transcriptomics for RNA evidence, Plant Proteomics for protein evidence and Plant Metabolomics for chemical measurements.
A useful integration table records what each layer measures, its uncertainty, its sampling unit and the model variable it constrains. Leave a blank when no defensible mapping exists. A missing connection stated honestly is more informative than a continuous arrow built from incompatible measurements.
CHAPTER 4 OF 17 · Follow networks into growth
4. Give each network edge a meaning and an evidence trail
An arrow can mean conversion of material, transport, regulation or statistical association. These are different relationships. A transport edge carries an amount between locations. A regulatory edge changes a process. An association edge records a pattern in observations without establishing what causes it.
Suppose two genes rise whenever plants are sampled later in the morning. A network algorithm may connect them, but both may be responding to time of day. That edge is a candidate relationship. It becomes a stronger mechanistic claim only when evidence separates direct influence from shared timing, tissue composition or other explanations.
Pathway resources help organise prior knowledge. The official EMBL-EBI Plant Reactome introduction explains the use of rice as a reference and homology-based pathway projection to other species. That distinction matters when deciding whether a relationship was studied in the plant under investigation or transferred from another context.
An evidence ledger for a small network
For each proposed edge, write its biological meaning, species, tissue, evidence type and alternative explanation. Mark directly measured relationships separately from inferred ones. If an edge is included solely to make a model work, label it as an assumption rather than allowing the finished diagram to disguise its origin.
An informative network can contain uncertainty. The objective is not to remove every tentative connection but to keep the strength of the public claim aligned with the evidence. This also identifies the next experiment: test an uncertain connection that materially changes the predicted response.
CHAPTER 5 OF 17 · Follow networks into growth
5. Use auxin transport to see why location matters
Plant development provides a concrete reason to represent space. A regulator’s total amount does not specify where it is concentrated, which cells receive it or how transport connects neighbouring cells. Two tissues with the same total can therefore have different local patterns and different downstream behaviour.
Jönsson and colleagues’ 2006 primary study combined mathematical modelling with imaging to investigate phyllotactic patterning. Their model explored feedback between auxin and the polar distribution of transport machinery. It demonstrated that specified local transport rules could generate organised patterns. This is an example of testing whether a proposed mechanism is sufficient under a model’s assumptions, not proof that every plant pattern has one universal explanation.
A three-cell transport calculation
Imagine three model cells in a row with amounts 10, 10 and 10. During one short accounting interval, two units move from the left cell to the middle, and three move from the middle to the right. There is no synthesis or loss. The final amounts are 8, 9 and 13. Their sum remains 30.
Now reverse the second transfer while keeping its size. Two units still move left to middle, but three move right to middle. The result is 8, 15 and 7. Again the sum is 30. Changing direction has produced a different local maximum without changing the total.
These prescribed transfers are not an auxin transport law. They isolate the accounting consequence of direction. A biological model must additionally explain how transfer rates depend on transporters, concentrations, cell interfaces and time. Whole-tissue averages alone cannot distinguish the two constructed patterns.
CHAPTER 6 OF 17 · Follow networks into growth
6. Translate unequal growth into a visible shape
A molecular response becomes a change in form only through intermediate processes. For a bending shoot, one useful geometric question is whether opposite sides of a region lengthen by different amounts. The familiar picture of a curved organ can then be connected to an explicit calculation.
Take an idealised thin strip whose opposite sides remain separated by a fixed distance w and become concentric circular arcs. If both sides subtend an angle θ in radians, their length difference is ΔL = wθ. This follows because each arc length equals its radius multiplied by θ, and the two radii differ by w.
Let w = 1 millimetre and ΔL = 0.2 millimetres. Then θ = 0.2 radians, approximately 11.5 degrees. If the same length difference is distributed across a strip 2 millimetres wide, θ is 0.1 radians, approximately 5.7 degrees. Geometry changes the angle produced by the same differential extension.
Keep the idealisation visible
Real organs can change thickness, twist, bend nonuniformly and contain tissues with different mechanical properties. The calculation is therefore a teaching model, not a universal conversion from gene expression to root angle. It makes the missing bridge visible: molecular regulation must affect tissue behaviour, and tissue behaviour acts within geometry.
The specialist article on Phototropism, Auxin and Plant-Growth Evidence develops directional stimuli, growth and fair comparisons. Use it to distinguish the observation of a bend from the evidence required for its proposed cause. A final angle is an outcome, not a complete record of how it arose.
CHAPTER 7 OF 17 · Connect time and resources
7. Treat biological time as part of the system
Plants experience recurring changes in light and temperature, while internal timing mechanisms can organise responses across the day. A measurement taken at one clock time can reflect the state of a rhythm rather than a permanent difference between plants. Period, phase, amplitude and average level must be separated.
The 2012 Arabidopsis clock study by Pokhilko and colleagues used a mechanistic mathematical model to connect feedback relationships with timing data. It revised earlier relationships as evidence changed. The broader lesson is that a model is a versioned explanation: new molecular evidence can change its wiring, not merely adjust a numerical parameter.
Consider the original teaching functions A(t) = 10 + 4 cos(2πt/24) and B(t) = 10 + 4 cos(2π(t − 6)/24), with t in hours. Both have a 24-hour period, mean 10 and amplitude 4. B is delayed by six hours relative to A.
At t = 0, A equals 14 and B equals 10. At t = 6, A equals 10 and B equals 14. A single sampling time would reverse which group appears higher. Neither constructed group has a higher cycle mean.
These curves are not a fitted plant clock. They teach why a time series is needed to distinguish a phase shift from altered abundance. The existing Chronobiology guide provides the vocabulary for period, phase, entrainment and rhythmic outputs across biological systems.
CHAPTER 8 OF 17 · Connect time and resources
8. Design sampling around the difference between explanations
Collecting more observations is useful only if they address the uncertainty. If two candidate models predict the same endpoint but different early responses, another endpoint measurement may add precision while leaving the mechanism unresolved. Sampling during the predicted divergence is more informative.
Suppose a rapid-response model predicts a signal of 8 units after one hour and 10 after four hours. A delayed-response model predicts 3 after one hour and also 10 after four hours. The four-hour observation cannot separate these particular predictions. The one-hour observation can, provided measurement uncertainty and biological variability are small enough relative to the five-unit difference.
Experimental units matter as much as timing. Imaging six leaves from each of four plants gives 24 leaf observations, but the leaves do not automatically become 24 independent plants. If treatment was applied to one shared chamber, chamber conditions may also be entangled with treatment. Analysis and replication must reflect the level at which comparisons were established.
Preserve the time reference
Record whether time means hours after dawn, hours after transfer, chronological age or developmental stage. “Sampled at noon” does not guarantee comparable biological timing when light schedules differ. Likewise, equal days after germination need not imply equal developmental state under different growth conditions.
Before collecting data, draw the predictions and mark where they differ. Add the planned observation times, expected noise and independent units. That small design exercise can reveal a weak experiment before a large dataset is produced. It also makes a later negative result interpretable: was the proposed mechanism challenged, or were its distinguishing predictions never observed?
CHAPTER 9 OF 17 · Connect time and resources
9. Use flux constraints to rule out impossible accounts
A metabolic network connects reactions through material balances. At a steady state for a particular internal pool, total input equals total output. This does not mean that reactions stop. It means the pool does not accumulate over the interval represented by the approximation.
Consider a fictional carbon-equivalent pool with input 12 units per hour. It supplies a maintenance route fixed at 3 and two other routes, growth g and storage s. If the pool is steady and these are the only exits, 12 = 3 + g + s, so g + s = 9. Both g = 6, s = 3 and g = 2, s = 7 satisfy the balance.
The equation excludes impossible combinations but does not identify the actual allocation. If an additional measured capacity limits growth to at most 4, both earlier examples are reconsidered: the first becomes infeasible, while the second remains feasible. If a model chooses the largest feasible growth flux, it selects g = 4 and s = 5 under the stated constraints.
That selection depends on the objective. A plant is not established to maximise the chosen quantity merely because the optimiser does. Alternative objectives and omitted demands may yield different solutions. Predicted fluxes require comparison with appropriate physiological or tracer evidence.
The existing How Science Works: Systems Biology introduces flux balance alongside other model types. Plant Biochemistry supplies the reaction and compartment foundations. Keep the distinction between a chemically valid account, a feasible numerical solution and a measured plant behaviour.
CHAPTER 10 OF 17 · Connect time and resources
10. Couple day and night instead of optimising one snapshot
A daytime solution can look successful while leaving no resources for the following night. Connecting phases changes the question from “What is possible now?” to “What can be sustained across the specified cycle?” A storage variable creates an explicit connection between the two intervals.
Cheung and colleagues’ 2014 diel flux-balance study linked light and dark metabolism in one optimisation framework. Requirements during darkness constrained the stores accumulated during light. The study illustrates why separating phases without their connecting reserves can change model predictions. Its detailed network is far richer than the teaching ledger below.
Suppose a fictional plant model acquires 120 carbon-equivalent units during the day. Daytime maintenance consumes 30. Nighttime demand is 40. Assume storing and recovering material causes no loss and all night demand must come from the day’s input. At most 50 units remain for the other daytime allocation: 120 − 30 − 40 = 50.
Now assume only 80% of stored material is recoverable for the specified night demand. Delivering 40 requires storing 40/0.8 = 50. The remaining daytime allocation falls to 120 − 30 − 50 = 40. The extra ten units must be assigned to the model’s stated storage or recovery loss; they cannot vanish from the carbon ledger.
Change one demand and recalculate
If night demand rises to 48 under the same recovery assumption, storage must be 60 and the remaining daytime allocation becomes 30. A plan that maximises immediate allocation without reserving enough would violate its own night requirement. These arbitrary numbers teach coupled constraints, not real storage efficiencies or universal plant priorities.
CHAPTER 11 OF 17 · Challenge the model
11. Explain feedback through the variable it changes
Feedback exists when a downstream state influences an earlier process in the same causal loop. The sign of one arrow does not determine the entire response. Timing, strength, saturation and other inputs can change whether a disturbance is reduced, amplified or followed by oscillation.
Use a deliberately simple discrete teaching rule for a model reserve X. In one step, new X equals current X plus an input correction k(10 − X). There are no other flows in this constructed rule, and 10 is a target value expressed in the same units as X.
With k = 0.5 and X initially 6, successive values are 8, 9 and 9.5. The difference from 10 halves each step. With k = 1.5, the sequence is 12, 9 and 10.5: the response crosses the target but its error shrinks. With k = 2.5, the first values are 16 and 1, and the error grows in magnitude with alternating sign.
The error rule is e(next) = (1 − k)e(current). For this specific recurrence, error shrinks when the absolute value of 1 − k is below one. This is a mathematical property of the chosen update, not a universal stability threshold for biological feedback.
In a plant model, identify the sensed quantity and the process being regulated. Does the proposed loop change synthesis, transport, degradation or growth? If an equation eventually produces an impossible negative amount, examine its assumptions and numerical treatment. Calling the loop “negative feedback” does not excuse an invalid model state or prove perfect adaptation.
CHAPTER 12 OF 17 · Challenge the model
12. Separate parameter uncertainty from a missing mechanism
An uncertain parameter is a quantity whose value is not well determined within the model. Structural uncertainty concerns the model itself: which processes and relationships belong in it. Repeatedly adjusting parameters cannot always compensate for a missing process without creating misleading explanations.
Take an original model of a pulse-labelled pool after input stops: X(t) = X(0) exp(−kt). Here k is a removal coefficient in inverse hours. If X(0) is 100 and X(2) is 50, k = ln(2)/2, approximately 0.347 per hour. The model then predicts 25 at four hours.
Suppose the four-hour observation is instead 45. That discrepancy could involve measurement uncertainty, continued input, multiple pools with different removal rates or another violated assumption. It does not by itself prove which alternative is correct. The next investigation should target the plausible causes rather than quietly selecting a new coefficient for each time point.
Check what the data can actually identify
If the only observation is a stable amount of 100 in a production–removal model, production 10 with k = 0.1 and production 20 with k = 0.2 both fit. The endpoint identifies a ratio under that model, not the two rates independently. A suitable perturbation or independent rate measurement could distinguish them.
Sensitivity analysis asks how outputs change when assumptions or parameters vary. Report which conclusions remain stable and which depend on a narrow choice. A model that honestly identifies an unresolved rate can be more useful than one reporting many precise-looking parameters unsupported by the available observations.
CHAPTER 13 OF 17 · Challenge the model
13. Test prediction on a genuinely new comparison
A model fitted to a dataset has already used information from that dataset. Reproducing those observations is an important check, but it is weaker evidence of generalisation than a prediction made before seeing a new comparison. Keep fitting, model selection and final evaluation distinct.
For an original example, imagine fitting growth responses using several plants under two light schedules. Holding out random photographs from those same plants leaves considerable shared information between training and evaluation. Holding out whole plants asks a different question. Holding out an entire light schedule asks whether the model transfers across that environmental change.
Choose the split to match the intended claim. A model meant to interpolate among plants under familiar conditions need not meet the same test as one intended to predict a new genotype in a new environment. Describe both the successful comparison and the boundary it leaves untested.
Compare with a simple baseline. If yesterday’s measured growth predicts tomorrow’s observation nearly as well as a complicated model, establish what extra capability the complexity provides. It might improve prediction after an intervention, identify a mechanism or quantify uncertainty. A larger network alone is not that capability.
When a prediction fails, inspect the residual pattern. Errors confined to one developmental stage suggest a different problem from errors that alternate with time of day. Do not erase a structured failure by reporting only one average error. The pattern can reveal a missing context and guide the next revision of the plant model.
CHAPTER 14 OF 17 · Challenge the model
14. Cross scales through explicit measurements
Molecular evidence, cell behaviour, organ shape and whole-plant performance belong to connected but distinct scales. Moving between them requires a rule and a test. A change in expression cannot simply be relabelled as faster growth, and faster growth in one region cannot automatically be relabelled as higher final yield.
Suppose an original model predicts that a local transport change increases extension in a marked shoot region. The first check concerns that region’s extension. A second concerns resulting geometry. A third might concern leaf placement and light interception. A fourth concerns a defined whole-plant outcome over an appropriate duration. Failure at any connection changes the wider conclusion.
Aggregation also matters. Two equally weighted tissue regions may have signals of 2 and 8, giving an average of 5. Another plant may have 5 in both regions, also averaging 5. The average cannot reveal whether a response is spatially concentrated. If one region controls the measured developmental event, the difference may be essential.
Preserve the environment during transfer
A result in a controlled seedling system may motivate a crop hypothesis, but transfer must consider genotype, architecture, stage, environmental variation and the performance measure. “The mechanism exists” and “this intervention improves a field outcome” require different evidence.
This is where plant systems biology becomes useful beyond the laboratory. It can identify which connection must be checked before a molecular finding becomes a practical claim. The result may be a narrower conclusion, a better measurement plan or a reason to reject an apparently attractive shortcut.
CHAPTER 15 OF 17 · Transfer and continue
15. Transfer task: did a treatment improve the whole system?
A fictional study compares untreated and treated seedlings. The treatment group has twice the RNA signal for a transport-related gene in one sampled region. Both groups bend through the same final angle after four hours, but the treated group bends earlier. At a single sampling time, a clock-related reporter is also higher in treated plants.
A separate resource model assigns 120 carbon-equivalent units of daytime input to both groups. Day maintenance is 30. Untreated night demand is 40, with complete storage recovery. Treated night demand is 48, and its assumed storage recovery is 80%. No growth outcome has been measured beyond the bending experiment.
The proposed conclusion is: “The treatment improves transport, accelerates the clock and leaves more carbon for growth.” Identify which parts are supported, calculate the remaining daytime allocation under each resource model and propose one next comparison that addresses a specific uncertainty.
Answer and reasoning
The earlier bending supports a difference in the observed trajectory under the experiment’s conditions. Equal final angles do not erase that timing difference. However, RNA abundance is not a direct measurement of transport rate, and the observed trajectory does not by itself identify the causal route from the gene to growth.
A higher clock reporter at one time does not establish a faster clock. A phase difference, amplitude difference or baseline change could produce it. A suitably replicated time series, with a clear environmental time reference, would help distinguish these alternatives. That comparison answers a different question from measuring transport activity.
The untreated ledger leaves 120 − 30 − 40 = 50 units. The treated model must store 48/0.8 = 60, leaving 120 − 30 − 60 = 30. Under these assumptions, the treated model leaves 20 fewer units for the other daytime allocation. The proposed carbon conclusion is therefore opposite to the constructed calculation.
Even the corrected ledger does not establish actual growth. Its inputs, demands and recovery assumptions need evidence, and allocation is not identical to measured biomass. A defensible report separates the observed bending trajectory, the untested transport mechanism, the unresolved clock interpretation and the conditional resource calculation. One attractive sentence cannot replace those four distinct claims.
CHAPTER 16 OF 17 · Transfer and continue
16. Diagnose weak claims and repair the missing connection
A useful diagnosis names the exact inferential gap. “More research is needed” leaves the next action unclear. “Measure transport because the present result measures RNA” states what is missing and why it matters.
| Claim or symptom | Missing distinction | Useful repair |
|---|---|---|
| The network is dense, so the explanation is strong | Number of edges versus evidence for their meanings | Label edge types and inspect the relationships driving predictions |
| The reporter is higher, so the clock runs faster | Level, phase, amplitude and period | Collect a time series with a defined timing reference |
| Total regulator amount is unchanged, so nothing changed | Whole-tissue total versus spatial distribution | Measure or test the relevant local pattern |
| A flux solution exists, so the plant uses it | Feasibility versus realised behaviour | Compare with independent physiological or tracer constraints |
| The model fits every endpoint | Endpoint agreement versus identifiable dynamics | Test a time or perturbation where alternatives diverge |
| One organ grows faster, so the crop performs better | Local response versus whole-plant outcome | Define and measure the intended outcome in the relevant context |
Is plant systems biology just collecting several omics datasets?
No. Multiple datasets can support a systems question, but the essential work is connecting them through explicit relationships and testable predictions. A small well-defined model can answer a systems question without measuring every molecular layer.
Does every model need differential equations?
No. The representation should match the question and evidence. Logical rules, spatial models, constraint-based models and statistical models answer different kinds of questions. Their assumptions must remain visible; a simpler representation is not automatically less scientific.
Can a model be useful when some parameters are unknown?
Yes, if the useful conclusions survive the stated uncertainty. It can rule out impossible accounts, identify informative experiments or show that available data cannot distinguish alternatives. Unsupported numerical precision should not be substituted for that result.
Does a successful prediction prove the mechanism?
It supports the model under the tested comparison. Other mechanisms may make similar predictions. Confidence grows when informative tests challenge competing explanations, especially when those tests were chosen before seeing their results.
CHAPTER 17 OF 17 · Transfer and continue
17. Build a practical learning route and preserve the model
Begin with a paper ledger. Draw two compartments, assign amounts and volumes, and check every transfer. Next use a spreadsheet to plot the two clock functions from chapter 7. Sample them at one time, then several times, and explain why the initial comparison was incomplete. These activities build the reasoning before software complexity enters.
Continue with the three-cell example. Change one transport direction while preserving total amount. Then connect unequal extension to the idealised bending calculation. The objective is to explain each connection in ordinary language, with units, before increasing the number of cells or parameters.
For a research-oriented route, select one published model and recover its question, equations or rules, initial conditions, parameter values, input schedule and comparison data. Reproducing its figure is a computational check. Understanding why the chosen comparison tests the biological claim is a separate learning task.
The official SBML specifications describe a standard for representing biological models, including extensions for different model classes. A shared format helps communicate the model; it does not establish that its biological assumptions are correct. Preserve the model version, units, software settings, data transformations and the exact experiment being simulated alongside any reusable file.
Continue through connected evidence
Use Plant Biochemistry for reaction and compartment reasoning, the linked omics guides for what instruments measure, and Systems Biology for the general modelling framework. The Gold Standard master and How X Works hub connect the broader subject routes.
The linked primary studies and official resources were checked on 25 October 2026. They support the specific research examples and representation guidance; the arithmetic cases in this article are original simplified constructions. No new analysis of the studies’ raw data is claimed. The lasting skill is to trace a plant response across quantities, places and times, then choose a comparison that could show the explanation needs to change.