A Nature study presents ProteinTalks, a model estimating drug responses from changes in proteins over time. It is not a complete simulation of a person.

The evidence includes 18 breast-cancer cell lines. Evaluation also uses archived specimens from 501 women; such retrospective comparison alone does not demonstrate benefits from model-guided treatment.

Our editorial perspective: imagine a laboratory with hundreds of combinations but limited experimental capacity. A useful model could help prioritize tests. Its value would lie in fewer dead ends, not in an infallible answer.

Independent teams first need to establish whether the proposed experiments yield more useful findings than a simple comparator. Measurement costs, transfer between laboratories and separation of training data from new tests are important.

Our optimistic editorial estimate is 6–18 months for a suitably equipped research team to assess a limited laboratory pilot. This is not a treatment rollout date: patient use would require further safety work and prospective clinical studies.