Skip to main content
iT
Guide

Guide & Use Cases

Everything you can do with InTree — input patterns, search modes, filters, and real-world use cases. Click any example to try it live.

Three search modes

Explore

1 entity

Pick a lens (everything, treatments, causes, biomarkers, drug targets…) and one entity. Get an interactive knowledge graph of its relationships, with novel connections and drug mechanisms.

Verify

2 entities

Get every piece of evidence classified as SUPPORTS, CONTRADICTS, or INDIRECT. See study quality breakdown and paper metadata.

Trace Pathway

3+ steps

Build an ordered chain of three or more entities. Verify the evidence for each hop, set how each step connects, and spot the weakest link.

What you can type

Natural language questions

How each mode's input works

ExploreA guided phrase: “[ Everything ▾ ] of [ entity ]”. Pick a lens (or leave it on Everything), enter one entity, and press Explore → relationship graph.
VerifyA structured triple: fill the Subject and Object slots, then optionally pick the relationship between them (the ⇄ button flips direction) → labeled evidence.
Trace PathwayA numbered step builder: add three or more ordered steps and set each hop's relationship (or leave it on auto-detect). Use “+ add step” to extend the chain → evidence per hop.
Plain EnglishPrefer to type? Questions like “does aspirin prevent stroke?” are parsed automatically and fill the right fields for the active mode.

Relationship types (optional)

TREATSINHIBITSACTIVATESBIOMARKER OFCAUSESASSOCIATED WITHTARGETSREGULATESPREVENTSBINDS

These power the lens in Explore, the relationship selector in Verify, and each hop in Trace. Leave them on the default to include every relationship type.

What you get in results

Evidence Page

  • Each sentence labeled SUPPORTS, CONTRADICTS, or INDIRECT
  • • Paper title, year, DOI link, journal
  • • Study type (RCT, Review, Clinical, Preclinical)
  • • Citation count per paper
  • • Verdict: Strong support / Contested / Contradicted
  • • Quality score out of 100
  • • Real paper count from 240M+ index

Filters & Lens

  • All — show everything
  • Contested — only supporting + contradicting
  • Recent (2023+) — papers from 2023 onwards
  • High impact — papers with 10+ citations
  • • Plus: filter by direction (Supports / Contradicts / Indirect)

Stats & Actions

  • • Evidence distribution donut chart
  • • Study types breakdown
  • • Publications per year trend
  • • Total papers found (real count)
  • • Average citations
  • Export to CSV
  • Share link
  • Save to workspace

Real use cases — click any to try

Who uses InTree

Researchers
Validate hypotheses with classified evidence before running experiments
Clinicians
Quickly check drug-disease evidence with study quality breakdown
Drug Discovery
Map all targets, biomarkers, and mechanisms for any drug or disease
Students
Learn relationship biology with real paper citations and DOI links
Grant Writers
Find and export supporting evidence for research proposals
Reviewers
Get structured evidence overview filterable by study type and year
Pharmacologists
Check drug interactions, side effects, and mechanism pathways
Bioinformaticians
Explore gene/protein networks and validate signaling pathways

How it works

1

Pick a mode and fill it in

Choose Explore, Verify, or Trace Pathway — each gives you a purpose-built input so you always know what to enter. Prefer typing? Plain-English questions are parsed and fill the right fields for you.

2

We search 926M+ sentences

InTree searches using both BM25 keyword matching and BGE 1024-dim semantic vectors across 170M+ papers. Results are re-ranked with a cross-encoder (bge-reranker-v2-m3) for precision.

3

AI classifies each sentence

Each evidence sentence is classified by Claude Haiku into SUPPORTS, CONTRADICTS, INDIRECT, or NOT_RELEVANT with HIGH/LOW confidence. The prompt distinguishes treatment vs prevention, biomarker vs therapeutic target.

4

Enriched with paper metadata

Study type (RCT, review, clinical), citation count, journal, year, and DOI are joined from Semantic Scholar. Study quality breakdown lets you weigh RCT evidence higher than observational.

How InTree is different

Feature
InTree
PubMed / Consensus
Evidence granularity
Sentence-level, labeled
Paper-level / abstract
Direction labels
SUPPORTS / CONTRADICTS / INDIRECT
None / yes-no meter
Study quality
RCT vs review vs clinical breakdown
Not distinguished
Relationship graph
Interactive, auto-computed
Not available
Pathway verification
Per-hop evidence chain
Manual reading
Coverage
240M+ papers, 926M+ sentences
Abstracts only
Cost per query
~$0.004 (under 1 cent)
Free / subscription
Speed
5-10 seconds with analysis
Instant but no analysis
240M+
Papers
PubMed + Semantic Scholar
926M+
Sentences
Searchable with semantic + keyword
292K
Entities
Genes, drugs, diseases, proteins