Unit: Connectome Analysis and NeuroAI

Learning goals

• Explain why connectome analysis is a core bridge between neuroscience and AI.
• Describe the motif-search workflow from hypothesis to query to interpretation.
• Compare practical tool choices for motif queries and subgraph isomorphism.
• Evaluate how analysis scale and compute constraints shape scientific conclusions.

Scope boundaries

• In scope:
  • graph/motif-centric connectome analysis
  • NeuroAI motivation and analysis pipeline
  • practical tooling examples (DotMotif/GrandIso-style workflows)
• Out of scope:
  • full cell segmentation methods (covered in Unit 08)
  • wet-lab imaging/acquisition detail (Units 02/03)

Source synthesis

• Core claims to preserve:
  • biological networks motivate new AI priors and architectures
  • motif search is a practical lens for candidate computational primitives
  • scalable query tooling is required for realistic connectome-scale analysis
• Historical context to reframe:
  • neuroAI deck framing from July 12, 2021 should be labeled historical baseline
  • outreach examples used as context, not as current-state authority
• Needs verification:
  • benchmark claims and speed comparisons before publication-ready copy
  • current status of specific repositories/tools

Website draft blocks

Hero framing

Connectome analysis asks a direct question: can biological circuit structure reveal reusable computational motifs that improve AI? This unit introduces the analysis workflow that links graph theory, query languages, and large-scale neural datasets.

Core concepts

• From connectome graph to candidate motif hypotheses.
• Subgraph search and isomorphism as scientific instruments.
• Why data scale, storage models, and query overhead matter.

Practical workflow

1. Define motif hypothesis from neuroscience literature.
2. Express query in a readable motif language.
3. Run motif search over connectome graph(s).
4. Compare prevalence to null models / control datasets.
5. Interpret motifs as candidate functional primitives.

Tools and methods

• DotMotif-style motif specification.
• Cypher-translation and graph backends.
• GrandIso-style accelerated subgraph search.

Output and impact

• Candidate circuit primitives for hypothesis generation.
• Structured comparisons across regions/species/development.
• NeuroAI design cues for architecture constraints.

Slide draft sequence (v1)

1. Why connectome analysis for NeuroAI? (motivation)
2. Where current ML falls short (problem framing)
3. What brain data looks like at analysis level
4. Reverse-engineering analogy and limits
5. NeuroAI pipeline overview
6. Motif hypothesis and subgraph framing
7. Query language design (human-readable -> executable)
8. GrandIso/subgraph isomorphism mechanics
9. Performance and scale tradeoffs
10. Atlas-style motif scan examples
11. Developmental/connectome comparison example
12. Interpretation limits and verification checklist
13. Current applications and integration points
14. Summary + reading/resources

Figure candidates

See: course/units/figures/09-connectome-analysis-neuroai-selected-v1.md

Cross-links

• Related modules: module10, module13, module14, module15, module20
• Related tools: /tools/ask-an-expert/, /tools/connectome-quality/
• Related frameworks: /models/, /education/models/

Open issues

• Confirm citation/attribution text for selected figures before web integration.
• Decide how many historical figures to keep vs replace with newer references.
• Add one hands-on notebook activity aligned to motif query practice.

Scope check (expert pass)

• Distinguish motif prevalence from causal mechanism; include null-model comparisons.
• Require statistical controls for multiple comparisons and dataset bias.
• Frame NeuroAI transfer as constrained inspiration, not direct biological equivalence.

Technical anchors to preserve

• Query language usability is part of scientific throughput.
• Subgraph isomorphism complexity drives practical design tradeoffs.
• Cross-dataset comparability requires standardized preprocessing and annotation assumptions.