OnCo
ideasIdea

An organotropism atlas that predicts where a cancer will spread

Different cancers favour different organs, and so do different patients. A model that predicts which organ is at risk could target surveillance and prevention.

Site of relapse is recorded in registries and trial datasets but almost never modelled as an outcome. Linking primary tumour genomics, transcriptomics and digital pathology to first-relapse site across tens of thousands of patients would produce an organotropism predictor, and would identify the features that drive lung, liver, bone and brain tropism in humans rather than in mice.

Hypothesis
A multimodal model trained on primary tumour features predicts first metastatic site with an AUC above 0.7 in held-out cohorts, and its top features nominate druggable organ-specific seeding mechanisms.
Rationale
Human organotropism is strikingly reproducible by subtype — HER2-positive breast to brain, luminal breast to bone, uveal melanoma to liver — but has never been systematically explained. Existing registry and trial data can answer it without collecting new samples.
What would test it
Federated analysis across three national registries plus genomics-linked outcome data; publish the model, then validate it prospectively by predicting relapse site in an ongoing adjuvant cohort.
Maturity
speculative
Who has to act
data
Cost to try
Medium ($1M to $50M)
Years to first evidence
6
Bottlenecks it attacks

Connected

11top

Pages like this

not linked directly; found by shared links