Improving response rate
to first-line therapy in HGSOC.
A consortium to predict platinum response in high-grade serous ovarian cancer.
Join the consortium →
A consortium to predict platinum response in high-grade serous ovarian cancer.
Join the consortium →The stakes
women die of ovarian cancer every year, worldwide
of HGSOC patients relapse after platinum
median survival once platinum resistance sets in
The treatment gap
The FDA has cleared real therapies for platinum-refractory ovarian cancer — but every one of them is gated behind the same thing: documented failure on platinum chemotherapy. For the 20–30% of patients whose tumors were never going to respond, that means months of toxicity, disease progression, and lost time before they reach a drug that might have worked from day one. We can identify them upfront.
First therapy to improve overall survival in platinum-resistant disease. Requires FRα-positive tumor selection and 1–3 prior regimens.
FDA approval →Biomarker-agnostic combination. ROSELLA showed improved PFS and OS over nab-paclitaxel alone in previously treated patients.
OncLive →The long-standing backbone for resistant disease, added to single-agent chemo. Modest benefit, still reached only after platinum fails.
NCI background →Every option above starts the clock after the failure. A validated chemo-response predictor moves the decision to day one — before the first wasted cycle.
The solution
So we're building one — trained by leading cancer centers on H&E histology, genomics, and clinical data from a pool of 10,000 HGSOC patients, to predict platinum response before the first wasted cycle.
Why scale
Language models got predictably better with scale. Real clinical tasks haven't seen that curve yet. There are two ways to scale — more patients, or more modalities per patient. We're scaling both.
Every patient × every modality — H&E, RNA-seq, WES, methylation, proteomics, clinical — scaling both axes toward 10,000.
Founding institutions
Worldwide effort
Twelve founding institutions across four continents, tied together by one shared cohort.
Evidence
Toward 10,000
How we work
Every contributing institution receives co-authorship on flagship papers. Cohort PIs sit on the publications committee.
Federated-first. We ship the training and evaluation container; slides and clinical data never leave your firewall unless you opt in.
Template DTA, IRB protocol language, and a HIPAA-compliant pipeline on day one. The legal groundwork is done.
Ways to participate
Become a founding partner. Co-develop the foundation model. Co-author the science.
Bring your HGSOC patients into the 10,000. More data in, stronger model out. Shared credit, shared model.
Run the model on your cohort. No data leaves your site. We ship the container, you return the metrics.
Team
PI
ML & Multimodal Modeling
Clinical Oncology
Supported by
Join the consortium
Leave your email and we'll be in touch about becoming a partner, contributing a cohort, or running federated validation at your site.