WHY MATHEMATICS?

Why Mathematics? | SingleCellSignalR Interaction Scoring: Regularized Expression Products, Stable Thresholds and Directed Communication Graphs

How SingleCellSignalR converts two group means into a bounded regularized score, why background scaling changes threshold behaviour and how directed graphs should be read without turning expression into proof of signalling.

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

Calculate a geometric mean that requires both molecular partners
Regularize the joint signal against matrix-wide expression
Evaluate cutoffs with ROC curves and external reference evidence
Build a directed graph while preserving every score component

Chapter map

1. Sender and receiver groups
2. Ligand mean
3. Receptor mean
4. Geometric joint signal
5. Background mean mu
6. Bounded LRscore
7. Monotonicity
8. Saturation
9. Directionality
10. LRdb eligibility
11. Specificity label
12. ROC curve
13. Stable cutoff
14. Graph construction
15. Pathway context
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. Sender and receiver groups

The central mathematical idea is concrete. Cell annotations define which cells supply ligand and receptor summaries.

A concrete situation

Group A sends to group B.

What the mathematics does not license

Mislabelled cells change both means. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Audit annotations. 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. Ligand mean

Here the geometry becomes operational. Average ligand expression is computed in the sender.

A concrete situation

The ligand mean is four.

What the mathematics does not license

A few outliers can inflate a mean. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Show distribution and prevalence. 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. Receptor mean

This step deserves a careful pause. Average receptor expression is computed in the receiver.

A concrete situation

The receptor mean is nine.

What the mathematics does not license

Dropout depresses the group mean. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Report detected fraction. 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. Geometric joint signal

The useful question is not only how to calculate, but what the calculation assumes. The square root of ligand times receptor balances the partners.

A concrete situation

sqrt(4 times 9) equals six.

What the mathematics does not license

Zero in either partner gives zero. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Calculate by hand. 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. Background mean mu

A small example makes the mechanism visible. The expression-matrix mean supplies the regularization scale.

A concrete situation

mu equals two.

What the mathematics does not license

Normalization changes mu. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Record preprocessing. 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. Bounded LRscore

This is where modelling choices begin to affect evidence. Joint signal is divided by itself plus mu.

A concrete situation

Six divided by eight gives 0.75.

What the mathematics does not license

Bounded does not mean probability. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Label the score correctly. 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. Monotonicity

The central mathematical idea is concrete. Holding mu fixed, the score rises with either partner.

A concrete situation

Raising ligand increases LRscore.

What the mathematics does not license

Changing mu can reverse cross-dataset comparisons. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Separate within and across datasets. 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. Saturation

Here the geometry becomes operational. Large joint signals approach one.

A concrete situation

A signal of 100 and 200 score similarly.

What the mathematics does not license

Compression hides magnitude differences. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Keep raw means. 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. Directionality

This step deserves a careful pause. A-to-B and B-to-A use different ligand and receptor groups.

A concrete situation

One direction scores 0.8, reverse 0.2.

What the mathematics does not license

An arrow is a model orientation. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Test both directions. 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. LRdb eligibility

The useful question is not only how to calculate, but what the calculation assumes. Only curated pairs in the selected resource enter scoring.

A concrete situation

One known pair is eligible.

What the mathematics does not license

Missing knowledge creates false negatives. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Version LRdb. 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. Specificity label

A small example makes the mechanism visible. Differential enrichment can mark partners as cell-type specific.

A concrete situation

Both genes are enriched in their groups.

What the mathematics does not license

Specific expression is not specific binding. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Keep labels separate. 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. ROC curve

This is where modelling choices begin to affect evidence. Thresholds trade sensitivity against false positives under a reference.

A concrete situation

A curve summarizes many cutoffs.

What the mathematics does not license

The reference is incomplete. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

State the benchmark. 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. Stable cutoff

The central mathematical idea is concrete. Regularization reduces cutoff variation across tested datasets.

A concrete situation

A 0.5 rule is explored.

What the mathematics does not license

Stability is empirical, not universal. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Run sensitivity checks. 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. Graph construction

Here the geometry becomes operational. Passing pairs become weighted directed edges.

A concrete situation

Three cell types form six possible directions.

What the mathematics does not license

Dense graphs depend on thresholds. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Publish the edge table. 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. Pathway context

This step deserves a careful pause. Receptors may be linked to intracellular pathway knowledge.

A concrete situation

A receptor maps to a signaling route.

What the mathematics does not license

Prior pathways are not observed responses. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Validate downstream activity. 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 useful question is not only how to calculate, but what the calculation assumes. Version data, annotations, means, mu, LRdb and cutoff.

A concrete situation

A student recomputes one edge.

What the mathematics does not license

Reproducibility needs all components. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Write an LRscore audit 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

SingleCellSignalR begins with two group-level quantities: the average ligand expression l in a proposed sender and the average receptor expression r in a proposed receiver. Their geometric mean, the square root of l times r, is large only when both partners contribute. The method then regularizes that joint signal with the mean expression mu of the full matrix: LRscore equals square root of lr divided by mu plus square root of lr. The result lies between zero and one and is directed because reversing sender and receiver changes which group supplies each partner.

Regularization matters because raw products vary sharply with library depth, normalization and dropout. Dividing by the sum of the joint signal and a background scale compresses very large values and makes the score more comparable across cluster pairs. The original study used external transcriptomic and proteomic references to examine receiver-operating-characteristic curves and the stability of thresholds. A practical cutoff of 0.5 was illustrated, but the authors explicitly did not claim a universal false-positive guarantee.

A stable threshold is useful only if the upstream data contract is stable. Cell types, normalization, human orthologue mapping, mean expression and the LRdb version all determine the score. A threshold inherited from another dataset is a starting rule, not a natural constant. Students can see the transfer lesson clearly: monotone rescaling preserves within-dataset order but can change whether a fixed cutoff is crossed.

The final output is an oriented communication graph supported by known ligand-receptor pairs and observed expression. It is not a direct measurement of secretion, receptor occupancy or downstream response. The strongest workflow stores l, r, their geometric mean, mu, LRscore, specificity labels and validation evidence separately. That separation lets a parent, student or researcher trace how a biological claim grew from arithmetic rather than treating a coloured arrow as proof.

Frequently asked questions

What is SingleCellSignalR?

It is an R framework that uses a curated ligand-receptor database and group-level expression to infer and visualize potential directed communication between cell populations.

What is the LRscore?

For ligand mean l, receptor mean r and matrix mean mu, LRscore is sqrt(lr) divided by mu plus sqrt(lr). It lies between zero and one.

Why use the geometric mean?

It requires both partners and is less dominated by one large value than a sum or arithmetic mean.

Why regularize with mu?

The background scale compresses raw products and was designed to make practical thresholds more stable across datasets with different depth and dropout.

Is 0.5 a universal cutoff?

No. The study illustrated 0.5 as a useful consensus threshold, while showing that no cutoff guarantees the same false-positive rate in every context.

Does an LRscore prove communication?

No. It prioritizes expression-supported known pairs; secretion, protein abundance, spatial contact, binding and downstream response need independent evidence.

Primary source and fact-check note

Simon Cabello-Aguilar, Mélissa Alame, Fabien Kon-Sun-Tack, Caroline Fau, Matthieu Lacroix and Jacques Colinge, “SingleCellSignalR: inference of intercellular networks from single-cell transcriptomics,” Nucleic Acids Research 48(10), e55 (2020), DOI 10.1093/nar/gkaa183. 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

Input audit

Record normalization, cluster labels and the complete expression matrix.

Mean audit

Keep ligand mean, receptor mean and prevalence.

Scale audit

Record mu and show how preprocessing changes it.

Formula audit

Recalculate the geometric signal and bounded 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.

Threshold audit

Sweep cutoffs instead of treating 0.5 as universal.

Benchmark audit

Name reference evidence and its limits.

Graph audit

Preserve direction, edge table and isolated nodes.

Final claim

LRscore is a prioritization score, not a communication probability.

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: Sender and receiver groups

Begin with this mechanism: Cell annotations define which cells supply ligand and receptor summaries. Recreate the concrete case: Group A sends to group B. Write ligand means, receptor means, prevalence, geometric signals, background means, LRscores, cutoffs and graph weights 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: Mislabelled cells change both means. Complete the transfer step: Audit annotations. 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: Ligand mean

Begin with this mechanism: Average ligand expression is computed in the sender. Recreate the concrete case: The ligand mean is four. Write ligand means, receptor means, prevalence, geometric signals, background means, LRscores, cutoffs and graph weights 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 few outliers can inflate a mean. Complete the transfer step: Show distribution and prevalence. 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: Receptor mean

Begin with this mechanism: Average receptor expression is computed in the receiver. Recreate the concrete case: The receptor mean is nine. Write ligand means, receptor means, prevalence, geometric signals, background means, LRscores, cutoffs and graph weights 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 depresses the group mean. Complete the transfer step: Report detected fraction. 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: Geometric joint signal

Begin with this mechanism: The square root of ligand times receptor balances the partners. Recreate the concrete case: sqrt(4 times 9) equals six. Write ligand means, receptor means, prevalence, geometric signals, background means, LRscores, cutoffs and graph weights 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: Zero in either partner gives zero. Complete the transfer step: Calculate by hand. 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: Background mean mu

Begin with this mechanism: The expression-matrix mean supplies the regularization scale. Recreate the concrete case: mu equals two. Write ligand means, receptor means, prevalence, geometric signals, background means, LRscores, cutoffs and graph weights 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: Normalization changes mu. Complete the transfer step: Record preprocessing. 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: Bounded LRscore

Begin with this mechanism: Joint signal is divided by itself plus mu. Recreate the concrete case: Six divided by eight gives 0.75. Write ligand means, receptor means, prevalence, geometric signals, background means, LRscores, cutoffs and graph weights 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: Bounded does not mean probability. Complete the transfer step: Label the score correctly. 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: Monotonicity

Begin with this mechanism: Holding mu fixed, the score rises with either partner. Recreate the concrete case: Raising ligand increases LRscore. Write ligand means, receptor means, prevalence, geometric signals, background means, LRscores, cutoffs and graph weights 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: Changing mu can reverse cross-dataset comparisons. Complete the transfer step: Separate within and across datasets. 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: Saturation

Begin with this mechanism: Large joint signals approach one. Recreate the concrete case: A signal of 100 and 200 score similarly. Write ligand means, receptor means, prevalence, geometric signals, background means, LRscores, cutoffs and graph weights 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: Compression hides magnitude differences. Complete the transfer step: Keep raw means. 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: Directionality

Begin with this mechanism: A-to-B and B-to-A use different ligand and receptor groups. Recreate the concrete case: One direction scores 0.8, reverse 0.2. Write ligand means, receptor means, prevalence, geometric signals, background means, LRscores, cutoffs and graph weights 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 arrow is a model orientation. Complete the transfer step: Test both directions. 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: LRdb eligibility

Begin with this mechanism: Only curated pairs in the selected resource enter scoring. Recreate the concrete case: One known pair is eligible. Write ligand means, receptor means, prevalence, geometric signals, background means, LRscores, cutoffs and graph weights 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: Missing knowledge creates false negatives. Complete the transfer step: Version LRdb. 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: Specificity label

Begin with this mechanism: Differential enrichment can mark partners as cell-type specific. Recreate the concrete case: Both genes are enriched in their groups. Write ligand means, receptor means, prevalence, geometric signals, background means, LRscores, cutoffs and graph weights 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: Specific expression is not specific binding. Complete the transfer step: Keep labels separate. 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: ROC curve

Begin with this mechanism: Thresholds trade sensitivity against false positives under a reference. Recreate the concrete case: A curve summarizes many cutoffs. Write ligand means, receptor means, prevalence, geometric signals, background means, LRscores, cutoffs and graph weights 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 reference is incomplete. Complete the transfer step: State the benchmark. 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: Stable cutoff

Begin with this mechanism: Regularization reduces cutoff variation across tested datasets. Recreate the concrete case: A 0.5 rule is explored. Write ligand means, receptor means, prevalence, geometric signals, background means, LRscores, cutoffs and graph weights 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: Stability is empirical, not universal. Complete the transfer step: Run sensitivity checks. 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: Graph construction

Begin with this mechanism: Passing pairs become weighted directed edges. Recreate the concrete case: Three cell types form six possible directions. Write ligand means, receptor means, prevalence, geometric signals, background means, LRscores, cutoffs and graph weights 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: Dense graphs depend on thresholds. Complete the transfer step: Publish the edge table. 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: Pathway context

Begin with this mechanism: Receptors may be linked to intracellular pathway knowledge. Recreate the concrete case: A receptor maps to a signaling route. Write ligand means, receptor means, prevalence, geometric signals, background means, LRscores, cutoffs and graph weights 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: Prior pathways are not observed responses. Complete the transfer step: Validate downstream activity. 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 data, annotations, means, mu, LRdb and cutoff. Recreate the concrete case: A student recomputes one edge. Write ligand means, receptor means, prevalence, geometric signals, background means, LRscores, cutoffs and graph weights 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: Reproducibility needs all components. Complete the transfer step: Write an LRscore audit 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.