# AI in oncology roadmap: pattern readers → foundation models → agents in the workflow

Source: https://onco.cc/roadmaps/ai-oncology-roadmap/  
OnCo record `ai-oncology-roadmap` (Roadmap). Data CC BY-NC 4.0, attribute "Data from OnCo (onco.cc)"; commercial use needs a licence.

## TL;DR

Artificial intelligence in cancer started as software that flagged spots on a mammogram. It now designs molecules, reads slides better than any single pathologist for some tasks, and is beginning to match patients to trials and draft the tumour board summary; the question is which of it will be proven to help.

## Summary

Three strands of AI are converging on oncology. In discovery, structure prediction (AlphaFold 3, Boltz, Chai) and generative chemistry have produced the first AI-designed candidates in trials, and perturbation-scale single-cell datasets are training models that try to predict what a drug will do to a cell. In diagnosis, foundation models trained on millions of slides and scans (Virchow, Prov-GigaPath, UNI, TITAN, CT-FM) underpin the first AI tests cleared to predict treatment benefit (ArteraAI Prostate 2025, ArteraAI Breast 2026) and the first randomised evidence that AI reading improves screening (MASAI). In the clinic, language models are entering trial matching, documentation and tumour-board support, with radiotherapy auto-contouring as the most mature deployed use.

The gap between the thousands of published models and the handful in clinical use is the defining feature of the field. Prospective, ideally randomised, evidence that an AI-guided decision improves an outcome exists for a few tools; a regulatory route for models that keep updating, payment codes for AI-derived biomarkers, and data that can be shared or federated across hospitals are all unsettled.

This roadmap covers the whole stack from molecule to clinic; the companion roadmaps go deeper on the AI-assisted clinic and on the virtual cell.

## Fields

- Kind: Roadmap
- Last checked: 2026-09-10

## Sources

- MASAI randomised trial of AI-supported mammography screening (Lancet Digital Health 2025): https://doi.org/10.1016/S2589-7500(24)00267-X
- FDA: Artificial intelligence-enabled medical devices list: https://www.fda.gov/medical-devices/software-medical-device-samd/artificial-intelligence-enabled-medical-devices

## Connected records

- roadmaps: [AI in the oncology clinic: from narrow cleared tools to multimodal decision support](https://onco.cc/roadmaps/ai-oncology-clinic/), [Drug discovery roadmap: screening in mice → maps of dependency → designing in silico](https://onco.cc/roadmaps/drug-discovery-roadmap/), [Virtual cell roadmap: from bulk omics to a predictive model of a cancer cell](https://onco.cc/roadmaps/virtual-cell/)
- ideas: [A registry of external validation datasets for cancer AI models, with mandatory reporting](https://onco.cc/ideas/idea-tr2-ai-external-validation-registry/), [AI-assisted central imaging reads to cut endpoint cost and variability](https://onco.cc/ideas/idea-tr1-ai-central-imaging-reads/), [AI-first reading for high-volume common cancer diagnoses, pathologist for the exceptions](https://onco.cc/ideas/idea-acc-ai-first-pathology-common-cases/), [Every routine CT scan checked by AI for early cancer signs, with a tracked follow-up pathway](https://onco.cc/ideas/idea-prev-opportunistic-ct-ai-registry/), [In silico trials to choose the dose before the first patient](https://onco.cc/ideas/idea-bio1-in-silico-trials-dose/), [Patient-level multimodal foundation models for treatment selection](https://onco.cc/ideas/idea-multimodal-foundation-model/), [Trial matching inside the electronic record at the moment a treatment is chosen](https://onco.cc/ideas/idea-tr1-ehr-point-of-care-trial-alert/)
- collections: [Arc Virtual Cell Atlas](https://onco.cc/collections/arc-virtual-cell-atlas/), [Flatiron Health and Foundation Medicine Clinico-Genomic Database](https://onco.cc/collections/flatiron-foundation-cgdb/), [NCI Imaging Data Commons (IDC)](https://onco.cc/collections/imaging-data-commons/), [Pathology AI benchmarks (CAMELYON, PANDA, TCGA slide tasks)](https://onco.cc/collections/pathology-benchmarks/), [Tahoe-100M](https://onco.cc/collections/tahoe-100m/)
- technologies: [AI auto-contouring and adaptive planning](https://onco.cc/technologies/auto-contouring-ai/), [AI compute and model platforms for oncology](https://onco.cc/technologies/ai-compute-platforms/), [AI in radiology](https://onco.cc/technologies/radiology-ai-screening/), [AI trial matching & clinical decision support](https://onco.cc/technologies/ai-trial-matching/), [AI-driven drug & target discovery](https://onco.cc/technologies/ai-drug-design/), [Aidoc CARE (clinical radiology foundation model)](https://onco.cc/technologies/aidoc-care/), [AlphaFold 3](https://onco.cc/technologies/alphafold3/), [Boltz-1 / Boltz-2 (MIT, open)](https://onco.cc/technologies/boltz/), [Chai-1 / Chai-2](https://onco.cc/technologies/chai-1/), [Chemistry42 and Pharma.AI (Insilico)](https://onco.cc/technologies/chemistry42/), [CHIEF (Harvard, Yu Lab)](https://onco.cc/technologies/chief/), [CT-FM (whole-body CT foundation model)](https://onco.cc/technologies/ct-fm/), [Digital pathology & AI](https://onco.cc/technologies/digital-pathology-ai/), [Digital twins and virtual control arms](https://onco.cc/technologies/digital-twins-trials/), [ESM3 (EvolutionaryScale)](https://onco.cc/technologies/esm3/), [Federated learning and privacy-preserving AI](https://onco.cc/technologies/federated-learning-medical-ai/), [Foresight (generative EHR model)](https://onco.cc/technologies/foresight-ehr/), [Geneformer](https://onco.cc/technologies/geneformer/), [H-optimus (Bioptimus)](https://onco.cc/technologies/h-optimus/), [Mammography & tomosynthesis](https://onco.cc/technologies/mammography/), [Med-Gemini and MedLM (Google)](https://onco.cc/technologies/med-gemini/), [Merlin (Stanford abdominal CT vision-language model)](https://onco.cc/technologies/merlin-ct/), [Mirai (MIT breast cancer risk from mammograms)](https://onco.cc/technologies/mirai/), [Multidisciplinary tumour boards](https://onco.cc/technologies/multidisciplinary-tumour-board/), [MUSK (Stanford, vision-language pathology)](https://onco.cc/technologies/musk/), [NVIDIA BioNeMo](https://onco.cc/technologies/bionemo/), [Pathology & radiology foundation models](https://onco.cc/technologies/pathology-foundation-model/), [Phenom-2 and Recursion OS](https://onco.cc/technologies/phenom-2/), [Prov-GigaPath (Microsoft, Providence)](https://onco.cc/technologies/prov-gigapath/), [RFdiffusion / RFdiffusion2 and ProteinMPNN (Baker Lab)](https://onco.cc/technologies/rfdiffusion/), [scGPT](https://onco.cc/technologies/scgpt/), [State (Arc Institute perturbation model)](https://onco.cc/technologies/state-arc/), [Sybil (MIT/MGH lung cancer risk from CT)](https://onco.cc/technologies/sybil/), [Tempus multimodal models](https://onco.cc/technologies/tempus-multimodal/), [TITAN (whole-slide multimodal model)](https://onco.cc/technologies/titan/), [UNI and CONCH (Harvard, Mahmood Lab)](https://onco.cc/technologies/uni-conch/), [Virchow / Virchow2 (Paige, MSK)](https://onco.cc/technologies/virchow/), [Whole-slide scanners and image management](https://onco.cc/technologies/whole-slide-scanners/)
- fronts: [AI & Computation](https://onco.cc/fronts/ai-computation/)
- drugs: [ArteraAI Breast](https://onco.cc/drugs/artera-ai-breast/), [ArteraAI Prostate](https://onco.cc/drugs/artera-ai-prostate/), [Paige Prostate Detect](https://onco.cc/drugs/paige-prostate/)
- companies: [Aidoc](https://onco.cc/companies/aidoc/), [Artera](https://onco.cc/companies/artera/), [Flatiron Health (Roche)](https://onco.cc/companies/flatiron-health/), [Google DeepMind (and Google Research)](https://onco.cc/companies/google-deepmind/), [Insilico Medicine](https://onco.cc/companies/insilico-medicine/), [Isomorphic Labs](https://onco.cc/companies/isomorphic-labs/), [Limbus AI](https://onco.cc/companies/limbus-ai/), [Massive Bio](https://onco.cc/companies/massive-bio/), [Microsoft (Research and Health AI)](https://onco.cc/companies/microsoft-research/), [Noetik](https://onco.cc/companies/noetik/), [NVIDIA](https://onco.cc/companies/nvidia/), [Owkin](https://onco.cc/companies/owkin/), [Paige AI](https://onco.cc/companies/paige/), [PathAI](https://onco.cc/companies/pathai/), [Pathos AI](https://onco.cc/companies/pathos-ai/), [Recursion Pharmaceuticals](https://onco.cc/companies/recursion/), [Tempus AI](https://onco.cc/companies/tempus/), [TheraPanacea](https://onco.cc/companies/therapanacea/), [Trial Library](https://onco.cc/companies/trial-library/), [Xaira Therapeutics](https://onco.cc/companies/xaira-therapeutics/)
- trials: [MASAI (Mammography Screening with Artificial Intelligence)](https://onco.cc/trials/masai/)
- bottlenecks: [AI that is built but not validated or deployed](https://onco.cc/bottlenecks/b-ai-validation/), [Data silos](https://onco.cc/bottlenecks/b-data-silos/), [Not enough oncologists, nurses, pathologists, physicists](https://onco.cc/bottlenecks/b-workforce/), [Preclinical results do not reproduce](https://onco.cc/bottlenecks/b-reproducibility/), [Weak real-world evidence and registries](https://onco.cc/bottlenecks/b-real-world-evidence/)

---
JSON: https://onco.cc/api/v1/entities/ai-oncology-roadmap.json