Module 11: Synapses and Circuit Logic

Interpret synaptic organization and local circuit motifs from connectomics data with evidence-aware reasoning.

Stylized vector art: two apposed synaptic boutons with vesicles across a cleft.

Lesson Flow

Learn

Goals and Concepts

Start with the capability target and concept set for this module.

Practice

Studio Activity

Apply the ideas in a guided activity tied to realistic outputs.

Check

Assessment Rubric

Use the rubric to verify competency and identify improvement targets.

Interactive Lab

Practice in short loops: checkpoint quiz, microtask decision, and competency progress tracking.

Circuit Logic Checkpoint

Q1. Why is "reconstruction errors just add noise to my motif counts" a dangerous assumption?

Noise that averages out is survivable; a bias is not. Because motif counting is combinatorial, one merged interneuron manufactures many loops at once, so the error has a direction. That is why the checks are re-counting at a higher synapse threshold and on a later proofreading version, rather than simply widening an error bar.

Q2. E to I to E loops are enriched 1.7x over a degree-preserving null, but a cell-type-stratified null expects almost the observed count. What does that tell you?

Each null removes a different explanation. Preserving degree leaves cell-type composition free to generate the motif; preserving E/I rates removes that and the enrichment thins to almost nothing. The disagreement is itself a finding because it localises where the signal came from, so report both and let the stricter null carry the claim.

Q3. Which statement about synapse placement is supported by EM evidence alone?

Placement is measured; function is inferred. The first option asserts an effect on spiking, which needs physiology. The third leans on the spine-size correlation, which holds across populations and is not a strength readout for one connection. The reportable result is the compartment tally, with any functional reading marked as interpretation.

Motif Claim Microtask

Your motif is enriched 1.7x over a cell-type-stratified null. Which sentence belongs in the results?

Progress Tracker

State is saved locally in your browser for this module.

0% complete

Capability target

Generate one synapse-to-motif interpretation with explicit evidence chain and one alternative explanation.

Concept set

1) Synaptic organization as circuit logic

Synapses are not randomly placed. Their location on the postsynaptic neuron (soma, proximal dendrite, distal dendrite, spine, axon initial segment) determines their functional impact:

This compartment-specific targeting is a fundamental organizing principle of cortical circuits. In EM connectomics, you can directly observe where each synapse lands, making this a uniquely powerful approach for studying circuit logic.

2) Circuit motifs: recurring wiring patterns

Beyond individual synapses, the pattern of connections between neurons forms circuit motifs — small subgraph patterns that may implement computational primitives:

3) From observation to claim: the evidence chain

To claim that a motif is “enriched” or “functionally relevant,” you need:

  1. Detection: Identify the motif instances in the connectome graph.
  2. Quantification: Count occurrences.
  3. Comparison: Compare to a null model (degree-preserving random, spatially constrained, cell-type-stratified).
  4. Statistical test: z-score, p-value, multiple comparison correction.
  5. Biological interpretation: What computation could this motif implement?
  6. Alternative explanation: What non-functional explanation could produce the same enrichment? (e.g., spatial proximity, cell-type structure)

4) Annotation errors create false motifs

Segmentation and synapse detection errors can create or destroy motif instances:

Always ask: “Could this motif be an artifact of reconstruction errors?” Sensitivity analysis across proofreading versions helps: if a motif finding changes substantially between data versions, it may not be robust.

Misconception guardrails

Each of these is a belief a learner plausibly holds on arriving. Name it, then check your own work against it.

Worked example: the feedback-inhibition claim, walked to its honest size

The numbers below are illustrative — they show the shape of the reasoning, not results from a specific published dataset.

A student finds that feedback inhibition loops (pyramidal → interneuron → same pyramidal) look common in a 200-neuron L2/3 subgraph with 1,450 directed edges, and drafts the claim “this circuit is organized for gain control.” Here is how an expert walks that claim down to what the data supports.

Step 1 — Count against a null before describing function. Observed: 68 E→I→E feedback loops. Against 1,000 degree-preserving rewirings: null mean 41, sd 6, so z = 4.5 and enrichment 1.7x. Promising — but the degree-preserving null shuffles cell types, and E→I and I→E edges are common regardless of specific feedback wiring. A cell-type-stratified null preserving E/I connection rates expects 61 loops: enrichment 1.11x, z = 1.0. Most of the apparent enrichment was type composition; what survives is modest and needs the next two checks to mean anything.

Step 2 — Use the synapse-level evidence the graph threw away. This is where EM earns its keep. In 49 of the 68 loops the I→E synapses land perisomatically — on soma or proximal dendrite, the basket-cell placement positioned to control spiking. In 12 loops they land on distal dendrites, and in 7 the compartment is uncertain. The perisomatic subset is the one consistent with the gain-control story; the distal subset is a different circuit motif with different functional implications. A graph-only analysis would have averaged these together. The compartment cues and their reliability limits are Technical Unit 05.

Step 3 — Ask what reconstruction error does to this count, in this direction. Merges inflate dense motifs: a merged interneuron inherits two cells’ partner lists and manufactures loops. Check the interneurons carrying the most loops — the top one participates in 11. Its morphology passes inspection, but 19 of the 68 loops depend on at least one 1-synapse edge, the least reliable edge class. Re-run at a ≥2 synapse threshold: 51 loops survive, and the stratified-null comparison moves to 47 expected — the residual enrichment thins further. Re-count on the next proofreading materialization: 64 loops, with the composition roughly stable, so the count is not an artifact of one data version.

Step 4 — Write the claim and its boundary. Supported: “Perisomatic-targeting feedback loops are present and constitute the majority of E→I→E motifs in this subgraph; loop frequency is largely explained by cell-type connection rates, with at most a small residual above the stratified null.” Non-claim: “This does not show the circuit performs gain control — that requires functional data — and does not show loop-specific wiring selection, which the stratified null does not support.” Alternative explanation, stated in the report: spatial proximity was never controlled, and interneuron arbors overlap densely with their neighbors.

What the walk bought. The claim shrank from “organized for gain control” to a compartment-resolved anatomical statement with a quantified null comparison and a stated confound. The smaller claim is publishable and durable; the original was neither.

Core workflow

  1. Identify synapse candidates: find synapses in the region of interest with correct pre/post assignment.
  2. Build local connectivity motif: extract the subgraph connecting the pre and post neurons and their immediate neighbors.
  3. Classify the motif: reciprocal pair, feed-forward loop, feedback inhibition, convergent input, etc.
  4. Evaluate against null: is this motif more common than expected?
  5. State supported claim (what the data shows) + caveat (what it doesn’t prove and what could confound it).

60-minute tutorial run-of-show

Pre-class preparation (10 min async)

Minute-by-minute plan

  1. **00:00-10:00 Synapse cue recap**
    • Quick review: asymmetric (Type I, excitatory) vs symmetric (Type II, inhibitory) synapses.
    • Show 3 synapses in EM: spine synapse, perisomatic synapse, AIS synapse. “Where the synapse lands tells you about circuit function.”
  2. **10:00-24:00 Motif construction examples**
    • Walk through 3 motifs in the MICrONS dataset:
      • Reciprocal pair between two L2/3 pyramidal cells (mutual excitation)
      • Feed-forward loop: L4 stellate → L2/3 pyramidal → L5 pyramidal, with L4 also connecting directly to L5
      • Feedback inhibition: pyramidal → basket cell → same pyramidal
    • For each: show the EM evidence (synapses), draw the circuit diagram, discuss functional implication.
  3. **24:00-38:00 Learner motif analysis**
    • Learners receive a small subgraph (15 neurons, 50 synapses) and identify all 3-node motifs.
    • Count each motif type. Which are most common?
    • Compare to expectations: “If these were randomly connected with the same degree distribution, how many of each motif would you expect?”
  4. **38:00-50:00 Alternative explanation challenge**
    • For each enriched motif, learners must propose one alternative (non-functional) explanation:
      • “Reciprocal connections are enriched because nearby neurons are more likely to connect” (spatial proximity)
      • “Feed-forward loops are enriched because of cell-type structure” (E→I and I→E are common)
    • Group discussion: how would you test whether the spatial explanation is sufficient?
  5. **50:00-60:00 Competency check**
    • Each learner writes a motif claim/caveat pair:
      • “In this circuit, [motif] is enriched [X]× compared to [null model]. This is consistent with [functional interpretation]. However, [alternative explanation] could also account for this enrichment.”
    • Exit ticket: “One motif claim and one plausible confound.”

Studio activity: motif discovery and interpretation (60-75 minutes)

Scenario: You are analyzing a 200-neuron subgraph from the MICrONS dataset, spanning L2/3 and L4 of mouse visual cortex. Your goal: characterize the local circuit motif profile and identify any enriched patterns that suggest specific wiring rules.

Task sequence:

  1. Enumerate all 2-node and 3-node motifs in the subgraph (use DotMotif or equivalent tool).
  2. Generate 1,000 degree-preserving random rewirings. Count motifs in each.
  3. Compute z-scores for each motif type.
  4. Identify the top 3 most enriched motifs. For each: draw the circuit diagram, propose a functional interpretation, and state one alternative explanation.
  5. Write a 1-page “circuit logic brief” summarizing the motif profile of this circuit.

Expected outputs:

Assessment rubric

Common errors and how to recover

What this module does not cover

Content library references

Teaching resources

References

Quick practice prompt

Write one motif claim and one plausible confound.

Teaching Materials

Activity Worksheet

Learner worksheet aligned to the studio activity and rubric.

Open worksheet

Slide Source

Marp source file for editing and rendering.

course/decks/marp/modules/module11.marp.md

Related Content