Reported isn’t inferred.
A protein assay is preserved as an assay. A proposed genetic interpretation stays a separate, reviewable claim.
A CLEARER LENS ON CANCER RESEARCH
Every insight starts
with evidence.
Extraordinary responses deserve a closer look. OncoRift is building a way to connect scattered cancer research to the original evidence—and the people who can make sense of it.
+A shared foundation. Disease-specific research.
22 concept tracks · GBM and PDAC remain core priorities
22 concept tracks · GBM and PDAC remain core priorities
CONNECT THE EVIDENCE.
PRESERVE THE CONTEXT.
FIG. 01 / THE EVIDENCE NETWORKPRESERVE THE CONTEXT.
Built for scientific scrutinyExact source passages ↗Versioned interpretations ↗Human review at the core ↗
01 / THE APPROACH
Follow the evidence.
Keep the nuance.
Research is more than a collection of conclusions. It’s the wording of an observation, the timing of a treatment, the uncertainty in a diagnosis.
OncoRift is designed to keep those details visible. Original observations stay separate from interpretations, so every proposed connection can be inspected, challenged, and revisited.
01 Original observations02 Versioned interpretations03 Pinned knowledge
RESEARCH TRACKS / SCOPE BEFORE SCALE
More cancers.
The same evidence standard.
22 concept tracks cover the 20 leading global cancer mortality categories, plus a dedicated GBM track and melanoma. GBM and PDAC remain our initial pilot priorities.
Selection basis
Global number of cancer deaths, both sexes, all ages. IARC · GLOBOCAN 2024, version 1.0 ↗ · Accessed October 19, 2026.
Mortality categories are broader than some research scopes. PDAC is a pancreatic subset; GBM sits within brain/CNS. This is not a ranking of individual prognosis.
All tracks are concepts. Only GBM and PDAC currently have illustrative source-passage examples. Other tracks provide proposed scope, extraction fields, and review requirements; no live corpus or evaluated extraction is claimed.
02 / THE RESEARCH WORKSPACE
A question.
A trail you can follow.
Explore how a researcher could move from a paper to a source-linked observation—and into independent review.
INTERACTIVE CONCEPT · SYNTHETIC DATAResearch workspace R
PANCREATIC DUCTAL ADENOCARCINOMA
Evidence explorer
Find the observation. Inspect the context. Keep the uncertainty.
Concept previewEvery record and passage below is fictional. They demonstrate the workflow, not clinical findings.
Independent review, made visible.
Inspect an observation before recording a demo decision. Actions stay in this browser session and do not validate real evidence.
A connection is only useful if you can trace it.
Three distinct layers preserve what was reported, what was interpreted, and which knowledge version was used.
01
→Frozen source
Article version + immutable text
UTF-8 SHA-256 fingerprint
02
→Observation
Exact quote + character offsets
Literal finding + assay context
03
Interpretation
Versioned mapping + run identity
Uncertainty + reviewer decision
Knowledge is a versioned dependency.
HGNC, MONDO, and diagnostic classifications would be pinned separately. This demo has no installed ontology releases.
03 / INTEGRITY BY DESIGN
Uncertainty is part
of the evidence.
A useful research tool should make it easier to see what a source supports—and where it stops.
Missing isn’t negative.
Not reported, not tested, indeterminate, and explicitly negative findings need different meanings.
Timing changes the question.
Induction and maintenance treatments stay distinct. Survival and objective response stay distinct, too.
04 / WHERE WE ARE
Building the foundation.
Then earning the insight.
This site shows the product direction.
The research software is an experimental prototype.
Evidence tooling
Metadata indexing, source validation, PDAC cohort safeguards, citation checking, and annotation tools.
89 software tests passed in the audited snapshot. Live integrations still need verification.A 10 + 10 paper pilot
Start with a reproducible GBM and PDAC pilot, then establish disease-specific protocols for additional tracks. Verify citations, source versions, and article-level rights.
No completed real-paper pilot or adjudicated reference corpus yet.Independent evaluation
Two qualified reviewers, preserved annotation drafts, adjudication, and separate measurements for each disease track.
No scientific extraction benchmark scores have been established.THE PURPOSE BEHIND THE PLATFORM
Less time chasing sources.
More time asking better questions.
Built to help researchers inspect, challenge, and reproduce evidence.