# Digital twins for treatment selection, validated by predicting before observing

Source: https://onco.cc/ideas/idea-data-digital-twin-predict-then-observe/  
OnCo record `idea-data-digital-twin-predict-then-observe` (Idea). Data CC BY-NC 4.0, attribute "Data from OnCo (onco.cc)"; commercial use needs a licence.

## TL;DR

Build a computer model of each patient's cancer that forecasts how it will respond to each treatment option, and prove it by writing the forecast down before the real result is known.

## Summary

Patient digital twins (mechanistic, statistical or hybrid models of a patient's tumour and physiology) are proposed for treatment selection, but validation is almost entirely retrospective. The proposal is a validation programme with a strict protocol: for each enrolled patient, the twin's prediction (response, progression time, toxicity) for the chosen treatment is locked before treatment; predictions are compared with observed outcomes; calibration and discrimination are published. Only twins that pass proceed to trials where predictions inform choices.

## Fields

- Kind: Idea
- Last checked: 2026-09-08
- Hypothesis: Prospectively locked twin predictions will achieve clinically useful calibration for at least one common decision (for example, response to first-line chemo-immunotherapy in NSCLC), and a randomised trial of twin-informed selection will improve response rates.
- Rationale: Weather and engineering models earned trust through routine, scored forward prediction; medical models have skipped this step and are trusted or dismissed on retrospective fits.
- Proposed test: Enrol 500 patients across two cancers; lock predictions; publish calibration plots and Brier scores; proceed to a randomised trial only if pre-specified thresholds are met.
- Maturity: preclinical-evidence
- Actor: research

## Sources

- Bottleneck evidence (AI that is built but not validated or deployed): Wu et al., How medical AI devices are evaluated: limitations and recommendations from an analysis of FDA approvals (Nature Medicine 2021): https://doi.org/10.1038/s41591-021-01312-x

## Connected records

- ideas: [In silico trials to prioritise combinations, scored against later real trials](https://onco.cc/ideas/idea-data-in-silico-trials-calibrated/), [Patient-level multimodal foundation models for treatment selection](https://onco.cc/ideas/idea-multimodal-foundation-model/)
- fronts: [AI & Computation](https://onco.cc/fronts/ai-computation/)
- technologies: [Functional (ex vivo) drug testing](https://onco.cc/technologies/functional-drug-testing/), [Patient-derived organoids](https://onco.cc/technologies/organoids/)
- bottlenecks: [AI that is built but not validated or deployed](https://onco.cc/bottlenecks/b-ai-validation/), [Tumour heterogeneity and clonal evolution](https://onco.cc/bottlenecks/b-tumor-heterogeneity/)
- key papers: [How medical AI devices are evaluated: limitations and recommendations from an analysis of FDA approvals](https://onco.cc/key-papers/paper-wu-nat-med/)

---
JSON: https://onco.cc/api/v1/entities/idea-data-digital-twin-predict-then-observe.json