How the Paper Collection Is Built

This page explains how the 2,000-paper curated connectomics literature corpus was retrieved, screened, stratified across 12 canonical research domains, structured into 3 nested tiers ($500 \subset 1,000 \subset 2,000$), and annotated with OCAR cards and 3-level pedagogical summaries.

🏛️ Corpus Architecture & Nested Tiers

The collection is structured into three strictly nested materializations:

Tier	Corpus Size	Primary Target Audience & Role	Metadata Depth
500 Key Papers	500 papers	Curriculum flagships, course reading lists, seminar deep-dives	Full abstract, complete author list, verified 5-part OCAR research cards, 3-level summaries, seminar discussion prompts
1000 Key Papers	1,000 papers	Comprehensive scholarly survey, methods reference, subfield tracking	Full abstract, complete OCAR cards, citation metrics (in/out degree, k-core), domain classifications, organism tags
2000 Key Papers	2,000 papers	Global bibliometric network, citation lineage modeling, AI synthesis	Complete directed citation graph ($5,460+$ internal links), complete OCAR cards, full abstract and author/venue metadata, facet views

🔬 Multi-Channel Retrieval & Strict Scope Screening

The candidate pool was compiled using positive nanoscale/synaptic-resolution inclusion gates across Semantic Scholar, OpenAlex, Europe PMC, and PubMed:

Direct Synaptic Connectomics: Dense EM wiring diagrams, synaptic resolution imaging, automated segmentation pipelines (FFN, U-Net, affinity prediction, flood-filling).
First-Class Scientific Axes: Covers tissue preparation, FIB-SEM/SBEM acquisition, synapse detection, proofreading tools (CAVE, CATMAID, FlyWire), graph analysis (motifs, modularity, network topology), structure-function modeling, NeuroAI, cell census, health-translation, and training/outreach.
Positive Nanoscale Boundary: Macroscale non-synaptic methods (such as standard low-resolution fMRI or whole-brain fiber tractography without synaptic validation) are filtered out, preserving a clean nanoscale focus.

📊 Stratified 12-Domain Literature Taxonomy

Candidate papers are classified into 12 mutually exclusive primary domains using a strict decision-order hierarchy:

circuit-structure (15.0% target share / 75 in Top 500 / 300 in Top 2,000)
pipeline (15.0% target share / 75 in Top 500 / 300 in Top 2,000)
physiology (12.0% target share / 60 in Top 500 / 240 in Top 2,000)
behaviour (12.0% target share / 60 in Top 500 / 240 in Top 2,000)
imaging (8.0% target share / 40 in Top 500 / 160 in Top 2,000)
cell-types (8.0% target share / 40 in Top 500 / 160 in Top 2,000)
neuroanatomy (8.0% target share / 40 in Top 500 / 160 in Top 2,000)
synthesis (5.0% target share / 25 in Top 500 / 100 in Top 2,000)
dataset (5.0% target share / 25 in Top 500 / 100 in Top 2,000)
neuroai (5.0% target share / 25 in Top 500 / 100 in Top 2,000)
health (5.0% target share / 25 in Top 500 / 100 in Top 2,000)
training-outreach (2.0% target share / 10 in Top 500 / 40 in Top 2,000)

🃏 What Each Record Carries

Bibliographic Identity: Title, complete author list as published (authors, names separated by ; ), publication year, full venue/journal name, clean lowercase DOI, and a plain-text citation in the form First-author surname et al. (year). Title. Journal. DOI URL (two authors are written Surname & Surname). These fields are re-derived from the corpus record with the same DOI by scripts/derive_journal_papers.py; the validator scripts/validate_paper_counts.rb fails if any record ships without authors, with a placeholder citation, or with a year that disagrees with the corpus.
Abstract: The published abstract (abstract), present on all 2,000 records. Most are single paragraphs; structured or multi-paragraph abstracts keep their paragraph breaks.
Era: inclusion_role is assigned from the publication year – history (up to 2018), contemporary (2019–2023), sota (2024 onward) – and drives the era filter on the journal club page.
OCAR Structure:
- Opportunity: Scientific/technological opening addressed.
- Challenge: Key bottlenecks, scale limits, or biological ambiguities.
- Action: Experimental, imaging, computational, or theoretical methodology executed.
- Resolution: Findings, benchmarks, connectome maps, or models delivered.
- Future Work: Open problems and next-generation research horizons.
Three-Level Pedagogical Summaries: Beginner (no prerequisites), Intermediate (foundational knowledge), Advanced (active researcher).
Discussion Prompts: Facilitator questions for journal clubs and research seminars.
Graph Topological Placement: In-degree (inbound citations within corpus), Out-degree (references within corpus), k-core centrality.
Directed Citation Links: List of DOIs cited by the publication.

🧭 Exploring the Collection

Interactive Citation Graph Explorer: Live WebGL/Canvas network graph with organic force-directed physics, weighted directed citation arrows, and AI research synthesis prompt generator.
Journal Club Card Browser: Filterable cards by tier, domain, expertise level, and publication era.
Literature Corpus Taxonomy Hub: Domain breakdown with direct links to standalone JSON downloads.