# An open model of every cancer cell state, built from perturbation atlases

Source: https://onco.cc/ideas/idea-moon-open-cancer-cell-state-model/  
OnCo record `idea-moon-open-cancer-cell-state-model` (Idea). Data CC BY-NC 4.0, attribute "Data from OnCo (onco.cc)"; commercial use needs a licence.

## TL;DR

Map every state a cancer cell can be in, and how drugs and the surrounding tissue move it between states, into an open computational model anyone can query and improve.

## Summary

Single-cell and spatial atlases (Human Tumor Atlas Network, Human Cell Atlas) describe cell states; perturbation screens and foundation models trained on them begin to predict responses. The proposal is a coordinated, openly licensed effort to generate perturbation-response single-cell data across hundreds of models and patient samples, train and release a foundation model of cancer cell state transitions, and benchmark it prospectively against drug response in organoids and trials, with the data, weights and benchmarks all public.

## Fields

- Kind: Idea
- Last checked: 2026-09-08
- Hypothesis: An open cell-state model predicts drug response and resistance transitions in held-out patient samples better than existing biomarkers and shortens target and combination discovery cycles measurably.
- Rationale: Heterogeneity and plasticity defeat single-marker approaches; only a model of states and transitions captures them. Open weights and benchmarks avoid the reproducibility failures of closed models.
- Proposed test: Release version one with prospective benchmark on organoid response; then a biomarker-defined trial in which model-predicted responders are enriched and outcomes compared with standard selection.
- Maturity: preclinical-evidence
- Actor: research

## Sources

- Human Tumor Atlas Network: https://humantumoratlas.org/
- Human Cell Atlas: https://www.humancellatlas.org/

## Connected records

- technologies: [CRISPR functional genomics](https://onco.cc/technologies/crispr-screens/), [Pathology & radiology foundation models](https://onco.cc/technologies/pathology-foundation-model/), [Patient-derived organoids](https://onco.cc/technologies/organoids/), [Single-cell & spatial profiling](https://onco.cc/technologies/single-cell-spatial/)
- companies: [10x Genomics](https://onco.cc/companies/10x-genomics/)
- institutions: [Broad Institute of MIT and Harvard](https://onco.cc/institutions/broad-institute/)
- bottlenecks: [Lab models that fail to predict what happens in patients](https://onco.cc/bottlenecks/b-preclinical-models/), [Preclinical results do not reproduce](https://onco.cc/bottlenecks/b-reproducibility/), [Tumour heterogeneity and clonal evolution](https://onco.cc/bottlenecks/b-tumor-heterogeneity/)

---
JSON: https://onco.cc/api/v1/entities/idea-moon-open-cancer-cell-state-model.json