aDNA for Researchers

Your citations are only as good as your context graph. aDNA gives research labs, computational teams, and independent researchers a structured, agent-navigable knowledge layer — so every session starts oriented, every protocol is traceable, and knowledge accumulated over months doesn't evaporate when the context window closes.

The research knowledge problem

Research projects accumulate knowledge faster than any one person — or agent — can track. Protocols drift. Analysis decisions get made in a Slack thread and forgotten. A new collaborator spends a week reconstructing context from scattered READMEs. An agent session produces a finding the next session contradicts, because neither had the full picture.

Research knowledge is relational: a finding only makes sense relative to its protocol, its dataset version, its assumptions. Without structure that preserves those relationships, knowledge isn't reproducible — it's anecdotal.

How aDNA helps

aDNA structures research knowledge as a typed, navigable graph — not a flat document pile. Three properties matter most:

• Knowledge provenance. Every context file, decision record, and session log carries frontmatter — who created it, when, what it links to, what standard it targets. Agents update provenance automatically. The audit trail is structural.
• FAIR metadata, built in. Every lattice object includes a fair block: license, creators, keywords, identifier, provenance. Publishing to the registry is one command — FAIR compliance is not a separate checklist.
• Multi-agent context fidelity. The convergence model loads exactly the context a task needs — the active dataset, the active protocol, the prior findings — not the entire project. Parallel agents work without context bleed.

Typical ontology extensions

Research teams extend the base 14 entity types with domain-specific ones. Common extensions:

• experiment — a bounded empirical test with hypothesis, method, and outcome fields
• protocol — a versioned, reusable procedure with preconditions and reproducibility notes
• finding — a validated result linked to the experiment and dataset that produced it
• dataset — a data object with FAIR metadata, lineage, and access notes

The Extend the Ontology tutorial walks through adding any of these in 25 minutes.

Research reading path

Four tutorials that build a complete research knowledge architecture — from curated context files to executable workflow graphs.

Reference and depth