I'm looking for a tool that lets us do Kaplan Meier for subtype of breast cancer. (or publication showing this graph for TCGA, most of the time it's ER+/ER- but it should exists for PAM50 somewhere) Something like http://kmplot.com/analysis/.

At the same time I writing the question I finally found an answer. https://portal.gdc.cancer.gov/ has a tool to do the survival based on list of patient. But still main question is open, at the end of the message. See next.

From publications, Luminal A has the best prognosis. Basal-like (contains most of the triple negative which lack Hormone Receptor and HER2 amplification) should have the worst prognosis.

I don't retrieve this using the following datasets. (and also just using the cancer.gov tool)

Using time of survival from data from An Integrated TCGA Pan-Cancer Clinical Data Resource to Drive High-Quality Survival Outcome Analytics and PAM50 subtype annotation from A Comprehensive Pan-Cancer Molecular Study of Gynecologic and Breast Cancers . Both recent ressources from 2018.

I'm using the the twenty-year survival time to compute the survival (as it's done on the portal of gdc.cancer.gov) Can we truncate the data to do the KM only to 5 years or 10 years. (and censure all the deaths after these period) .Is it wrong or can we do that ?