AI reveals new types of breast cancer that respond differently to treatment

Scientists have used artificial intelligence to recognize the characteristics of bad cancer and have discovered five new types of the disease, each corresponding to a personalized treatment.

Their study applied AI and machine learning to the gene sequences and molecular data of bad tumors to reveal crucial differences between cancers previously grouped into one type.

The new study, led by a team from the Institute of Cancer Research in London, found that two of the types were more likely to respond to immunotherapy than others, while one was more likely to relapse with tamoxifen .

Researchers are currently developing tests for these types of bad cancer that will be used to select patients for different drugs during clinical trials, with the goal of making personalized therapy a standard part of treatment.

The researchers previously used AI in the same way to discover five different types of bowel cancer and oncologists now evaluate their application in clinical trials.

The goal is to apply the AI ​​algorithm to many types of cancer and to provide everyone with information about their sensitivity to treatment, the likely pathways of evolution and the how to fight against drug resistance.

The new research, published today (Friday) in the journal Breast cancer NPJ, could not only help choose treatments for women with bad cancer, but also identify new therapeutic targets.

The Cancer Institute (ICR) – a charitable organization and research institute – funded the study itself from its own donations.

The majority of bad cancers develop in the internal cells lining the bad ducts and are "fed" by estrogen or progesterone hormones. These are clbadified as "luminal A" tumors and often have the best cure rates.

However, patients in these groups respond very differently to standard treatments, such as tamoxifen, or to new treatments, needed in case of relapse of patients, such as immunotherapy.

The researchers applied AI-trained computer software to a wide range of available data on the genetics, molecular and cellular composition of luminal primitive mammary tumors, as well as data on patient survival.

Once formed, artificial intelligence was able to identify five different types of diseases with particular treatment response patterns.

Women with a type of cancer labeled "inflammatory" had immune cells present in their tumors and high levels of a protein called PD-L1 – suggesting that they were likely to respond immunotherapies.

Another group of patients had "triple negative" tumors – which do not respond to standard hormone treatments – but various indicators suggest that they might also respond to immunotherapy.

Patients with tumors containing a specific change in chromosome 8 had worse survival than other groups treated with tamoxifen and tended to relapse much earlier – after an average of 42 months versus 83 months in patients with a tumor type different stem cells. These patients may benefit from additional or new treatment to delay or prevent late relapses.

The markers identified in this new study do not question the general clbadification of bad cancer, but they discover additional differences in the current subdivisions of the disease, with important implications for treatment.

The use of AI to understand the complexity and evolution of cancer is one of the main strategies pursued by CI as part of a pioneering research program to combat cancer's ability to adapt and to become resistant to drugs. The ICR has raised the final sum of a £ 75 million investment in a new £ 15 million GBP Cancer Research Center, which will house the world's first ever global cancer program. "anti-evolution" therapies.

Dr. Anguraj Sadanandam, Head of System Medicine and Precision Cancer Research at the Institute of Cancer Research in London, said:

"We are at the dawn of a revolution in the health sector because we are really seizing the opportunities that are available to AI and the US. machine learning.

"Our new study showed that AI was able to recognize the characteristics of bad cancer that exceeded the limit of the human eye and to point us to new ways of treatment among those who had stopped responding to conventional hormonal therapies, to be used much more widely, and we believe that we can apply this technique to all cancers, even opening up new treatment options for cancers for which there is currently no effective option. "

Dr. Maggie Cheang, a pioneer in identifying different types of bad cancer and head of the genomic badysis clinical trials team at the London Cancer Institute, said:

"Doctors have been using the current clbadification of bad cancer as a treatment guide for years, but it is a fairly rudimentary method and patients who seem to have the same type of illness often respond very differently to drugs.

"Our study used artificial intelligence algorithms to pinpoint bad cancer trends that human badysis had until now not noticed – and found out about the effects of bad cancer. other types of the disease that respond very specifically to treatment.

"Among the interesting implications of this research, there is its ability to identify women who might respond well to immunotherapy, even when the general clbadification of their cancer would suggest that these treatments would not work for them.

"The AI ​​used in our study could also be used to discover new drugs for people at highest risk of late relapse, beyond age 5, who are common in estrogen-related bad cancer and may cause considerable anxiety in the patients. "

In addition to funding the ICR charity, the work was also supported by the NIHB Biomedical Research Center at the Cancer Institute in London, and by the Royal Marsden NHS Foundation Trust.