Evidex uses AI to structure clinical signals across disease, stage, subtype, line of therapy, prior treatments, and genomic variants. It unifies publications, guidelines, regulatory approvals, public databases, and conference abstracts into one evidence base, so every match can be traced to its source.
Search in the workspace or integrate the same capabilities through an API.
Live evidence workflow
Select a supported cancer type and variant. The result comes from the same Evidex API used by the workspace.
Ready for a structured query
No patient identity or narrative medical record is sent.
The current narrative workspace is available in Chinese. Structured labels are localized here.
Open Chinese workspaceNo full medical record or patient identity data is required. Start with an available cancer type and variant.
Treatment evidence search is available now. Multidimensional matching, clinical trials, and additional sources are product directions.
Simple rules and manual search struggle to deliver precision, evidence coverage, and speed at the same time.
The same variant can carry different evidence and limitations across cancer type, stage, subtype, line of therapy, and treatment history.
Matches lack clinical context
Reviewers repeat secondary filtering
Interpretation criteria drift between people
Publications, labels, guidelines, databases, trial registries, and conference abstracts live in separate systems with different structures.
Teams repeat searches across sites
Evidence language and citations vary
Quality review must rebuild the source trail
Search, synthesis, citation, and review compete for the same limited medical resources as sample volume grows.
Report turnaround is affected
Repetitive work grows with volume
Senior reviewers remain a bottleneck
Evidex is designed to turn authorized clinical inputs into structured signals that can inform matching.
More specific candidate matches and limitations, not a context-free drug list.
Medical knowledge becomes searchable, comparable, and citable evidence while preserving sources, populations, conclusions, and limitations.
One reviewable result for reporting, quality control, and historical traceability.
PubMed and FDA available now. Additional sources are planned.
Product direction illustration
Organize drugs and regimens with evidence direction, regulatory status, study populations, clinical context, and rationale.
Limited treatment evidence matching is available now.
Match studies using disease, stage, subtype, line of therapy, prior treatments, biomarkers, and key eligibility criteria.
Planned
Candidate matches would not confirm eligibility. Final eligibility remains with the study team.
Start with structured fields. Authorized clinical documents remain a future direction after privacy and security requirements are met.
Identify disease, stage, subtype, line of therapy, prior treatments, and biomarkers while preserving source context.
Search approved knowledge releases. The model cannot add sources outside the retrieved evidence.
Organize rationale, populations, evidence strength, limitations, and sources for professional review.
Select case criteria, then review treatment matches, study limitations, and original sources.
Submit structured requests and receive evidence fields, citation relationships, and knowledge versions for internal workflows.
Pilot discussion only. Production authentication, quotas, SDKs, and service commitments are not yet available.
POST /api/v1/evidence-answer{
"disease": "NSCLC",
"biomarkers": [
{
"gene": "EGFR",
"alterationType": "SNV",
"hgvsp": "p.L858R"
}
],
"jurisdiction": "US",
"locale": "zh-CN"
}View the API example
{
"status": "ANSWERED",
"knowledge": {
"release": "v0.2.0"
},
"resultGroups": "PublicEvidenceResultGroup[]",
"answer": "EvidenceAnswerDraft",
"disclaimer": "string"
}Each candidate explains why it was matched, which population the evidence applies to, its strength, its limitations, and its source.
Current real example
Non-small cell lung cancer + EGFR p.L858R
Structured query
Reviewed knowledge release
Deterministic evidence pack
Citation allowlist
Reviewable result
Treatment evidence organized by therapy
Same-disease and exact-variant cross-indication evidence kept separate
Sensitivity, resistance, and exploratory directions
Study design, sample size, stage, line, and prior treatment
PMID, FDA records, permitted passages, and knowledge release
Narrative summary kept separate from structured evidence
These figures describe the current demonstration release, not the final product scope or all public medical evidence.
2
Cancer types
2
Genes
5
Variants
6
Direct query combinations
20
Clinical evidence claims
9
FDA approval records
22
Source documents
Only reviewed and published evidence enters retrieval
The same input and release return a deterministic evidence set
The model summarizes the Evidence Pack and does not decide retrieval
Every treatment statement cites evidence from the current request
References outside the retrieved set are rejected
Structured evidence remains if the narrative summary fails
Missing and out-of-scope results are shown explicitly
v0.2.0
Evidex is designed for genomic testing, clinical interpretation, and medical teams that need evidence search, matching, report support, and internal review.
No. Evidex organizes patient signals, matches evidence, and explains sources. Final interpretation and medical review remain with qualified professionals.
Not in the current demonstration release. Clinical trial matching is a product direction built around patient criteria, registry data, and eligibility requirements.
The current release validates human review, deterministic retrieval, citation controls, and source traceability before expanding coverage.
No. It means the current Evidex release does not contain eligible evidence for that query, or the query is outside the current scope.
A structured request and response example is available. Production integration still requires authentication, quotas, logging, security, and service commitments.
The model summarizes an Evidence Pack built from reviewed knowledge. It cannot decide retrieval, regulatory status, or evidence levels, and citations pass an allowlist check.
No. The current release does not accept narrative records or collect names, identity numbers, contact details, or medical record numbers.
Evidex supports professional interpretation and review. Each organization decides whether and how outputs enter reports under its own medical, quality, and compliance process.
Start with evidence search and source traceability, then evaluate how patient context, clinical trials, and integration could fit your reporting workflow.
The external demo request channel is being configured. No contact request is collected on this page yet.
No system integration is required. Start with one supported cancer type and variant.