DoubleChecks

What was claimed

New AI model COMPASS proves your tumor can lie to doctors, dramatically outperforming old biomarkers for immunotherapy response across 33 cancer types

Our verdict

Needs Caution

The study shows tumors can have misleading or incomplete biomarker signals for response prediction, but it does not prove a tumor can 'lie.' That is rhetorical wording, not a scientific claim. The study reports better performance than baseline methods and notes that biomarkers like TMB and PD-L1 often fail across settings. 'Dramatically outperforming old biomarkers' is stronger than the sources state and is not quantified in the claim. (Only 2 of 3 AI systems responded.)

All 2 AI systems agree8 sources citedChecked Jul 4, 2026

Check your own claim

Paste any statement, headline, or AI answer — 3 independent AIs verify it in seconds, with sources.

Key findings

COMPASS dramatically outperforms old biomarkers for immunotherapy response across 33 cancer types.

Misleading86%
1 of 2 AIs agree·ChatGPT: Can’t verify

COMPASS 'proves your tumor can lie to doctors.'

Misleading92%
All 2 AIs agree

New AI model COMPASS exists and was trained using data from 33 cancer types.

Verified98%
All 3 AIs agree

Verified against 8 sources

pmc.ncbi.nlm.nih.govpubmed.ncbi.nlm.nih.govhms.harvard.edulinkedin.comnature.comx.comtheengineer.co.ukfirstwordpharma.com

Detailed Analysis

The core claim about COMPASS is broadly accurate, but the wording is exaggerated. The model did train on 33 cancer types and outperformed existing methods and common biomarkers in multiple cohorts, but the source says it improves prediction rather than proving tumors can “lie” to doctors. The phrase “dramatically outperforming old biomarkers” is subjective and not stated that way in the cited sources.

Why this verdict

  • The core claim about COMPASS is broadly accurate, but the wording is exaggerated.
  • The model did train on 33 cancer types and outperformed existing methods and common biomarkers in multiple cohorts, but the source says it improves prediction rather than proving tumors can “lie” to doctors.
  • The phrase “dramatically outperforming old biomarkers” is subjective and not stated that way in the cited sources.

Claims checked

dramatically outperforming old biomarkers for immunotherapy response

Misleading86%
The study reports better performance than baseline methods and notes that biomarkers like TMB and PD-L1 often fail across settings. 'Dramatically outperforming old biomarkers' is stronger than the sources state and is not quantified in the claim.

This claim may mislead readers without added context.

proves your tumor can lie to doctors

Misleading94%
The study shows tumors can have misleading or incomplete biomarker signals for response prediction, but it does not prove a tumor can 'lie.' That is rhetorical wording, not a scientific claim.

This claim may mislead readers without added context.

New AI model COMPASS

Verified99%
The sources describe a new model named COMPASS for immunotherapy response prediction.

What other AI think

ClaudeDid not respond
The COMPASS model is a real, recently published AI that was trained on transcriptomes from 33 cancer types and evaluated on immunotherapy cohorts. The headline phrasing 'tumor can lie to doctors' is a metaphor and misleading. Claims about 'dramatically outperforming old biomarkers' are partly supported by reported improvements but are context-dependent and not uniformly quantified across all cancers.

Why this verdict

  • The COMPASS model is a real, recently published AI that was trained on transcriptomes from 33 cancer types and evaluated on immunotherapy cohorts.
  • The headline phrasing 'tumor can lie to doctors' is a metaphor and misleading.
  • Claims about 'dramatically outperforming old biomarkers' are partly supported by reported improvements but are context-dependent and not uniformly quantified across all cancers.

Claims checked

COMPASS 'proves your tumor can lie to doctors.'

Misleading90%
This is a rhetorical/metaphorical claim; COMPASS shows some patients with low PD-L1 or low TMB may still respond, but it does not 'prove' tumors are intentionally deceptive.

This claim may mislead readers without added context.

COMPASS dramatically outperforms old biomarkers for immunotherapy response across 33 cancer types.

Can’t verify70%
Public summaries report COMPASS improves prediction versus some existing biomarkers in many cohorts, but 'dramatically' and 'across 33 cancer types' are broad, unquantified claims that depend on metric, cohort, and cancer type and cannot be universally confirmed without detailed per-cancer results.

New AI model COMPASS exists and was trained using data from 33 cancer types.

Verified95%
Published reports and the COMPASS project documentation state the model was pretrained on TCGA transcriptomes covering 33 cancer types.

Share this result

Try DoubleChecks