WHY MATHEMATICS?

Why Mathematics? | scSeqComm Probabilistic Expression Scores: Fuzzy AND, Personalized PageRank and Intracellular Evidence

How scSeqComm quantifies both sides of communication, from expression evidence under a cluster reference to receptor-centered pathway proximity and receiver transcription-factor activity.

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

Convert group expression into a cluster-calibrated zero-to-one score
Combine required subunits with a geometric mean
Use a minimum as a transparent fuzzy logical AND
Weight receiver TF activity by personalized PageRank proximity

Chapter map

1. Cluster matrix
2. Ground mean
3. Ground spread
4. Mean standard error
5. Normal approximation
6. CDF score
7. Subunit geometric mean
8. Ligand evidence
9. Receptor evidence
10. Minimum fuzzy AND
11. Pathway network
12. Restart distribution
13. TF proximity
14. TF activity
15. Weighted intracellular score
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. Cluster matrix

A small example makes the mechanism visible. Each cell type has a gene-by-cell expression matrix.

A concrete situation

Cluster A contains 200 cells.

What the mathematics does not license

Cells from one donor are not 200 replicates. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Report donor structure. 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. Ground mean

This is where modelling choices begin to affect evidence. The whole cluster matrix supplies a background location.

A concrete situation

Mean expression is 0.8.

What the mathematics does not license

Normalization changes the reference. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Freeze 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.

3. Ground spread

The central mathematical idea is concrete. Matrix variability supplies a background scale.

A concrete situation

Standard deviation is 1.2.

What the mathematics does not license

Zeros and outliers affect spread. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Inspect the 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.

4. Mean standard error

Here the geometry becomes operational. Cluster size shrinks uncertainty for average expression.

A concrete situation

Spread divides by a size term.

What the mathematics does not license

Cell dependence violates simple scaling. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Use 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.

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. Normal approximation

This step deserves a careful pause. The central limit theorem motivates a reference for random-gene means.

A concrete situation

A bell curve is fitted.

What the mathematics does not license

Rare or heavy-tailed data can deviate. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Check approximation. 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. CDF score

The useful question is not only how to calculate, but what the calculation assumes. The observed gene mean is converted to cumulative probability.

A concrete situation

A ligand scores 0.92.

What the mathematics does not license

This is not signaling probability. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Name it expression evidence. 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. Subunit geometric mean

A small example makes the mechanism visible. Multiple chain scores combine multiplicatively.

A concrete situation

sqrt(0.9 times 0.4)=0.6.

What the mathematics does not license

One zero closes the complex. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

List every subunit. 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. Ligand evidence

This is where modelling choices begin to affect evidence. Sender ligand or complex receives one zero-to-one score.

A concrete situation

Ligand score is 0.8.

What the mathematics does not license

A high score is cluster-relative. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Keep raw expression. 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. Receptor evidence

The central mathematical idea is concrete. Receiver receptor receives its own calibrated score.

A concrete situation

Receptor score is 0.6.

What the mathematics does not license

Different clusters have different backgrounds. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Report both references. 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. Minimum fuzzy AND

Here the geometry becomes operational. Intercellular score is the smaller partner score.

A concrete situation

min(0.8,0.6)=0.6.

What the mathematics does not license

The stronger partner magnitude is discarded. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Retain components. 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. Pathway network

This step deserves a careful pause. Prior receptor-to-TF edges define possible intracellular routes.

A concrete situation

R reaches TF1 and TF2.

What the mathematics does not license

Database edges may be context-inactive. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Version the graph. 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. Restart distribution

The useful question is not only how to calculate, but what the calculation assumes. Personalized PageRank repeatedly restarts at the receptor.

A concrete situation

Restart probability is fixed.

What the mathematics does not license

Parameter choice changes locality. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Sweep restart. 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. TF proximity

A small example makes the mechanism visible. Stationary mass ranks TF closeness to the receptor.

A concrete situation

TF1 receives twice TF2 mass.

What the mathematics does not license

Hubs can attract flow. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Use degree-aware 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. TF activity

This is where modelling choices begin to affect evidence. Target-gene evidence estimates whether each TF is active.

A concrete situation

TF1 activity is 0.7.

What the mathematics does not license

Target sets overlap. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Show leading targets. 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. Weighted intracellular score

The central mathematical idea is concrete. TF activities are averaged using receptor-centered PageRank weights.

A concrete situation

Weighted score is 0.65.

What the mathematics does not license

Priors and expression are dependent. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Separate components. 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

Here the geometry becomes operational. Version matrices, Normal parameters, subunits, minima, graphs, restart and activity.

A concrete situation

A student solves a three-TF example.

What the mathematics does not license

Every zero-to-one score has a denominator. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Write an scSeqComm 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

scSeqComm starts by turning a ligand or receptor’s average expression within a cell cluster into a probability-like score between zero and one. The background distribution is the distribution of average expression expected for random genes in that cluster. Using the central limit theorem, the method approximates that background with a Normal distribution whose location comes from the whole cluster matrix and whose spread shrinks with cluster size. The score is a cumulative probability under that reference, not the probability that signaling occurred.

For multisubunit ligands and receptors, subunit scores are combined with a geometric mean, implementing a soft all-components requirement. The intercellular score for a ligand-receptor pair is the minimum of the ligand and receptor scores. This minimum is a fuzzy logical AND: the weaker partner limits the result. The rule is transparent and monotone, but it deliberately discards how much larger the stronger partner is.

Inside the receiver, scSeqComm connects receptors to transcription factors through signaling pathways. Personalized PageRank measures the proximity of a TF to a receptor in a directed prior network. TF activity is estimated from target-gene expression evidence. The intracellular score is a weighted average of TF activities, using receptor-to-TF PageRank proximity as weights. A high value therefore needs both an accessible route and active downstream regulators.

The framework can attach functional enrichment to the prioritized communication. Gene-set meaning remains conditional on the database and tested universe. Combining intercellular and intracellular evidence strengthens coherence, but shared priors create dependence. The responsible conclusion is a quantified, functionally annotated hypothesis whose weakest mathematical component remains visible.

Frequently asked questions

What is scSeqComm?

scSeqComm identifies ligand-receptor communication, quantifies receiver intracellular signaling and attaches functional interpretation from single-cell expression and prior networks.

How is one gene scored?

Its cluster-average expression is placed in a Normal approximation to the background distribution of random-gene averages for that cluster, producing a zero-to-one cumulative score.

Why does cluster size matter?

The spread of an average decreases with more cells, so the same observed mean can receive stronger evidence in a larger, less variable cluster.

How are multisubunit molecules handled?

Subunit scores are combined by geometric mean, so a zero component makes the whole complex score zero.

Why take the minimum of ligand and receptor scores?

The minimum implements a fuzzy AND in which the weaker partner limits the intercellular score.

What does personalized PageRank contribute?

It quantifies how strongly a receptor is connected to candidate TFs through a directed pathway network, which then weights downstream activity evidence.

Primary source and fact-check note

Giacomo Baruzzo, Giulia Cesaro and Barbara Di Camillo, “Identify, quantify and characterize cellular communication from single-cell RNA sequencing data with scSeqComm,” Bioinformatics 38(7), 1920–1929 (2022), DOI 10.1093/bioinformatics/btac036. 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

Matrix audit

Record preprocessing, cluster sizes and background parameters.

Approximation audit

Check the Normal reference against empirical summaries.

Subunit audit

List all components and geometric aggregation.

Fuzzy audit

Keep ligand and receptor scores beside their minimum.

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.

Graph audit

Version pathway directions and receptor seeds.

PageRank audit

Publish restart, weights and convergence.

Activity audit

Expose TF targets and weighted contributions.

Final claim

A zero-to-one score is calibrated evidence, not signaling 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: Cluster matrix

Begin with this mechanism: Each cell type has a gene-by-cell expression matrix. Recreate the concrete case: Cluster A contains 200 cells. Write cluster means, background means, spreads, cell counts, CDF scores, geometric subunit scores, minima, PageRank weights and TF activities with direction, denominators and graph or probability assumptions. Calculate the smallest non-trivial example by hand, then reproduce it in code. Challenge this limitation: Cells from one donor are not 200 replicates. Complete the transfer step: Report donor structure. Change exactly one annotation, edge threshold, graph aggregation, path transform, damping value, gene universe, expression cutoff, cluster size, Normal scale, subunit rule, probability bin, summary statistic or combination weight. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, transformation, null evidence, graph summary, pathway interpretation and biological conclusion on separate notebook lines. Reserve whole donors, tissues or conditions outside tuning when testing transfer.

Workshop 2: Ground mean

Begin with this mechanism: The whole cluster matrix supplies a background location. Recreate the concrete case: Mean expression is 0.8. Write cluster means, background means, spreads, cell counts, CDF scores, geometric subunit scores, minima, PageRank weights and TF activities with direction, denominators and graph or probability assumptions. Calculate the smallest non-trivial example by hand, then reproduce it in code. Challenge this limitation: Normalization changes the reference. Complete the transfer step: Freeze preprocessing. Change exactly one annotation, edge threshold, graph aggregation, path transform, damping value, gene universe, expression cutoff, cluster size, Normal scale, subunit rule, probability bin, summary statistic or combination weight. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, transformation, null evidence, graph summary, pathway interpretation and biological conclusion on separate notebook lines. Reserve whole donors, tissues or conditions outside tuning when testing transfer.

Workshop 3: Ground spread

Begin with this mechanism: Matrix variability supplies a background scale. Recreate the concrete case: Standard deviation is 1.2. Write cluster means, background means, spreads, cell counts, CDF scores, geometric subunit scores, minima, PageRank weights and TF activities with direction, denominators and graph or probability assumptions. Calculate the smallest non-trivial example by hand, then reproduce it in code. Challenge this limitation: Zeros and outliers affect spread. Complete the transfer step: Inspect the distribution. Change exactly one annotation, edge threshold, graph aggregation, path transform, damping value, gene universe, expression cutoff, cluster size, Normal scale, subunit rule, probability bin, summary statistic or combination weight. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, transformation, null evidence, graph summary, pathway interpretation and biological conclusion on separate notebook lines. Reserve whole donors, tissues or conditions outside tuning when testing transfer.

Workshop 4: Mean standard error

Begin with this mechanism: Cluster size shrinks uncertainty for average expression. Recreate the concrete case: Spread divides by a size term. Write cluster means, background means, spreads, cell counts, CDF scores, geometric subunit scores, minima, PageRank weights and TF activities with direction, denominators and graph or probability assumptions. Calculate the smallest non-trivial example by hand, then reproduce it in code. Challenge this limitation: Cell dependence violates simple scaling. Complete the transfer step: Use sensitivity checks. Change exactly one annotation, edge threshold, graph aggregation, path transform, damping value, gene universe, expression cutoff, cluster size, Normal scale, subunit rule, probability bin, summary statistic or combination weight. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, transformation, null evidence, graph summary, pathway interpretation and biological conclusion on separate notebook lines. Reserve whole donors, tissues or conditions outside tuning when testing transfer.

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: Normal approximation

Begin with this mechanism: The central limit theorem motivates a reference for random-gene means. Recreate the concrete case: A bell curve is fitted. Write cluster means, background means, spreads, cell counts, CDF scores, geometric subunit scores, minima, PageRank weights and TF activities with direction, denominators and graph or probability assumptions. Calculate the smallest non-trivial example by hand, then reproduce it in code. Challenge this limitation: Rare or heavy-tailed data can deviate. Complete the transfer step: Check approximation. Change exactly one annotation, edge threshold, graph aggregation, path transform, damping value, gene universe, expression cutoff, cluster size, Normal scale, subunit rule, probability bin, summary statistic or combination weight. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, transformation, null evidence, graph summary, pathway interpretation and biological conclusion on separate notebook lines. Reserve whole donors, tissues or conditions outside tuning when testing transfer.

Workshop 6: CDF score

Begin with this mechanism: The observed gene mean is converted to cumulative probability. Recreate the concrete case: A ligand scores 0.92. Write cluster means, background means, spreads, cell counts, CDF scores, geometric subunit scores, minima, PageRank weights and TF activities with direction, denominators and graph or probability assumptions. Calculate the smallest non-trivial example by hand, then reproduce it in code. Challenge this limitation: This is not signaling probability. Complete the transfer step: Name it expression evidence. Change exactly one annotation, edge threshold, graph aggregation, path transform, damping value, gene universe, expression cutoff, cluster size, Normal scale, subunit rule, probability bin, summary statistic or combination weight. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, transformation, null evidence, graph summary, pathway interpretation and biological conclusion on separate notebook lines. Reserve whole donors, tissues or conditions outside tuning when testing transfer.

Workshop 7: Subunit geometric mean

Begin with this mechanism: Multiple chain scores combine multiplicatively. Recreate the concrete case: sqrt(0.9 times 0.4)=0.6. Write cluster means, background means, spreads, cell counts, CDF scores, geometric subunit scores, minima, PageRank weights and TF activities with direction, denominators and graph or probability assumptions. Calculate the smallest non-trivial example by hand, then reproduce it in code. Challenge this limitation: One zero closes the complex. Complete the transfer step: List every subunit. Change exactly one annotation, edge threshold, graph aggregation, path transform, damping value, gene universe, expression cutoff, cluster size, Normal scale, subunit rule, probability bin, summary statistic or combination weight. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, transformation, null evidence, graph summary, pathway interpretation and biological conclusion on separate notebook lines. Reserve whole donors, tissues or conditions outside tuning when testing transfer.

Workshop 8: Ligand evidence

Begin with this mechanism: Sender ligand or complex receives one zero-to-one score. Recreate the concrete case: Ligand score is 0.8. Write cluster means, background means, spreads, cell counts, CDF scores, geometric subunit scores, minima, PageRank weights and TF activities with direction, denominators and graph or probability assumptions. Calculate the smallest non-trivial example by hand, then reproduce it in code. Challenge this limitation: A high score is cluster-relative. Complete the transfer step: Keep raw expression. Change exactly one annotation, edge threshold, graph aggregation, path transform, damping value, gene universe, expression cutoff, cluster size, Normal scale, subunit rule, probability bin, summary statistic or combination weight. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, transformation, null evidence, graph summary, pathway interpretation and biological conclusion on separate notebook lines. Reserve whole donors, tissues or conditions outside tuning when testing transfer.

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: Receptor evidence

Begin with this mechanism: Receiver receptor receives its own calibrated score. Recreate the concrete case: Receptor score is 0.6. Write cluster means, background means, spreads, cell counts, CDF scores, geometric subunit scores, minima, PageRank weights and TF activities with direction, denominators and graph or probability assumptions. Calculate the smallest non-trivial example by hand, then reproduce it in code. Challenge this limitation: Different clusters have different backgrounds. Complete the transfer step: Report both references. Change exactly one annotation, edge threshold, graph aggregation, path transform, damping value, gene universe, expression cutoff, cluster size, Normal scale, subunit rule, probability bin, summary statistic or combination weight. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, transformation, null evidence, graph summary, pathway interpretation and biological conclusion on separate notebook lines. Reserve whole donors, tissues or conditions outside tuning when testing transfer.

Workshop 10: Minimum fuzzy AND

Begin with this mechanism: Intercellular score is the smaller partner score. Recreate the concrete case: min(0.8,0.6)=0.6. Write cluster means, background means, spreads, cell counts, CDF scores, geometric subunit scores, minima, PageRank weights and TF activities with direction, denominators and graph or probability assumptions. Calculate the smallest non-trivial example by hand, then reproduce it in code. Challenge this limitation: The stronger partner magnitude is discarded. Complete the transfer step: Retain components. Change exactly one annotation, edge threshold, graph aggregation, path transform, damping value, gene universe, expression cutoff, cluster size, Normal scale, subunit rule, probability bin, summary statistic or combination weight. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, transformation, null evidence, graph summary, pathway interpretation and biological conclusion on separate notebook lines. Reserve whole donors, tissues or conditions outside tuning when testing transfer.

Workshop 11: Pathway network

Begin with this mechanism: Prior receptor-to-TF edges define possible intracellular routes. Recreate the concrete case: R reaches TF1 and TF2. Write cluster means, background means, spreads, cell counts, CDF scores, geometric subunit scores, minima, PageRank weights and TF activities with direction, denominators and graph or probability assumptions. Calculate the smallest non-trivial example by hand, then reproduce it in code. Challenge this limitation: Database edges may be context-inactive. Complete the transfer step: Version the graph. Change exactly one annotation, edge threshold, graph aggregation, path transform, damping value, gene universe, expression cutoff, cluster size, Normal scale, subunit rule, probability bin, summary statistic or combination weight. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, transformation, null evidence, graph summary, pathway interpretation and biological conclusion on separate notebook lines. Reserve whole donors, tissues or conditions outside tuning when testing transfer.

Workshop 12: Restart distribution

Begin with this mechanism: Personalized PageRank repeatedly restarts at the receptor. Recreate the concrete case: Restart probability is fixed. Write cluster means, background means, spreads, cell counts, CDF scores, geometric subunit scores, minima, PageRank weights and TF activities with direction, denominators and graph or probability assumptions. Calculate the smallest non-trivial example by hand, then reproduce it in code. Challenge this limitation: Parameter choice changes locality. Complete the transfer step: Sweep restart. Change exactly one annotation, edge threshold, graph aggregation, path transform, damping value, gene universe, expression cutoff, cluster size, Normal scale, subunit rule, probability bin, summary statistic or combination weight. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, transformation, null evidence, graph summary, pathway interpretation and biological conclusion on separate notebook lines. Reserve whole donors, tissues or conditions outside tuning when testing transfer.

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: TF proximity

Begin with this mechanism: Stationary mass ranks TF closeness to the receptor. Recreate the concrete case: TF1 receives twice TF2 mass. Write cluster means, background means, spreads, cell counts, CDF scores, geometric subunit scores, minima, PageRank weights and TF activities with direction, denominators and graph or probability assumptions. Calculate the smallest non-trivial example by hand, then reproduce it in code. Challenge this limitation: Hubs can attract flow. Complete the transfer step: Use degree-aware checks. Change exactly one annotation, edge threshold, graph aggregation, path transform, damping value, gene universe, expression cutoff, cluster size, Normal scale, subunit rule, probability bin, summary statistic or combination weight. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, transformation, null evidence, graph summary, pathway interpretation and biological conclusion on separate notebook lines. Reserve whole donors, tissues or conditions outside tuning when testing transfer.

Workshop 14: TF activity

Begin with this mechanism: Target-gene evidence estimates whether each TF is active. Recreate the concrete case: TF1 activity is 0.7. Write cluster means, background means, spreads, cell counts, CDF scores, geometric subunit scores, minima, PageRank weights and TF activities with direction, denominators and graph or probability assumptions. Calculate the smallest non-trivial example by hand, then reproduce it in code. Challenge this limitation: Target sets overlap. Complete the transfer step: Show leading targets. Change exactly one annotation, edge threshold, graph aggregation, path transform, damping value, gene universe, expression cutoff, cluster size, Normal scale, subunit rule, probability bin, summary statistic or combination weight. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, transformation, null evidence, graph summary, pathway interpretation and biological conclusion on separate notebook lines. Reserve whole donors, tissues or conditions outside tuning when testing transfer.

Workshop 15: Weighted intracellular score

Begin with this mechanism: TF activities are averaged using receptor-centered PageRank weights. Recreate the concrete case: Weighted score is 0.65. Write cluster means, background means, spreads, cell counts, CDF scores, geometric subunit scores, minima, PageRank weights and TF activities with direction, denominators and graph or probability assumptions. Calculate the smallest non-trivial example by hand, then reproduce it in code. Challenge this limitation: Priors and expression are dependent. Complete the transfer step: Separate components. Change exactly one annotation, edge threshold, graph aggregation, path transform, damping value, gene universe, expression cutoff, cluster size, Normal scale, subunit rule, probability bin, summary statistic or combination weight. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, transformation, null evidence, graph summary, pathway interpretation and biological conclusion on separate notebook lines. Reserve whole donors, tissues or conditions outside tuning when testing transfer.

Workshop 16: Governance and learning

Begin with this mechanism: Version matrices, Normal parameters, subunits, minima, graphs, restart and activity. Recreate the concrete case: A student solves a three-TF example. Write cluster means, background means, spreads, cell counts, CDF scores, geometric subunit scores, minima, PageRank weights and TF activities with direction, denominators and graph or probability assumptions. Calculate the smallest non-trivial example by hand, then reproduce it in code. Challenge this limitation: Every zero-to-one score has a denominator. Complete the transfer step: Write an scSeqComm audit card. Change exactly one annotation, edge threshold, graph aggregation, path transform, damping value, gene universe, expression cutoff, cluster size, Normal scale, subunit rule, probability bin, summary statistic or combination weight. Explain which mathematical object changes, which conclusion survives and which claim must weaken. Keep observation, transformation, null evidence, graph summary, pathway interpretation and biological conclusion on separate notebook lines. Reserve whole donors, tissues or conditions outside tuning when testing transfer.