A test may be in sight that can identify breast cancers most likely to return after treatment.
Scientists have identified genetic variants that generally only occur in relapsing cancers.
Spotting them at diagnosis could forewarn doctors and help them deliver personalised treatments tailored to at-risk patients.
Dr Lucy Yates, from the Wellcome Trust Sanger Institute at Hinxton, Cambridgeshire, said: "We have found that some of the genetic mutations that drive breast cancers that relapse are relatively uncommon amongst cancers that do not relapse at the point of primary diagnosis.
"We believe that the differences we have seen reflect genetic differences that can predispose a cancer to return, combined with mutations acquired throughout the period from first diagnosis to the subsequent relapse. Some of these genetic alterations are potentially targetable with drugs."
About one in five breast cancers come back after treatment, appearing either at the same place as the original tumour or spreading to other parts of the body.
The new research, to be presented at the 2015 European Cancer Congress in Vienna, shows that genetic factors driving recurring breast cancer are different from those responsible for non-returning tumours.
The Sanger team analysed DNA from tumour samples taken from 1,000 breast cancer patients. In 161 cases samples were taken from recurring tumours or spreading (metastatic) cancers.
Cancer genes found in tumours sampled at first diagnosis were compared with those in relapsed cancers.
Within an individual cancer, a wide range of accumulating factors influencing the function and activity of genes may affect tumour survival, the study found.
Professor Peter Naredi, the European Cancer Organisation scientific co-chair of the Congress, said: "Information such as that which Dr Yates will present is very important in the era of precision medicine.
"Not only can we better choose the right treatment combination as our information about the primary tumour increases, and hence prevent over-treating patients who will not benefit, but this will also help us select the right therapy for each breast cancer patient.
"This study also underlines the fact that we should consider a recurrence of a cancer as a new event, and carefully select the right treatment for the recurrent tumour as opposed to just relying on information from the first occurrence."