WHY MATHEMATICS?

Why Mathematics? | ICELLNET Communication Intensity: Maximum Scaling, Multisubunit Logic, Expression Products and Global Scores

How ICELLNET normalizes gene expression, respects required molecular subunits and aggregates individual interaction products into family and global views without confusing totals with biological certainty.

Why mathematics is important becomes easier to see when a real decision depends on it. This guide follows the mechanism from definitions and dimensions to a worked calculation, validation, practical learning and responsible use. It keeps factual evidence separate from interpretation and treats every illustrative scenario as an example rather than a promise of results.

Choose a reading route

Normalize each communication gene across the declared comparison set
Require the relevant subunits of multichain ligands and receptors
Multiply directional partner expression into interaction scores
Disaggregate global scores into families, edges and validation evidence

Chapter map

1. Central-cell question
2. Outward direction
3. Inward direction
4. Gene maximum
5. Scaled expression
6. Single-chain product
7. Zero rule
8. Multisubunit ligand
9. Multisubunit receptor
10. Family assignment
11. Family score
12. Global score
13. Reference partners
14. Condition normalization
15. Statistical comparison
16. Governance and learning

Visit the How Mathematics Works hub for the wider series and its connections to education, reasoning, science, computing, design and careers.

1. Central-cell question

The useful question is not only how to calculate, but what the calculation assumes. A focal population is compared with declared partners.

A concrete situation

Dendritic cells are central.

What the mathematics does not license

The setup shapes the candidate universe. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Write the biological question. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

2. Outward direction

A small example makes the mechanism visible. Central ligands pair with partner receptors.

A concrete situation

Central-to-partner score is computed.

What the mathematics does not license

Direction does not prove secretion. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Use candidate language. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

3. Inward direction

This is where modelling choices begin to affect evidence. Partner ligands pair with central receptors.

A concrete situation

Partner-to-central is separate.

What the mathematics does not license

Reversing roles changes values. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Calculate both routes. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

4. Gene maximum

The central mathematical idea is concrete. Each gene receives a comparison-set maximum.

A concrete situation

Gene X has maximum ten.

What the mathematics does not license

An outlier sets the denominator. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Inspect maxima. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

Did You Know? In this chapter, the most important number is useful only together with its construction rule. Changing scaling, sample membership or validation design can change the numerical answer without changing the arithmetic.

5. Scaled expression

Here the geometry becomes operational. Expression is divided by the gene maximum.

A concrete situation

Five divided by ten gives 0.5.

What the mathematics does not license

The value is panel dependent. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Freeze the partner set. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

6. Single-chain product

This step deserves a careful pause. Scaled ligand multiplies scaled receptor.

A concrete situation

0.5 times 0.8 gives 0.4.

What the mathematics does not license

A product is not a probability. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Retain both factors. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

7. Zero rule

The useful question is not only how to calculate, but what the calculation assumes. An absent required communication molecule yields zero.

A concrete situation

Missing receptor closes one edge.

What the mathematics does not license

Dropout can mimic absence. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Check detection. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

8. Multisubunit ligand

A small example makes the mechanism visible. Required ligand chains are combined by explicit logic.

A concrete situation

Two chains must co-occur.

What the mathematics does not license

One abundant chain cannot rescue another. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

List every component. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

9. Multisubunit receptor

This is where modelling choices begin to affect evidence. Receptor-chain requirements affect eligibility and strength.

A concrete situation

Three chains define one receptor.

What the mathematics does not license

Annotation of complexes can change. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Version the database. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

10. Family assignment

The central mathematical idea is concrete. Interactions are grouped into cytokine, checkpoint and other families.

A concrete situation

Five edges belong to one family.

What the mathematics does not license

Categories overlap conceptually. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Use documented labels. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

11. Family score

Here the geometry becomes operational. Individual products are aggregated within a family.

A concrete situation

Family total equals 2.4.

What the mathematics does not license

More listed pairs can raise totals. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Show count and distribution. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

12. Global score

This step deserves a careful pause. All selected interaction scores are combined for a cell pair.

A concrete situation

A-to-B totals 18.

What the mathematics does not license

One total hides composition. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Publish contribution plots. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

13. Reference partners

The useful question is not only how to calculate, but what the calculation assumes. A central profile can be compared with standardized partner profiles.

A concrete situation

Thirty-one reference types are available.

What the mathematics does not license

Platforms and contexts differ. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Treat cross-dataset scores cautiously. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

14. Condition normalization

A small example makes the mechanism visible. Totals may be normalized to a baseline condition.

A concrete situation

Treatment is divided by control.

What the mathematics does not license

A weak baseline inflates ratios. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Report raw and normalized. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

15. Statistical comparison

This is where modelling choices begin to affect evidence. Partner or condition scores can be compared with appropriate tests.

A concrete situation

Two partner distributions differ.

What the mathematics does not license

Interactions are dependent. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

State the testing unit. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

16. Governance and learning

The central mathematical idea is concrete. Version raw values, maxima, subunits, products and totals.

A concrete situation

A student rebuilds one family score.

What the mathematics does not license

Aggregation never repairs missing biology. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Write an ICELLNET score card. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

A seven-day practice lab

Day 1: define the decision, observation unit, X variables and Y or reference condition.
Day 2: reproduce the smallest worked example and check every dimension, unit and sign.
Day 3: build a transparent baseline and write what it assumes.
Day 4: fit the main method inside a training-only pipeline; record every parameter.
Day 5: challenge one assumption with a deliberately difficult dataset or operating regime.
Day 6: compare validation metrics with the practical workload created by errors.
Day 7: present one supported conclusion, one unresolved uncertainty and one safe next experiment.

Questions for a student, parent or mentor

Which part of the model is a mathematical definition, and which part is a choice?
What small calculation would catch a sign, scale or dimension error?
What evidence comes from training, validation and the untouched test?
What benign condition could look like a fault or prediction failure?
How would the answer change for another response, asset, school project or time period?
What should a human review before acting on the score?

A method-specific audit before deployment

ICELLNET treats communication as directional exchange between a central cell population and partner populations. For outward communication, ligands come from the central cells and receptors from partners; inward communication reverses those roles. Each gene is scaled by its maximum expression across the compared profiles so one highly expressed gene does not automatically dominate every score. Because the denominator depends on the comparison set, changing the partner panel can change the normalized values even when raw expression is unchanged.

For an individual ligand-receptor interaction, the scaled ligand and receptor values are combined by a product. A missing partner makes the score zero. Multisubunit molecules require explicit logical rules and joint representation, so a receptor complex is not credited merely because its easiest subunit is abundant. The arithmetic expresses an AND-like requirement: communication potential should weaken when any required component is absent or small.

Individual interaction scores can be summed into family, subfamily and global communication scores. This hierarchy helps readers move from one molecular edge to a cell-pair overview. It also creates an aggregation risk. A large global score can reflect many moderate interactions, a few dominant families, the size of the curated resource or the chosen scaling panel. ICELLNET therefore offers layered visualizations and contribution breakdowns rather than asking one total to carry every interpretation.

The primary paper validated selected predictions experimentally in a controlled dendritic-cell system, which is stronger than expression-only plausibility. Even so, users should not transfer every validated example to a new tissue. The responsible mathematics keeps raw expression, scaled expression, complex logic, interaction products, family totals, global totals and experimental evidence on separate levels. That makes the framework a lesson in normalization, aggregation and conditional interpretation.

Frequently asked questions

What is ICELLNET?

ICELLNET is a transcriptomics framework that combines an expert-curated ligand-receptor resource, directional communication scores, partner-cell references and layered visualizations.

What does maximum scaling do?

Each ligand or receptor gene is divided by its maximum expression across the compared profiles, keeping values comparable within that declared set.

How is one interaction scored?

The scaled ligand value in the source and scaled receptor value in the target are multiplied, with explicit rules for required molecular subunits.

What is a global communication score?

It is an aggregate of eligible individual interaction scores between a cell pair. Family and subfamily totals reveal which groups contribute.

Can scores from different partner panels be compared directly?

Not automatically. Maximum scaling depends on the profiles included, so the comparison set must be matched or the normalization rebuilt consistently.

Does the global score measure information in bits?

No. It is an expression-derived summary called communication intensity, not Shannon information, binding probability or causal effect.

Primary source and fact-check note

Floriane Noël, Lucile Massenet-Regad, Irit Carmi-Levy et al., “Dissection of intercellular communication using the transcriptome-based framework ICELLNET,” Nature Communications 12, 1089 (2021), DOI 10.1038/s41467-021-21244-x. Definitions and core claims in this article were checked against this primary publication. Calculations shown here were independently recomputed. Examples are educational illustrations; they are not reported industrial performance results.

Useful next reading

Keep the options open. Mathematics education supports clearer questions, transferable problem-solving skills and informed choices across science, engineering, computing, AI, statistics, design and many other careers. It does not guarantee intelligence, income, admission or employment; those outcomes depend on far more than one subject or model.

Extended laboratory: topic-specific validation and responsible use

Question audit

Declare central cells, partner cells and direction.

Panel audit

Freeze the comparison set used for gene maxima.

Subunit audit

List every required ligand and receptor chain.

Product audit

Keep scaled partners beside each interaction score.

Did You Know? A sophisticated adjustment can be algebraically correct and still answer the wrong scientific question when samples, labels, batches or validation boundaries are wrong.

Family audit

Show counts and contributions within each family.

Global audit

Disaggregate totals before interpreting them.

Validation audit

Separate paper validation from new-context evidence.

Final claim

Global communication intensity is an expression summary.

Did You Know? A sophisticated adjustment can be algebraically correct and still answer the wrong scientific question when samples, labels, batches or validation boundaries are wrong.

Mechanism workshop: calculate, challenge and transfer

Workshop 1: Central-cell question

Begin with this mechanism: A focal population is compared with declared partners. Recreate the concrete case: Dendritic cells are central. Write raw expression, gene maxima, scaled values, subunit states, products, family contributions and global totals with direction, units and denominators. Calculate the smallest non-trivial example by hand, then reproduce it in a spreadsheet or short script. Challenge this limitation: The setup shapes the candidate universe. Complete the transfer step: Write the biological question. Change exactly one annotation, normalization, comparison panel, required subunit, target set, correlation threshold, permutation block, random seed, score cutoff or multiplicity family. Explain which mathematical object changes, which interpretation survives and which claim must weaken. Keep observation, transformation, null evidence, summary label and biological conclusion on separate notebook lines. Reserve whole donors, tissues or conditions outside tuning when testing whether the reasoning transfers.

Workshop 2: Outward direction

Begin with this mechanism: Central ligands pair with partner receptors. Recreate the concrete case: Central-to-partner score is computed. Write raw expression, gene maxima, scaled values, subunit states, products, family contributions and global totals with direction, units and denominators. Calculate the smallest non-trivial example by hand, then reproduce it in a spreadsheet or short script. Challenge this limitation: Direction does not prove secretion. Complete the transfer step: Use candidate language. Change exactly one annotation, normalization, comparison panel, required subunit, target set, correlation threshold, permutation block, random seed, score cutoff or multiplicity family. Explain which mathematical object changes, which interpretation survives and which claim must weaken. Keep observation, transformation, null evidence, summary label and biological conclusion on separate notebook lines. Reserve whole donors, tissues or conditions outside tuning when testing whether the reasoning transfers.

Workshop 3: Inward direction

Begin with this mechanism: Partner ligands pair with central receptors. Recreate the concrete case: Partner-to-central is separate. Write raw expression, gene maxima, scaled values, subunit states, products, family contributions and global totals with direction, units and denominators. Calculate the smallest non-trivial example by hand, then reproduce it in a spreadsheet or short script. Challenge this limitation: Reversing roles changes values. Complete the transfer step: Calculate both routes. Change exactly one annotation, normalization, comparison panel, required subunit, target set, correlation threshold, permutation block, random seed, score cutoff or multiplicity family. Explain which mathematical object changes, which interpretation survives and which claim must weaken. Keep observation, transformation, null evidence, summary label and biological conclusion on separate notebook lines. Reserve whole donors, tissues or conditions outside tuning when testing whether the reasoning transfers.

Workshop 4: Gene maximum

Begin with this mechanism: Each gene receives a comparison-set maximum. Recreate the concrete case: Gene X has maximum ten. Write raw expression, gene maxima, scaled values, subunit states, products, family contributions and global totals with direction, units and denominators. Calculate the smallest non-trivial example by hand, then reproduce it in a spreadsheet or short script. Challenge this limitation: An outlier sets the denominator. Complete the transfer step: Inspect maxima. Change exactly one annotation, normalization, comparison panel, required subunit, target set, correlation threshold, permutation block, random seed, score cutoff or multiplicity family. Explain which mathematical object changes, which interpretation survives and which claim must weaken. Keep observation, transformation, null evidence, summary label and biological conclusion on separate notebook lines. Reserve whole donors, tissues or conditions outside tuning when testing whether the reasoning transfers.

Did You Know? A sophisticated adjustment can be algebraically correct and still answer the wrong scientific question when samples, labels, batches or validation boundaries are wrong.

Workshop 5: Scaled expression

Begin with this mechanism: Expression is divided by the gene maximum. Recreate the concrete case: Five divided by ten gives 0.5. Write raw expression, gene maxima, scaled values, subunit states, products, family contributions and global totals with direction, units and denominators. Calculate the smallest non-trivial example by hand, then reproduce it in a spreadsheet or short script. Challenge this limitation: The value is panel dependent. Complete the transfer step: Freeze the partner set. Change exactly one annotation, normalization, comparison panel, required subunit, target set, correlation threshold, permutation block, random seed, score cutoff or multiplicity family. Explain which mathematical object changes, which interpretation survives and which claim must weaken. Keep observation, transformation, null evidence, summary label and biological conclusion on separate notebook lines. Reserve whole donors, tissues or conditions outside tuning when testing whether the reasoning transfers.

Workshop 6: Single-chain product

Begin with this mechanism: Scaled ligand multiplies scaled receptor. Recreate the concrete case: 0.5 times 0.8 gives 0.4. Write raw expression, gene maxima, scaled values, subunit states, products, family contributions and global totals with direction, units and denominators. Calculate the smallest non-trivial example by hand, then reproduce it in a spreadsheet or short script. Challenge this limitation: A product is not a probability. Complete the transfer step: Retain both factors. Change exactly one annotation, normalization, comparison panel, required subunit, target set, correlation threshold, permutation block, random seed, score cutoff or multiplicity family. Explain which mathematical object changes, which interpretation survives and which claim must weaken. Keep observation, transformation, null evidence, summary label and biological conclusion on separate notebook lines. Reserve whole donors, tissues or conditions outside tuning when testing whether the reasoning transfers.

Workshop 7: Zero rule

Begin with this mechanism: An absent required communication molecule yields zero. Recreate the concrete case: Missing receptor closes one edge. Write raw expression, gene maxima, scaled values, subunit states, products, family contributions and global totals with direction, units and denominators. Calculate the smallest non-trivial example by hand, then reproduce it in a spreadsheet or short script. Challenge this limitation: Dropout can mimic absence. Complete the transfer step: Check detection. Change exactly one annotation, normalization, comparison panel, required subunit, target set, correlation threshold, permutation block, random seed, score cutoff or multiplicity family. Explain which mathematical object changes, which interpretation survives and which claim must weaken. Keep observation, transformation, null evidence, summary label and biological conclusion on separate notebook lines. Reserve whole donors, tissues or conditions outside tuning when testing whether the reasoning transfers.

Workshop 8: Multisubunit ligand

Begin with this mechanism: Required ligand chains are combined by explicit logic. Recreate the concrete case: Two chains must co-occur. Write raw expression, gene maxima, scaled values, subunit states, products, family contributions and global totals with direction, units and denominators. Calculate the smallest non-trivial example by hand, then reproduce it in a spreadsheet or short script. Challenge this limitation: One abundant chain cannot rescue another. Complete the transfer step: List every component. Change exactly one annotation, normalization, comparison panel, required subunit, target set, correlation threshold, permutation block, random seed, score cutoff or multiplicity family. Explain which mathematical object changes, which interpretation survives and which claim must weaken. Keep observation, transformation, null evidence, summary label and biological conclusion on separate notebook lines. Reserve whole donors, tissues or conditions outside tuning when testing whether the reasoning transfers.

Did You Know? A sophisticated adjustment can be algebraically correct and still answer the wrong scientific question when samples, labels, batches or validation boundaries are wrong.

Workshop 9: Multisubunit receptor

Begin with this mechanism: Receptor-chain requirements affect eligibility and strength. Recreate the concrete case: Three chains define one receptor. Write raw expression, gene maxima, scaled values, subunit states, products, family contributions and global totals with direction, units and denominators. Calculate the smallest non-trivial example by hand, then reproduce it in a spreadsheet or short script. Challenge this limitation: Annotation of complexes can change. Complete the transfer step: Version the database. Change exactly one annotation, normalization, comparison panel, required subunit, target set, correlation threshold, permutation block, random seed, score cutoff or multiplicity family. Explain which mathematical object changes, which interpretation survives and which claim must weaken. Keep observation, transformation, null evidence, summary label and biological conclusion on separate notebook lines. Reserve whole donors, tissues or conditions outside tuning when testing whether the reasoning transfers.

Workshop 10: Family assignment

Begin with this mechanism: Interactions are grouped into cytokine, checkpoint and other families. Recreate the concrete case: Five edges belong to one family. Write raw expression, gene maxima, scaled values, subunit states, products, family contributions and global totals with direction, units and denominators. Calculate the smallest non-trivial example by hand, then reproduce it in a spreadsheet or short script. Challenge this limitation: Categories overlap conceptually. Complete the transfer step: Use documented labels. Change exactly one annotation, normalization, comparison panel, required subunit, target set, correlation threshold, permutation block, random seed, score cutoff or multiplicity family. Explain which mathematical object changes, which interpretation survives and which claim must weaken. Keep observation, transformation, null evidence, summary label and biological conclusion on separate notebook lines. Reserve whole donors, tissues or conditions outside tuning when testing whether the reasoning transfers.

Workshop 11: Family score

Begin with this mechanism: Individual products are aggregated within a family. Recreate the concrete case: Family total equals 2.4. Write raw expression, gene maxima, scaled values, subunit states, products, family contributions and global totals with direction, units and denominators. Calculate the smallest non-trivial example by hand, then reproduce it in a spreadsheet or short script. Challenge this limitation: More listed pairs can raise totals. Complete the transfer step: Show count and distribution. Change exactly one annotation, normalization, comparison panel, required subunit, target set, correlation threshold, permutation block, random seed, score cutoff or multiplicity family. Explain which mathematical object changes, which interpretation survives and which claim must weaken. Keep observation, transformation, null evidence, summary label and biological conclusion on separate notebook lines. Reserve whole donors, tissues or conditions outside tuning when testing whether the reasoning transfers.

Workshop 12: Global score

Begin with this mechanism: All selected interaction scores are combined for a cell pair. Recreate the concrete case: A-to-B totals 18. Write raw expression, gene maxima, scaled values, subunit states, products, family contributions and global totals with direction, units and denominators. Calculate the smallest non-trivial example by hand, then reproduce it in a spreadsheet or short script. Challenge this limitation: One total hides composition. Complete the transfer step: Publish contribution plots. Change exactly one annotation, normalization, comparison panel, required subunit, target set, correlation threshold, permutation block, random seed, score cutoff or multiplicity family. Explain which mathematical object changes, which interpretation survives and which claim must weaken. Keep observation, transformation, null evidence, summary label and biological conclusion on separate notebook lines. Reserve whole donors, tissues or conditions outside tuning when testing whether the reasoning transfers.

Did You Know? A sophisticated adjustment can be algebraically correct and still answer the wrong scientific question when samples, labels, batches or validation boundaries are wrong.

Workshop 13: Reference partners

Begin with this mechanism: A central profile can be compared with standardized partner profiles. Recreate the concrete case: Thirty-one reference types are available. Write raw expression, gene maxima, scaled values, subunit states, products, family contributions and global totals with direction, units and denominators. Calculate the smallest non-trivial example by hand, then reproduce it in a spreadsheet or short script. Challenge this limitation: Platforms and contexts differ. Complete the transfer step: Treat cross-dataset scores cautiously. Change exactly one annotation, normalization, comparison panel, required subunit, target set, correlation threshold, permutation block, random seed, score cutoff or multiplicity family. Explain which mathematical object changes, which interpretation survives and which claim must weaken. Keep observation, transformation, null evidence, summary label and biological conclusion on separate notebook lines. Reserve whole donors, tissues or conditions outside tuning when testing whether the reasoning transfers.

Workshop 14: Condition normalization

Begin with this mechanism: Totals may be normalized to a baseline condition. Recreate the concrete case: Treatment is divided by control. Write raw expression, gene maxima, scaled values, subunit states, products, family contributions and global totals with direction, units and denominators. Calculate the smallest non-trivial example by hand, then reproduce it in a spreadsheet or short script. Challenge this limitation: A weak baseline inflates ratios. Complete the transfer step: Report raw and normalized. Change exactly one annotation, normalization, comparison panel, required subunit, target set, correlation threshold, permutation block, random seed, score cutoff or multiplicity family. Explain which mathematical object changes, which interpretation survives and which claim must weaken. Keep observation, transformation, null evidence, summary label and biological conclusion on separate notebook lines. Reserve whole donors, tissues or conditions outside tuning when testing whether the reasoning transfers.

Workshop 15: Statistical comparison

Begin with this mechanism: Partner or condition scores can be compared with appropriate tests. Recreate the concrete case: Two partner distributions differ. Write raw expression, gene maxima, scaled values, subunit states, products, family contributions and global totals with direction, units and denominators. Calculate the smallest non-trivial example by hand, then reproduce it in a spreadsheet or short script. Challenge this limitation: Interactions are dependent. Complete the transfer step: State the testing unit. Change exactly one annotation, normalization, comparison panel, required subunit, target set, correlation threshold, permutation block, random seed, score cutoff or multiplicity family. Explain which mathematical object changes, which interpretation survives and which claim must weaken. Keep observation, transformation, null evidence, summary label and biological conclusion on separate notebook lines. Reserve whole donors, tissues or conditions outside tuning when testing whether the reasoning transfers.

Workshop 16: Governance and learning

Begin with this mechanism: Version raw values, maxima, subunits, products and totals. Recreate the concrete case: A student rebuilds one family score. Write raw expression, gene maxima, scaled values, subunit states, products, family contributions and global totals with direction, units and denominators. Calculate the smallest non-trivial example by hand, then reproduce it in a spreadsheet or short script. Challenge this limitation: Aggregation never repairs missing biology. Complete the transfer step: Write an ICELLNET score card. Change exactly one annotation, normalization, comparison panel, required subunit, target set, correlation threshold, permutation block, random seed, score cutoff or multiplicity family. Explain which mathematical object changes, which interpretation survives and which claim must weaken. Keep observation, transformation, null evidence, summary label and biological conclusion on separate notebook lines. Reserve whole donors, tissues or conditions outside tuning when testing whether the reasoning transfers.