> For the complete documentation index, see [llms.txt](https://docs.neurolabai.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.neurolabai.io/start-using-neurolab/report-interpretation.md).

# How to Read a Neurolab Report

Read a Neurolab report as a set of typed research claims, not as one recommendation.

## A reliable reading order

1. **Confirm the input and route.** Check the resolved compound or submitted SMILES and whether the report is known, structure-derived, or receptor-conditioned.
2. **Read reliability first.** Before focusing on rankings, inspect display modes, applicability, warnings, novelty, and abstention.
3. **Classify each field.** Decide whether it is a model-ranked prediction, source-backed record, composed hypothesis, reference context, or caller-provided assumption.
4. **Inspect exact support.** A section-level evidence bundle does not necessarily support every item. Follow report pointers and support links.
5. **Keep retrieved literature separate.** Section Inspector can retrieve external context, but report-returned evidence and external literature are not one evidence tier.
6. **Read nulls and caveats.** Missing or withheld output is meaningful; do not silently substitute a neighboring value.
7. **Check brain-channel labels.** Predicted, density, circuit, literature, measured, and legacy channels have different meanings.
8. **Record release identity.** Keep model, facts-and-evidence, KG, report/schema, route, and generation metadata with the result.
9. **Return to primary sources.** Neurolab can organize evidence, but it cannot replace expert source review.

## Predictions and records can coexist

A known compound report can contain both model rankings and retrieved records. A model-ranked reported-usage label and an FDA-derived record may resemble one another, but they answer different questions:

* the ranking says what the model placed near the top under its learned objective;
* the record says what a named source asserts under that source’s scope.

A citation attached elsewhere in the report does not automatically validate the ranking.

## Rankings are not universal confidence

Top-K output answers “what did this head rank highest?” It does not, without a separate calibration contract, answer “what is the probability that this statement is true?”

Do not assume that:

* the first item is clinically dominant;
* rank implies effect size;
* scores from cognition and side effects share a scale;
* a high score overcomes an applicability warning;
* absence from the top five means biological absence; or
* ensemble agreement proves correctness.

## Structure-derived results need extra context

For a novel SMILES, chemical neighbors provide analog context. Their records belong to those neighbors, not automatically to the query structure. Preserve similarity, applicability, receptor-source, retrieval, and abstention metadata with every downstream interpretation.

If the user supplied receptor rows, every downstream conclusion is conditional on those assumptions. Changing the rows can change the report.

## Usage and safety sections

“Reported usage” is a ranking task, not a statement of approved indication or clinical efficacy. Side-effect output is sparse, route-specific, and non-exhaustive; it is not a safety screen. Source-derived label records should remain visibly distinct from model output.

Dose-related records are descriptive context. An entered milligram value does not resolve formulation, timing, pharmacokinetics, brain exposure, receptor occupancy, or individual response and must never be turned into instructions.

## Brain sections

Brain-related rows and maps are exploratory hypotheses unless a separately loaded and validated channel says otherwise. Color can encode rank, sign, density, or reference anatomy depending on the active channel. It must not be relabeled as measured activation, tract engagement, or causal propagation.

Readable parcel labels help locate the 360 Glasser cortical and 32 Tian S2 subcortical outputs. They are metadata, not a new model result or an accuracy improvement.

## See the rules applied

* [Donepezil: investigate a known compound](/examples-and-use-cases/known-compound-investigation.md) follows one report through the coordinated investigation views.
* [Donepezil: trace a report claim](/examples-and-use-cases/trace-a-report-claim.md) separates a model rank from binding, FDA-derived, ATC, and graph context.
* [Ketamine: explore a multi-system report](/examples-and-use-cases/ketamine-multisystem-investigation.md) shows why evidence quantity and evidence quality are different questions.
* [Modafinil: investigate uncertainty](/examples-and-use-cases/investigate-uncertainty.md) shows how low per-head agreement changes the language used for a result.

## A compact interpretation template

When recording a finding, use this form:

> On **\[route]** in **\[model/KG release]**, Neurolab returned **\[field and rank]** as a **\[evidence state]**. The attached support is **\[field-specific/section-level/contextual/unavailable]**. Reliability or applicability indicates **\[state]**. This supports **\[narrow research interpretation]**, but does not establish **\[key limitation]**.

When results are saved in a [Research Workbench](/organize-and-compare-research/research-workbenches.md) or compared in a [Molecular Portfolio](/organize-and-compare-research/molecular-portfolios.md), keep this interpretation record attached to the exact report version.

***

Next: [Evidence States](/understand-results/evidence-states.md)


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:

```
GET https://docs.neurolabai.io/start-using-neurolab/report-interpretation.md?ask=<question>&goal=<endgoal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.