WHY MATHEMATICS?

Why Mathematics? | Spateo for Molecular Holograms: Partial Non-Rigid Alignment, Optimal Mappings, Morphometric Vector Fields and Differential Geometry

How Spateo reconstructs large 3D molecular tissues and turns cross-stage spatial mappings into continuous morphometric fields for studying expansion, rotation, bending and developmental paths.

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

Build 3D molecular maps from partially overlapping sections
Separate rigid pose from continuous residual deformation
Fit a vector field to cross-stage displacement evidence
Interpret divergence, curl and trajectory geometry cautiously

Chapter map

1. Section point clouds
2. Partial overlap
3. Molecular dissimilarity
4. Spatial geometry
5. Probabilistic assignment
6. Rigid pose
7. Non-rigid residual
8. Multi-slice refinement
9. Mesh correction
10. Cross-stage mapping
11. Displacement samples
12. Kernel vector field
13. Field integration
14. Divergence and curl
15. Curvature and torsion
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. Section point clouds

Here the geometry becomes operational. Each slice contains coordinates and molecular features.

A concrete situation

Thousands of embryo sections contribute millions of cells.

What the mathematics does not license

Sampling density and segmentation vary. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Audit coverage and units. 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. Partial overlap

This step deserves a careful pause. Pairwise mapping permits unsupported tissue to remain weakly matched.

A concrete situation

A growing organ exists only at the later stage.

What the mathematics does not license

Overlap parameters affect correspondence. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Plot matched mass. 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. Molecular dissimilarity

The useful question is not only how to calculate, but what the calculation assumes. Gene profiles contribute a cross-cell cost.

A concrete situation

Low KL-like dissimilarity supports candidate matches.

What the mathematics does not license

Expression convergence can mimic ancestry. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Hold out marker programs. 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. Spatial geometry

A small example makes the mechanism visible. Relative coordinates constrain the assignment.

A concrete situation

Neighbouring cells should not scatter arbitrarily.

What the mathematics does not license

Growth changes scale and topology. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Test geometry weights. 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. Probabilistic assignment

This is where modelling choices begin to affect evidence. A matrix distributes mapping weight between cells.

A concrete situation

One earlier cell region maps to several later candidates.

What the mathematics does not license

Weights are not lineage probabilities. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Report entropy. 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. Rigid pose

The central mathematical idea is concrete. Rotation and translation remove global orientation differences.

A concrete situation

An embryo section is centred and rotated.

What the mathematics does not license

Rigid fit can be biased by partial overlap. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Use robust anchors. 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. Non-rigid residual

Here the geometry becomes operational. A continuous deformation explains local mismatch.

A concrete situation

A limb bud bends while the trunk stays stable.

What the mathematics does not license

Flexible fields overfit noise. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Inspect Jacobians. 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. Multi-slice refinement

This step deserves a careful pause. Neighbouring alignments are revisited across the stack.

A concrete situation

A middle section balances evidence from both sides.

What the mathematics does not license

Errors can propagate through order. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Run leave-one-slice tests. 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. Mesh correction

The useful question is not only how to calculate, but what the calculation assumes. A surface mesh is adjusted to remove reconstruction artifacts.

A concrete situation

A reconstructed organ boundary becomes continuous.

What the mathematics does not license

Smoothing can erase genuine ridges. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Compare raw and corrected meshes. 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. Cross-stage mapping

A small example makes the mechanism visible. Cells or regions between developmental stages receive transport relationships.

A concrete situation

E9.5 locations map toward E11.5 structures.

What the mathematics does not license

Different proliferation rates alter mass. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

State growth assumptions. 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. Displacement samples

This is where modelling choices begin to affect evidence. Mapped coordinates define observed direction vectors.

A concrete situation

A target minus source position gives a vector.

What the mathematics does not license

Many-to-many maps need weighted averages. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Calculate confidence. 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. Kernel vector field

The central mathematical idea is concrete. Nearby displacement samples influence one another.

A concrete situation

A Gaussian kernel predicts motion at an empty location.

What the mathematics does not license

Bandwidth controls smoothness. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Publish a bandwidth curve. 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. Field integration

Here the geometry becomes operational. An ODE traces positions through the predicted field.

A concrete situation

A path follows dx/dt equals v(x).

What the mathematics does not license

Numerical steps and field errors accumulate. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Check step convergence. 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. Divergence and curl

This step deserves a careful pause. Spatial derivatives summarize expansion and rotation.

A concrete situation

Positive divergence marks an expanding region.

What the mathematics does not license

Derivatives amplify noise. The same reasoning helps students move from an exercise to a real dataset.

Try it, check it, transfer it

Bootstrap and smooth transparently. 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. Curvature and torsion

The useful question is not only how to calculate, but what the calculation assumes. Trajectory derivatives describe bending and three-dimensional twisting.

A concrete situation

A cardiac path curves and leaves a plane.

What the mathematics does not license

Short noisy paths produce unstable values. That distinction matters because a correct formula can still answer the wrong operational question.

Try it, check it, transfer it

Use arc-length and uncertainty filters. 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

A small example makes the mechanism visible. Version genes, overlap, mapping, kernels, meshes and solvers.

A concrete situation

A student fits a two-dimensional kernel field.

What the mathematics does not license

A morphometric path is a model. The habit of checking assumptions is one of the practical benefits of learning mathematics.

Try it, check it, transfer it

Write a vector-field ledger. 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

Spateo builds molecular holograms from many spatial sections with scalable partial and non-rigid alignment, multi-slice refinement and mesh correction. Its alignment machinery estimates a probabilistic cell-to-cell assignment and a rigid transformation while allowing partial overlap, then represents residual deformation with a continuous vector field.

Across developmental stages, optimal mappings provide candidate directions from cells in one state toward locations in another. Kernel or Gaussian-process vector-field models turn discrete displacement evidence into a continuous morphometric field. Integrating that field generates trajectories, while divergence, curl, acceleration, curvature and torsion describe expansion, rotation and bending.

A vector field is an inferred population-level geometry, not a recorded lineage for each cell. Mapping, alignment, sampling, kernel width and developmental timing affect every path. Validate with known anatomy, alternative genes, subsampling, forward-reverse consistency, independent lineage evidence and sensitivity of differential-geometry summaries.

Students can calculate weighted displacement from a three-by-three assignment, fit a Gaussian-kernel average at one unobserved point and take finite differences to estimate divergence. Repeating the calculation after removing one mapping reveals how correspondence uncertainty reaches the final geometry. This is the habit to transfer: trace each scientific statement backward through derivatives, paths, fields, mappings and measurements.

Frequently asked questions

What is Spateo?

Spateo is a framework for reconstructing and analysing large 3D spatial-transcriptomic tissues over developmental time.

What does partial alignment mean?

Only supported overlap contributes strongly, so missing or stage-specific tissue is not forced into a complete one-to-one match.

What is a morphometric vector field?

It assigns a predicted developmental displacement vector to locations in a reconstructed organ.

How is a continuous field learned from cells?

Discrete mapping-derived directions are smoothed or regressed with kernels or Gaussian-process models, depending on the workflow.

What do divergence and curl describe?

Divergence measures local expansion or contraction; curl measures local rotational tendency of the vector field.

Are predicted trajectories true cell lineages?

No. They are field-integrated paths supported by spatial state mappings, not direct observation of individual-cell ancestry.

Primary source and fact-check note

Xiaojie Qiu, Daniel Y. Zhu, Yifan Lu et al., “Spatiotemporal modeling of molecular holograms,” Cell 187, 7351–7373.e61 (2024), DOI 10.1016/j.cell.2024.10.011. 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

Coverage audit

Map partial overlap and unmatched tissue.

Assignment audit

Report transport mass and entropy.

Warp audit

Inspect local Jacobians and reverse alignment.

Stack audit

Leave sections out during refinement.

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.

Mesh audit

Compare corrected and raw surfaces.

Field audit

Sweep kernel or GP settings.

Trajectory audit

Bootstrap divergence, curl and path geometry.

Final claim

Morphometric trajectories are inferred, not observed lineages.

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: Section point clouds

Begin with this mechanism: Each slice contains coordinates and molecular features. Work through the concrete case: Thousands of embryo sections contribute millions of cells. Make the limitation visible: Sampling density and segmentation vary. Complete the evidence step: Audit coverage and units. Write the coordinates, graphs, kernels, images, matrices, losses, transformations, vector fields, attention weights, distances or tree edges with named dimensions and units. Calculate one small example by hand, then vary one assumption and explain which output moves and why. Reserve full sections, donors, stages, modalities, genes or landmarks outside tuning when testing transfer. Keep measurement, fitted representation, statistical evidence, biological interpretation and decision on separate notebook lines so a student, parent or reviewer can audit the mathematics.

Workshop 2: Partial overlap

Begin with this mechanism: Pairwise mapping permits unsupported tissue to remain weakly matched. Work through the concrete case: A growing organ exists only at the later stage. Make the limitation visible: Overlap parameters affect correspondence. Complete the evidence step: Plot matched mass. Write the coordinates, graphs, kernels, images, matrices, losses, transformations, vector fields, attention weights, distances or tree edges with named dimensions and units. Calculate one small example by hand, then vary one assumption and explain which output moves and why. Reserve full sections, donors, stages, modalities, genes or landmarks outside tuning when testing transfer. Keep measurement, fitted representation, statistical evidence, biological interpretation and decision on separate notebook lines so a student, parent or reviewer can audit the mathematics.

Workshop 3: Molecular dissimilarity

Begin with this mechanism: Gene profiles contribute a cross-cell cost. Work through the concrete case: Low KL-like dissimilarity supports candidate matches. Make the limitation visible: Expression convergence can mimic ancestry. Complete the evidence step: Hold out marker programs. Write the coordinates, graphs, kernels, images, matrices, losses, transformations, vector fields, attention weights, distances or tree edges with named dimensions and units. Calculate one small example by hand, then vary one assumption and explain which output moves and why. Reserve full sections, donors, stages, modalities, genes or landmarks outside tuning when testing transfer. Keep measurement, fitted representation, statistical evidence, biological interpretation and decision on separate notebook lines so a student, parent or reviewer can audit the mathematics.

Workshop 4: Spatial geometry

Begin with this mechanism: Relative coordinates constrain the assignment. Work through the concrete case: Neighbouring cells should not scatter arbitrarily. Make the limitation visible: Growth changes scale and topology. Complete the evidence step: Test geometry weights. Write the coordinates, graphs, kernels, images, matrices, losses, transformations, vector fields, attention weights, distances or tree edges with named dimensions and units. Calculate one small example by hand, then vary one assumption and explain which output moves and why. Reserve full sections, donors, stages, modalities, genes or landmarks outside tuning when testing transfer. Keep measurement, fitted representation, statistical evidence, biological interpretation and decision on separate notebook lines so a student, parent or reviewer can audit the mathematics.

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: Probabilistic assignment

Begin with this mechanism: A matrix distributes mapping weight between cells. Work through the concrete case: One earlier cell region maps to several later candidates. Make the limitation visible: Weights are not lineage probabilities. Complete the evidence step: Report entropy. Write the coordinates, graphs, kernels, images, matrices, losses, transformations, vector fields, attention weights, distances or tree edges with named dimensions and units. Calculate one small example by hand, then vary one assumption and explain which output moves and why. Reserve full sections, donors, stages, modalities, genes or landmarks outside tuning when testing transfer. Keep measurement, fitted representation, statistical evidence, biological interpretation and decision on separate notebook lines so a student, parent or reviewer can audit the mathematics.

Workshop 6: Rigid pose

Begin with this mechanism: Rotation and translation remove global orientation differences. Work through the concrete case: An embryo section is centred and rotated. Make the limitation visible: Rigid fit can be biased by partial overlap. Complete the evidence step: Use robust anchors. Write the coordinates, graphs, kernels, images, matrices, losses, transformations, vector fields, attention weights, distances or tree edges with named dimensions and units. Calculate one small example by hand, then vary one assumption and explain which output moves and why. Reserve full sections, donors, stages, modalities, genes or landmarks outside tuning when testing transfer. Keep measurement, fitted representation, statistical evidence, biological interpretation and decision on separate notebook lines so a student, parent or reviewer can audit the mathematics.

Workshop 7: Non-rigid residual

Begin with this mechanism: A continuous deformation explains local mismatch. Work through the concrete case: A limb bud bends while the trunk stays stable. Make the limitation visible: Flexible fields overfit noise. Complete the evidence step: Inspect Jacobians. Write the coordinates, graphs, kernels, images, matrices, losses, transformations, vector fields, attention weights, distances or tree edges with named dimensions and units. Calculate one small example by hand, then vary one assumption and explain which output moves and why. Reserve full sections, donors, stages, modalities, genes or landmarks outside tuning when testing transfer. Keep measurement, fitted representation, statistical evidence, biological interpretation and decision on separate notebook lines so a student, parent or reviewer can audit the mathematics.

Workshop 8: Multi-slice refinement

Begin with this mechanism: Neighbouring alignments are revisited across the stack. Work through the concrete case: A middle section balances evidence from both sides. Make the limitation visible: Errors can propagate through order. Complete the evidence step: Run leave-one-slice tests. Write the coordinates, graphs, kernels, images, matrices, losses, transformations, vector fields, attention weights, distances or tree edges with named dimensions and units. Calculate one small example by hand, then vary one assumption and explain which output moves and why. Reserve full sections, donors, stages, modalities, genes or landmarks outside tuning when testing transfer. Keep measurement, fitted representation, statistical evidence, biological interpretation and decision on separate notebook lines so a student, parent or reviewer can audit the mathematics.

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: Mesh correction

Begin with this mechanism: A surface mesh is adjusted to remove reconstruction artifacts. Work through the concrete case: A reconstructed organ boundary becomes continuous. Make the limitation visible: Smoothing can erase genuine ridges. Complete the evidence step: Compare raw and corrected meshes. Write the coordinates, graphs, kernels, images, matrices, losses, transformations, vector fields, attention weights, distances or tree edges with named dimensions and units. Calculate one small example by hand, then vary one assumption and explain which output moves and why. Reserve full sections, donors, stages, modalities, genes or landmarks outside tuning when testing transfer. Keep measurement, fitted representation, statistical evidence, biological interpretation and decision on separate notebook lines so a student, parent or reviewer can audit the mathematics.

Workshop 10: Cross-stage mapping

Begin with this mechanism: Cells or regions between developmental stages receive transport relationships. Work through the concrete case: E9.5 locations map toward E11.5 structures. Make the limitation visible: Different proliferation rates alter mass. Complete the evidence step: State growth assumptions. Write the coordinates, graphs, kernels, images, matrices, losses, transformations, vector fields, attention weights, distances or tree edges with named dimensions and units. Calculate one small example by hand, then vary one assumption and explain which output moves and why. Reserve full sections, donors, stages, modalities, genes or landmarks outside tuning when testing transfer. Keep measurement, fitted representation, statistical evidence, biological interpretation and decision on separate notebook lines so a student, parent or reviewer can audit the mathematics.

Workshop 11: Displacement samples

Begin with this mechanism: Mapped coordinates define observed direction vectors. Work through the concrete case: A target minus source position gives a vector. Make the limitation visible: Many-to-many maps need weighted averages. Complete the evidence step: Calculate confidence. Write the coordinates, graphs, kernels, images, matrices, losses, transformations, vector fields, attention weights, distances or tree edges with named dimensions and units. Calculate one small example by hand, then vary one assumption and explain which output moves and why. Reserve full sections, donors, stages, modalities, genes or landmarks outside tuning when testing transfer. Keep measurement, fitted representation, statistical evidence, biological interpretation and decision on separate notebook lines so a student, parent or reviewer can audit the mathematics.

Workshop 12: Kernel vector field

Begin with this mechanism: Nearby displacement samples influence one another. Work through the concrete case: A Gaussian kernel predicts motion at an empty location. Make the limitation visible: Bandwidth controls smoothness. Complete the evidence step: Publish a bandwidth curve. Write the coordinates, graphs, kernels, images, matrices, losses, transformations, vector fields, attention weights, distances or tree edges with named dimensions and units. Calculate one small example by hand, then vary one assumption and explain which output moves and why. Reserve full sections, donors, stages, modalities, genes or landmarks outside tuning when testing transfer. Keep measurement, fitted representation, statistical evidence, biological interpretation and decision on separate notebook lines so a student, parent or reviewer can audit the mathematics.

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: Field integration

Begin with this mechanism: An ODE traces positions through the predicted field. Work through the concrete case: A path follows dx/dt equals v(x). Make the limitation visible: Numerical steps and field errors accumulate. Complete the evidence step: Check step convergence. Write the coordinates, graphs, kernels, images, matrices, losses, transformations, vector fields, attention weights, distances or tree edges with named dimensions and units. Calculate one small example by hand, then vary one assumption and explain which output moves and why. Reserve full sections, donors, stages, modalities, genes or landmarks outside tuning when testing transfer. Keep measurement, fitted representation, statistical evidence, biological interpretation and decision on separate notebook lines so a student, parent or reviewer can audit the mathematics.

Workshop 14: Divergence and curl

Begin with this mechanism: Spatial derivatives summarize expansion and rotation. Work through the concrete case: Positive divergence marks an expanding region. Make the limitation visible: Derivatives amplify noise. Complete the evidence step: Bootstrap and smooth transparently. Write the coordinates, graphs, kernels, images, matrices, losses, transformations, vector fields, attention weights, distances or tree edges with named dimensions and units. Calculate one small example by hand, then vary one assumption and explain which output moves and why. Reserve full sections, donors, stages, modalities, genes or landmarks outside tuning when testing transfer. Keep measurement, fitted representation, statistical evidence, biological interpretation and decision on separate notebook lines so a student, parent or reviewer can audit the mathematics.

Workshop 15: Curvature and torsion

Begin with this mechanism: Trajectory derivatives describe bending and three-dimensional twisting. Work through the concrete case: A cardiac path curves and leaves a plane. Make the limitation visible: Short noisy paths produce unstable values. Complete the evidence step: Use arc-length and uncertainty filters. Write the coordinates, graphs, kernels, images, matrices, losses, transformations, vector fields, attention weights, distances or tree edges with named dimensions and units. Calculate one small example by hand, then vary one assumption and explain which output moves and why. Reserve full sections, donors, stages, modalities, genes or landmarks outside tuning when testing transfer. Keep measurement, fitted representation, statistical evidence, biological interpretation and decision on separate notebook lines so a student, parent or reviewer can audit the mathematics.

Workshop 16: Governance and learning

Begin with this mechanism: Version genes, overlap, mapping, kernels, meshes and solvers. Work through the concrete case: A student fits a two-dimensional kernel field. Make the limitation visible: A morphometric path is a model. Complete the evidence step: Write a vector-field ledger. Write the coordinates, graphs, kernels, images, matrices, losses, transformations, vector fields, attention weights, distances or tree edges with named dimensions and units. Calculate one small example by hand, then vary one assumption and explain which output moves and why. Reserve full sections, donors, stages, modalities, genes or landmarks outside tuning when testing transfer. Keep measurement, fitted representation, statistical evidence, biological interpretation and decision on separate notebook lines so a student, parent or reviewer can audit the mathematics.