# Interpreting Results

Use this guide to interpret EVd3x outputs conservatively and consistently.

## 1) Read order

1. Confirm context in Summary (single node or collection).
2. Check EV Evidence source support.
3. Review Pathway and Disease.
4. Use Cell, LR, and PPI for candidate follow-up.

## 2) Direct versus target-derived context

For miRNA-driven states, some pathway and disease context is target-derived through miRNA-mRNA links.

- Direct evidence: rows linked directly to the queried molecule.
- Target-derived evidence: rows linked through supported target mRNAs.
- Direct overlap: queried miRNA supports a queried mRNA.
- Shared target: multiple queried miRNAs converge on a non-query mRNA.

Report direct and target-derived signals separately in manuscripts and internal notes.

## 3) Summary metrics versus full tables

Summary cards are first-pass views optimized for speed.

For final interpretation:
1. open the module tab,
2. inspect full rows,
3. export machine-readable tables,
4. verify that first-pass patterns persist.

## 4) Score boundaries

- Pathway: compare p-value and q-value only within the same filter state.
- Disease: higher support means stronger association context, not causality.
- Cell/LR/PPI: scores rank hypotheses for follow-up, not functional proof.

## 5) Interactive guards

- First-pass views can return fewer rows than maximum caps.
- Top-N and other limits are processing guards for responsiveness.
- If needed, expand scope iteratively and document each expansion step.

## 6) Reporting template

Include:
- query and mode,
- main EV evidence signal,
- top pathway and disease context,
- interpretation boundary,
- proposed follow-up validation step.