The differences in networks of proteins in breast cancer cells may hold the key to determining a patient's chances of recovery.
Research in Canada has resulted in a test that can help doctors predict the outcome for newly-diagnosed breast cancer with 80% accuracy.
Dr Jeff Wrana, at Mount Sinai Hospital, writes in the journal Nature Biotechnology that the new DyNeMo helps oncologists choose treatments best suited to individual circumstances.
Ed Yong, at Cancer Research UK, said: "Not all breast cancers are the same. By working out the differences between them at a molecular level, techniques like these could allow breast cancer patients to receive more personalised treatment from their doctor."
It is hoped that the tool will be available within the next five years, and that it might eventually be adapted for use with other types of cancer.
Breast cancer develops in milk-producing glands and ducts that deliver milk to the nipples. It may spread into the surrounding tissue and other parts of the body.
It is the most common cancer in women in the UK, with 44,600 new cases a year. But earlier detection and improved treatment has lowered the death rate by a fifth in the last 10 years.