OnCo
frontsFront

AI & Computation

Software that reads scans and slides, predicts outcomes, designs drugs, and matches patients to trials.

FDA-cleared digital pathology risk tools (ArteraAI), radiology triage and screening models, pathology and radiology foundation models, multimodal patient-level models, AI-driven target discovery and ADC design, and LLM-based trial matching and tumour-board support.

AI & Computation: how this front works · animated schematic, not to scale

Technologies

72top
Established
AI auto-contouring and adaptive planning

Software that draws organs and tumours on scans automatically, saving hours per patient and making daily plan adaptation practical.

Emerging
AI compute and model platforms for oncology

AI compute platforms are the GPUs, model libraries, and cloud services that pathology, radiology, and drug-design AI run on.

Established
AI in radiology

Software that reads scans alongside radiologists, catching cancers earlier and predicting who is at risk.

Established
AI trial matching & clinical decision support

Software, increasingly LLM-based, that reads a patient's record and finds trials or guideline options they qualify for.

Phase 2
AI-driven drug & target discovery

Using machine learning to pick targets, design molecules and antibodies, and predict which ADC will work.

Emerging
Aidoc CARE (clinical radiology foundation model)

Aidoc CARE is a single foundation model behind many FDA-cleared triage alerts in emergency radiology.

Established
AlphaFold 3

Predicts the 3D shape of proteins together with DNA, RNA, small molecules and antibodies, the starting point for much modern drug design.

Emerging
AlphaGenome

Reads a million letters of DNA at once and predicts how a mutation changes gene regulation, splicing and chromatin.

Established
AlphaMissense

Scored all 71 million possible single-letter protein changes in humans as likely harmful or benign.

Emerging
Atlas (Aignostics, Mayo Clinic, Charité)

Atlas is a pathology foundation model trained on 1.2 million slides from two of the world's largest hospitals.

Emerging
BioEmu (Microsoft)

Predicts the many shapes a protein moves between, not just one, thousands of times faster than simulation.

Emerging
Boltz-1 / Boltz-2 (MIT, open)

Open-source structure models that match AlphaFold 3, with Boltz-2 also predicting how strongly a drug binds.

Standard of care
Cancer registries and population surveillance

The public systems that count every cancer diagnosis and death in a country, which tell us whether incidence and survival are improving.

Established
Cancer variant knowledgebases and molecular tumour boards

Curated databases that say what each mutation means for treatment, and the expert meetings that use them to decide on therapy.

Established
Cell-therapy orchestration and chain-of-identity software

Scheduling and tracking software that makes sure each patient's cells come back to that patient, on time.

Emerging
Cell2Sentence / C2S-Scale (Yale, Google)

Turns a cell's gene expression into a sentence so a normal language model can reason about it; a 27-billion-parameter version proposed a cancer immunotherapy idea that was confirmed in the lab.

Emerging
CellFM

CellFM is an 800-million-parameter single-cell model trained on 100 million human cells.

Emerging
Chai-1 / Chai-2

Structure and antibody-design models from Chai Discovery, with Chai-2 reporting high zero-shot antibody hit rates.

Emerging
CHIEF (Harvard, Yu Lab)

A pathology model trained across 19 cancer types that predicts survival and mutations from slides.

Established
Clinical NGS bioinformatics and variant interpretation

Software that turns raw sequencer output into a report of which mutations matter and which drugs they point to.

Standard of care
Clinical trial software (EDC, eCOA, CTMS, RTSM)

Clinical trial software is the set of systems that collect trial data, randomise patients, and keep every form auditable.

Standard of care
Contract research organisations (CROs)

Companies that run clinical trials for sponsors: sites, monitoring, data, and regulatory filing.

Emerging
CT-FM (whole-body CT foundation model)

A model pretrained on 148,000 CT scans to segment organs and triage findings.

Established
Decentralised and hybrid clinical trials

Running parts of a trial at home or locally, with telehealth, home nursing, and remote monitoring, so patients far from big centres can take part.

Established
Dermoscopy, total-body photography & AI skin analysis

Magnified skin imaging and whole-body photo mapping, increasingly read by algorithms, to find melanoma early and avoid unnecessary biopsies.

Established
Digital pathology & AI

Scanning microscope slides and letting software measure things a pathologist cannot see, including predictions of who will benefit from a treatment.

Emerging
Digital twins and virtual control arms

Using a model of what would have happened to a patient on standard treatment, so fewer people have to be randomised to it.

Established
Electronic patient-reported outcome (ePRO) symptom monitoring

Patients report symptoms weekly through an app or web form, and nurses respond to alerts. Randomised trials showed this simple system improved quality of life, cut emergency visits and, in one trial, extended survival by five months.

Established
Electronic patient-reported outcomes and remote monitoring

Apps and sensors that let patients report symptoms between visits, which in trials improved survival and cut emergency visits.

Emerging
Enformer and Borzoi (DeepMind, Calico)

Models that predict how DNA sequence controls gene activity, used to interpret non-coding cancer mutations.

Emerging
ESM3 (EvolutionaryScale)

ESM3 is a generative protein model that designed a working fluorescent protein far from any natural sequence.

Emerging
Evo 2 (Arc Institute, NVIDIA)

A DNA language model trained on 9.3 trillion bases that can flag cancer-causing BRCA1 variants without being told about them.

Emerging
Federated learning and privacy-preserving AI

Training AI models across many hospitals without moving patient data, so the model learns from everyone while the data stay put.

Emerging
Foresight (generative EHR model)

A model trained on millions of hospital records that forecasts a patient's next diagnoses.

Emerging
GEARS and perturbation prediction benchmarks

GEARS is a graph model predicting the effect of gene knockouts; the perturbation benchmarks around it showed how hard the problem is.

Emerging
Geneformer

The first widely used transformer trained on millions of single cells, able to predict which genes matter in a disease.

Emerging
GenePT

Uses text embeddings of gene descriptions from a general LLM to represent cells, and performs surprisingly well.

Established
Genomics cloud and secure research environments

Cloud systems where hospitals and researchers store and analyse genomic data securely at petabyte scale.

Emerging
H-optimus (Bioptimus)

An open 1.1-billion-parameter pathology model from a French startup, among the strongest on public benchmarks.

Emerging
Hibou (HistAI)

Hibou is a family of open pathology foundation models under a permissive licence.

Emerging
Med-Gemini and MedLM (Google)

Google's medical versions of its Gemini models, able to reason over text, images, and long records.

Emerging
MedSAM / SAM-Med3D (segment anything for medicine)

Adaptations of Meta's Segment Anything model that outline tumours and organs on any scan with a click.

Emerging
Merlin (Stanford abdominal CT vision-language model)

Merlin is a model trained on 15,000 CT scans with their reports that can find and describe hundreds of findings.

Emerging
Midnight (kaiko.ai)

Midnight is a pathology model that matched the leaders while training on far fewer slides.

Emerging
Mirai (MIT breast cancer risk from mammograms)

Reads a mammogram to estimate five-year breast cancer risk, consistently across races and devices.

Emerging
MUSK (Stanford, vision-language pathology)

A model that reads slides and clinical text together to predict who will respond to immunotherapy.

Emerging
Nicheformer (spatial single-cell)

Nicheformer is a model trained on both dissociated and spatial data so it learns how a cell's neighbourhood shapes it.

Emerging
Nucleotide Transformer (InstaDeep)

DNA language models trained on thousands of genomes for variant and regulatory prediction.

Established
NVIDIA BioNeMo

BioNeMo is the software stack many biology foundation models are trained and served with.

Established
Oncology EHR and real-world data platforms

Databases built from millions of real patient records, used to see how treatments work outside trials and to run studies without new trials.

Standard of care
Oncology EHR modules and treatment pathways

The ordering and record systems oncologists use every day, including built-in treatment pathways that steer drug choice.

Emerging
Pathology & radiology foundation models

Very large AI models trained on millions of slides or scans that can be adapted to almost any diagnostic question.

Emerging
Phenom-2 and Recursion OS

A model trained on billions of cell microscopy images to read what a drug or gene knockout does to a cell.

Emerging
Phikon / Phikon-v2 (Owkin)

Owkin's open pathology models trained on TCGA and its federated hospital network.

Emerging
PLUTO (PathAI)

PathAI's compact pathology foundation model designed to run across many tasks and resolutions.

Emerging
Prov-GigaPath (Microsoft, Providence)

An open pathology model trained on 1.3 billion image tiles from a US health system, modelling whole slides at gigapixel scale.

Emerging
RadFM (generalist radiology foundation model)

An open generalist model that answers questions about 2D and 3D scans.

Standard of care
Radiotherapy treatment planning and QA software

The software that calculates exactly how radiation beams should be shaped and checks the machine delivered it.

Emerging
RFdiffusion / RFdiffusion2 and ProteinMPNN (Baker Lab)

The tools that design entirely new proteins to bind a chosen target, now used for cancer binders and antibodies.

Emerging
scFoundation (BioMap)

scFoundation is a 100-million-parameter model trained on 50 million cells, from China's BioMap.

Emerging
scGPT

A GPT-style model for single-cell data that predicts cell types, perturbation responses, and gene networks.

Emerging
Spatial-omics-guided treatment selection

Choosing treatment from a map of where each cell type sits in the tumour, not just from a list of its mutations.

Emerging
State (Arc Institute perturbation model)

Predicts how cells will respond to a drug or gene knockout, trained on over 100 million perturbed cells.

Emerging
Sybil (MIT/MGH lung cancer risk from CT)

Predicts a person's six-year lung cancer risk from one low-dose CT, even when no nodule is visible.

Established
Telehealth and hospital-at-home in oncology

Video visits, remote monitoring and home delivery of some cancer treatments expanded massively during COVID-19 and have stayed; they reduce travel burden, especially for rural patients, without evidence of worse outcomes.

Emerging
Tempus multimodal models

Models trained on Tempus's paired genomic, pathology, imaging and outcome data to predict response and prognosis.

Emerging
TITAN (whole-slide multimodal model)

TITAN is a model that summarises a whole slide, not just tiles, and can write a draft pathology report.

Emerging
TranscriptFormer and rBio (CZI virtual cell models)

CZI's open cross-species cell models and a reasoning model trained on them.

Emerging
UNI and CONCH (Harvard, Mahmood Lab)

Two open academic pathology models: UNI reads tissue images, CONCH links images with pathology text.

Emerging
Universal Cell Embedding (UCE)

Universal Cell Embedding maps any cell from any species into one shared space without retraining.

Emerging
Virchow / Virchow2 (Paige, MSK)

A pathology foundation model trained on millions of slides that can detect cancer and predict biomarkers from an ordinary H&E slide.

Established
Whole-slide scanners and image management

The scanners that turn glass slides into gigapixel images, and the software that stores and serves them, without which pathology AI cannot run.

Key papers

2top

Connected

258top

technologies

72
AI auto-contouring and adaptive planningAI compute and model platforms for oncologyAI in radiologyAI trial matching & clinical decision supportAI-driven drug & target discoveryAidoc CARE (clinical radiology foundation model)AlphaFold 3AlphaGenomeAlphaMissenseAtlas (Aignostics, Mayo Clinic, Charité)BioEmu (Microsoft)Boltz-1 / Boltz-2 (MIT, open)Cancer registries and population surveillanceCancer variant knowledgebases and molecular tumour boardsCell-therapy orchestration and chain-of-identity softwareCell2Sentence / C2S-Scale (Yale, Google)CellFMChai-1 / Chai-2CHIEF (Harvard, Yu Lab)Clinical NGS bioinformatics and variant interpretationClinical trial software (EDC, eCOA, CTMS, RTSM)Contract research organisations (CROs)CT-FM (whole-body CT foundation model)Decentralised and hybrid clinical trialsDermoscopy, total-body photography & AI skin analysisDigital pathology & AIDigital twins and virtual control armsElectronic patient-reported outcome (ePRO) symptom monitoringElectronic patient-reported outcomes and remote monitoringEnformer and Borzoi (DeepMind, Calico)ESM3 (EvolutionaryScale)Evo 2 (Arc Institute, NVIDIA)Federated learning and privacy-preserving AIForesight (generative EHR model)GEARS and perturbation prediction benchmarksGeneformerGenePTGenomics cloud and secure research environmentsH-optimus (Bioptimus)Hibou (HistAI)Med-Gemini and MedLM (Google)MedSAM / SAM-Med3D (segment anything for medicine)Merlin (Stanford abdominal CT vision-language model)Midnight (kaiko.ai)Mirai (MIT breast cancer risk from mammograms)MUSK (Stanford, vision-language pathology)Nicheformer (spatial single-cell)Nucleotide Transformer (InstaDeep)NVIDIA BioNeMoOncology EHR and real-world data platformsOncology EHR modules and treatment pathwaysPathology & radiology foundation modelsPhenom-2 and Recursion OSPhikon / Phikon-v2 (Owkin)PLUTO (PathAI)Prov-GigaPath (Microsoft, Providence)RadFM (generalist radiology foundation model)Radiotherapy treatment planning and QA softwareRFdiffusion / RFdiffusion2 and ProteinMPNN (Baker Lab)scFoundation (BioMap)scGPTSpatial-omics-guided treatment selectionState (Arc Institute perturbation model)Sybil (MIT/MGH lung cancer risk from CT)Telehealth and hospital-at-home in oncologyTempus multimodal modelsTITAN (whole-slide multimodal model)TranscriptFormer and rBio (CZI virtual cell models)UNI and CONCH (Harvard, Mahmood Lab)Universal Cell Embedding (UCE)Virchow / Virchow2 (Paige, MSK)Whole-slide scanners and image management

companies

54

institutions

20

ideas

107
A cancer data donor card: patient-controlled donation of records for researchA chatbot for pre-test genetic counselling so counsellors see only who needs themA common consent and material transfer template for tumour biobanksA consented commons of surgical video linked to pathology and outcomesA dedicated fund for randomised trials of cancer AI with patient outcomesA digital second-opinion network answering community oncologists within 72 hoursA liability framework for clinical AI: safe harbour for clinicians, liability for makersA machine-readable treatment summary handed to every patient and readable by any hospitalA mandatory silent (shadow) trial before any cancer AI goes liveA monthly-updated benchmark for AI answers to oncology questions with citation accuracyA national cancer data space with one legal front doorA national repository of radiotherapy dose plans linked to outcomesA neutral public evaluator for cancer AI, on the model of NISTA patient-held cancer record that travels across providers and bordersA plain-language explainer for every guideline recommendation, in every major languageA plain-language summary of every cancer trial result within a yearA pre-competitive consortium to train a shared multimodal cancer foundation modelA pre-registered standard for emulating trials with real-world dataA public API serving the current standard of care for any cancer, stage and biomarkerA public registry of every AI model used in cancer careA public registry of unanswered clinical questions linked to funding callsA public tracker of how long each country takes to adopt new evidenceA randomised trial of AI scribes in oncology clinics measuring errors and timeA randomised trial of AI-generated treatment recommendations versus tumour boardsA real-world sequencing analysis within a year of every new approvalA regulatory sandbox for continuously learning cancer AIA standard evaluation pathway for AI-assisted pathology, from reader study to deploymentA standard for monitoring AI performance drift with pause thresholdsA structured registry for every off-label cancer drug useA synthetic twin of every restricted cancer dataset for code developmentA tumour board assistant that cites its evidence and tracks outcomesAccredited trusted research environments with curated cancer tablesAmbient AI note-taking to give oncologists back a day a weekAn automatic electronic frailty index inside the oncology recordAn international consortium pooling the outcome of every treated child with cancerAn open foundation model of the cancer cell trained on perturbation dataAn open knowledge graph linking trials, results, biomarkers, drugs and recommendationsAn open library pairing completed cancer trials with real-world emulationsAutomatic flagging of retracted or corrected evidence in guidelines and decision supportAutomatic weekly linkage of cancer registries to deaths, prescriptions and imagingBroad research consent as a routine step of the cancer pathwayCertify every oncology software product for a standard bulk data exportDecision support that cites the exact trial and guideline line it relies onDeposit trial results as structured data, not just PDFsDigital twins for treatment selection, validated by predicting before observingDigitise the nation's pathology slides and link them to outcomesDynamic consent with usage receiptsEnd the abstract-to-paper gap: require full results with any conference presentationEnforce individual participant data sharing as a condition of publication and fundingEvery AI output logged in the record with input hash, version and clinician responseEvery patient on an accelerated-approval drug enrolled in a registry until confirmationEvery tumour genomic report machine-readable and deposited nationallyExpert-verified translation of guideline updates into 20 languages within 30 daysExternal validation at five or more sites in two countries before clearanceFederated training of pathology and radiology models across hospitalsFollow every trial participant for 30 years through registry linkageFund expert editors for the cancer pages of Wikipedia and WikidataFunded living systematic reviews for every major cancer indicationGood practice standards and inspection for real-world data sourcesIn silico trials to prioritise combinations, scored against later real trialsLink every national cancer registry to tumour genomicsLink single-cell and spatial tumour atlases to clinical outcomesLinked prescribing and outcome data to find drug interactions with cancer therapyLive guideline-concordance dashboards for every tumour board, generated from the recordLiving guidelines published as versioned, computable rulesMandatory post-market performance reporting for cancer AIMandatory real-world reporting for patients excluded from pivotal trialsMandatory subgroup performance reporting for cancer AIMap trial case report forms to the registry standard so trial and routine data joinMicro-learning pushed to community oncologists within 30 days of a practice changeNo mCODE, no payment: tie oncology reimbursement to a minimal structured recordOffline decision support for generalists treating common cancers in low-resource settingsOne certified open-source de-identification pipeline for scans and slidesOne legal framework for pooling rare cancer data across bordersOne outcome record for every donated or discounted cancer drug packOpen licences for publicly funded cancer guidelinesOpen-source cancer registry-in-a-box for low-resource settingsOpen, benchmarked algorithms for lines of therapy and progression from routine dataPatient-reported symptoms captured as standard structured data in every clinicPatients told which AI is used in their care, in plain languagePay for cancer AI only when it has outcome evidence, then pay properlyPayers fund cancer drugs conditionally on a registry with a pre-specified analysisPoint-of-care randomisation built into the oncology recordPre-registration and results reporting for real-world cancer studiesPrivacy-preserving linkage tokens for every cancer data holderPublic data-quality scorecards for every cancer centrePublic procurement clauses banning data export fees and lock-inPublicly funded cancer AI must release open weights and model cardsRandomised trials of how to spread proven cancer treatments, not just what worksReal-time guideline-concordance feedback for every cancer centreReal-world dose intensity and toxicity monitoring to revise labelled dosesReassess cancer drug prices at three years using real-world outcomesRed-team programmes that attack cancer AI before patients doRegistry-based randomised trials for oncology comparative effectivenessRequire stage-shift or interval-cancer endpoints for AI in cancer screeningRound-the-clock remote treatment-planning hubs for clinics without physicistsRules for retiring cancer AI when performance drops or the standard of care movesSend the code to the data: a federated analytics network of cancer centresSequestered, prospectively collected benchmark datasets that no one can train onSimulate each hospital's cancer pathway as a queue to find and remove the waitsStandards for external control arms built from federated real-world dataStructured, coded radiology reports for cancer response instead of free textSubscribable alerts when the standard of care changes for a patient's situationToxicity-management decision support for nurses and pharmacistsUpdate hospital order sets within 30 days of a guideline changeValidate real-world progression against central imaging reviewWearable activity data as a validated real-world endpoint

bottlenecks

3

key papers

2