Biology · Biological Frontiers & Open Questions
Single-cell and spatial biology across molecular and tissue scales
A tissue average gives marker signals of two and two. That average could come from four similar cells or from different cells whose signals balance. Even when the individual cell records are known, the cells could sit in different neighbourhoods. New molecular maps are useful partly because they let these different questions be asked separately.
Earth, Origins & the History of Life introduces the difference between biological history and processes possible today. Development, Differentiation & Stem Cells introduces cell states and organismal organisation. Genomics, Omics & Bioinformatics distinguishes molecular records, identity, proportions and validation. Systems & Quantitative Biology supplies state, observation and model-comparison reasoning. Each lesson also explains its local terminology; the original examples require only fractions, means and signed differences. Follow the concrete cases before returning to the broader research question.
Back to the Biology mapBy the end, you can
Distinguish a cell’s molecular state, its position and its ancestry, and identify the evidence each question requires.
A molecular profile describes a recorded state
A single-cell profile records selected molecular information associated with an included cell. A transcript profile concerns RNA represented in that observation; it is not a complete inventory of proteins, reaction rates or everything the cell can do. The record can help distinguish states and propose cell identities, but interpretation also needs the observation conditions and the biological question.
A cell type usually describes an identity that is useful across a specified context. A cell state describes features that can vary with activity, environment, development or the observation time. These terms need an explicit definition in a particular study. A computational group of similar profiles is a result of the chosen representation and grouping rule; it does not automatically establish a permanent natural boundary.
An atlas connects records across cells and, where supported, with location, morphology and other evidence. Its value depends on what was included and how the measurements relate. Cells omitted by collection or selection do not become known absent. A useful map can leave rare states, other tissues or other circumstances incompletely represented.
Composition does not determine a neighbourhood
In the four-cell example, cells 1 and 2 each have signals X = 0 and Y = 4. Cells 3 and 4 each have X = 4 and Y = 0. Both marker totals are eight, so both whole-inventory means are two. The markers are anonymous teaching signals; they do not identify a real gene, molecule count or cell type.
One layout puts cells 1 and 2 in the first row and 3 and 4 in the second. A second mixes the marker groups in each row. Neighbours in this grid are cells one horizontal or vertical grid step apart; diagonal pairs are not counted as neighbours. The entire molecular inventory is unchanged, while the regional means and neighbours differ. Spatial information therefore adds something that the whole-inventory average cannot recover.
A spatial measurement also has its own resolution and observation rule. A region can include more than one cell, and an inferred assignment can differ from directly observed position. Two nearby signals can suggest a question about interaction, but proximity alone does not demonstrate signalling, material exchange or influence. Those claims need appropriate additional evidence.
A shared tissue average can conceal different neighbours
Read the diagram in words
Both arrangements have X total 8 and Y total 8, giving means 2 and 2 over four cells. In the separated arrangement the two rows have means (0, 4) and (4, 0). In the mixed arrangement each row instead has (2, 2). The complete cell inventory is the same. Neighbours in this grid are cells one horizontal or vertical grid step apart; diagonal pairs are not counted as neighbours. The separated layout has adjacent pairs 1–2, 1–3, 2–4 and 3–4. The mixed layout has adjacent pairs 1–3, 1–4, 2–3 and 2–4.
| Same supplied cell identity | Assigned marker X signal units | Assigned marker Y signal units |
|---|---|---|
| Supplied cell 1 | 0 | 4 |
| Supplied cell 2 | 0 | 4 |
| Supplied cell 3 | 4 | 0 |
| Supplied cell 4 | 4 | 0 |
| Same supplied cell identity | x | y |
|---|---|---|
| Supplied cell 1 | 0 | 0 |
| Supplied cell 2 | 1 | 0 |
| Supplied cell 3 | 0 | 1 |
| Supplied cell 4 | 1 | 1 |
| Included grid row | Same supplied cell identity | X | Y |
|---|---|---|---|
| First grid row | Supplied cell 1 · Supplied cell 2 | 0 | 4 |
| Second grid row | Supplied cell 3 · Supplied cell 4 | 4 | 0 |
| Same supplied cell identity | Same supplied cell identity |
|---|---|
| Supplied cell 1 | Supplied cell 2 |
| Supplied cell 1 | Supplied cell 3 |
| Supplied cell 2 | Supplied cell 4 |
| Supplied cell 3 | Supplied cell 4 |
| Same supplied cell identity | x | y |
|---|---|---|
| Supplied cell 1 | 0 | 0 |
| Supplied cell 3 | 1 | 0 |
| Supplied cell 2 | 1 | 1 |
| Supplied cell 4 | 0 | 1 |
| Included grid row | Same supplied cell identity | X | Y |
|---|---|---|---|
| First grid row | Supplied cell 1 · Supplied cell 3 | 2 | 2 |
| Second grid row | Supplied cell 2 · Supplied cell 4 | 2 | 2 |
| Same supplied cell identity | Same supplied cell identity |
|---|---|
| Supplied cell 1 | Supplied cell 3 |
| Supplied cell 1 | Supplied cell 4 |
| Supplied cell 3 | Supplied cell 2 |
| Supplied cell 2 | Supplied cell 4 |
| n | ΣX | ΣY | X | Y |
|---|---|---|---|---|
| 4 | 8 | 8 | 2 | 2 |
Original four-cell, two-marker inventory in a two-by-two coordinate grid. X and Y are arbitrary compatible marker-signal units, not named genes or molecular counts. Each layout contains each cell exactly once. Regions are the two grid rows, each containing two cells; region averages are equal-weight cell averages. Coordinates supply adjacency only, with no metric length or tissue identity. Neighbours in this grid are cells one horizontal or vertical grid step apart; diagonal pairs are not counted as neighbours. Alternative arrangements are not two times in one tracked tissue. No state classification algorithm, signalling, fate, interaction or causal inference is provided.
Similarity is different from descent
A lineage concerns who descended from whom. A current molecular profile concerns a state at observation. Cells with different immediate ancestors can reach similar states, while siblings can diverge. Combining state and ancestry information can reveal relationships that neither record establishes alone.
The supplied tree has two intermediate parents and four terminal cells. Each parent has one descendant in current state A and one in B. Cells 1 and 3 share A across different parents. No earlier ancestor state was observed. Inferring that a parent was already in A simply because a descendant is in A would fill a missing record without evidence.
Ordering snapshot profiles into a proposed trajectory can be useful for a developmental hypothesis. It is still different from following the same cell through time or independently establishing its ancestry. A first research question is which parts of the proposed ordering are supported by measured time, lineage or another discriminating record.
Similar current states need not mean the same ancestry
Read the diagram in words
Cells 1 and 2 descend from parent 1, while cells 3 and 4 descend from parent 2. Cells 1 and 3 share current state A despite different immediate parents; 2 and 4 share B. Each sibling pair contains one A and one B. All three earlier ancestor states are unknown.
| Supplied ancestry identity | Current assigned molecular state |
|---|---|
| Shared earlier ancestor | ? |
| Intermediate ancestor 1 | ? |
| Intermediate ancestor 2 | ? |
| Terminal cell 1 | Assigned current state A |
| Terminal cell 2 | Assigned current state B |
| Terminal cell 3 | Assigned current state A |
| Terminal cell 4 | Assigned current state B |
| Supplied ancestry identity | Supplied ancestry identity | Cell-descent relationship |
|---|---|---|
| Shared earlier ancestor | Intermediate ancestor 1 | Supplied parent-to-descendant relationship |
| Shared earlier ancestor | Intermediate ancestor 2 | Supplied parent-to-descendant relationship |
| Intermediate ancestor 1 | Terminal cell 1 | Supplied parent-to-descendant relationship |
| Intermediate ancestor 1 | Terminal cell 2 | Supplied parent-to-descendant relationship |
| Intermediate ancestor 2 | Terminal cell 3 | Supplied parent-to-descendant relationship |
| Intermediate ancestor 2 | Terminal cell 4 | Supplied parent-to-descendant relationship |
| Supplied ancestry identity | Supplied ancestry identity |
|---|---|
| Terminal cell 1 | Terminal cell 3 |
| Terminal cell 2 | Terminal cell 4 |
| Supplied ancestry identity | Supplied ancestry identity |
|---|---|
| Terminal cell 1 | Terminal cell 2 |
| Terminal cell 3 | Terminal cell 4 |
Earlier ancestor state is unobserved
Original stipulated binary cell-descent tree with one root, two intermediate ancestors and four terminal cells. Tree layout is topology only; branch lengths are not elapsed time. Current states are supplied labels, not inferred cell types or measured developmental times. Ancestor states are unobserved. Neither transcript similarity nor ancestry by itself supplies a future fate.
Worked example
Compare the records before naming the cells
Add the X values 0 + 0 + 4 + 4 = 8, and divide by four to get two. The Y sum is also eight. Both layouts therefore share the same bulk marker means.
In the separated layout, the first row has mean (0, 4) and the second (4, 0). In the mixed layout, each row has (2, 2). Equal bulk values do not choose a layout.
In the ancestry tree, cells 1 and 3 share a state but have different immediate parents. Cells 1 and 2 share a parent but have different states. Their state and lineage records are complementary rather than interchangeable.
Check your understanding
Can a mean X signal of two identify the X value in every included cell?
Show reasoning
No. The complete example has values 0, 0, 4 and 4 with mean two. An average constrains a total per included unit; it does not specify each individual value or the distribution.
What changes when the four cell records are rearranged spatially?
Show reasoning
The cell identities, marker totals and bulk means stay the same. Row composition, regional means and neighbours change. This isolates the additional information supplied by location.
Does a similar current profile prove that two cells had the same immediate parent?
Show reasoning
No. Cells 1 and 3 share state A but have different supplied parents. A lineage claim requires ancestry evidence rather than being inferred solely from state similarity.
What to carry forward
Cell identity, present state, spatial arrangement and ancestry are related questions with different evidence needs.
References
Resources supporting this lesson’s explanations and examples.
- The Human Cell Atlas (eLife6:e27041,2017; doi:10.7554/eLife.27041)Aviv Regev, Sarah A. Teichmann and colleagues, with Human Cell Atlas Meeting ParticipantsResearch abstract
- Lineage tracing meets single-cell omics: opportunities and challenges (Nature Reviews Genetics21:410–427,2020; doi:10.1038/s41576-020-0223-2)Daniel E. Wagner and Allon M. KleinResearch abstract