ABSTRACTThe development of reliable, prognostically informative molecular tests to direct targeted cancer therapy is a major challenge. In 2019, the PI3Kα inhibitor alpelisib was approved for the treatment of advanced breast cancer, in combination with the oestrogen receptor degrader fulvestrant, with some evidence for improved therapeutic response in patients classified as having tumours which were positive for mutation in PIK3CA, the gene encoding PI3Kα. Using human pluripotent stem cells, we recently demonstrated that the PIK3CAH1047R oncogenic hotspot variant shows marked PIK3CA allele dose-dependent activation of PI3K signalling and induction of self-sustained stemness. Together with recent discoveries of multi-copy and double-cis PIK3CA mutations in human cancers, this calls for a re-evaluation of the PIK3CA genotype-phenotype relationship and the current use of binary stratification by PIK3CA mutation status. Using computational analyses, we thus investigated the relationship between PIK3CA mutational status, PI3K activity/signalling strength and stemness. Stemness and PI3K activity scores were calculated using open-source methods and well-established transcriptional signatures. We report that a high PI3K pathway activity score, but not the presence of PIK3CA mutation per se, predicts increased breast cancer dedifferentiation and higher stemness, correlating with reduced overall survival. Our data (1) corroborate reports that the presence of a PIK3CA mutation per se does not predict high PI3K pathway activation or poor prognosis; (2) suggest that stratification of breast cancer for PI3K-based therapy might benefit from the use of a PI3K pathway activity score rather than binary PIK3CA mutation status alone; (3) suggest that combination of PI3K pathway inhibitors with differentiation-promoting treatments warrants evaluation in aggressive breast cancers with high PI3K activity and stemness scores.