Machine learning AI tools: helping oncologists make better treatment decisions

When treating cancer patients, the goal of oncologists is to predict the patient’s disease progression and make key treatment decisions

Researchers at the Broad Institute of MIT, Harvard University, and the Dana-Farber Cancer Institute have developed a new model that can distinguish the genomic characteristics of deadly prostate cancer from those that are unlikely to cause symptoms or death

Eliezer (Eli) Van Allen, senior author of the study, an associate professor at Broad, an associate professor at the Dana-Farber Cancer Institute and Harvard Medical School, said that P-NET not only provides prognosis for patients, "we have not only improved the prediction" whether cancer will The ability to metastasize and which genes may be associated with this state, and as cancer researchers, we can use the interpretability of this model to understand the biology of these disease states

Build a better model

In order to build a model that can distinguish early and advanced prostate cancer tumors, the researchers developed a specialized deep learning model with a customized architecture that is more interpretable than other algorithms

Using this method, a team led by Haitham Elmarakeby, a lecturer at the Dana-Farber Cancer Institute, an affiliated researcher at the Broad Institute, and the first author of the study, combined relevant biological information-such as known genes and metabolism or signaling pathways.

In the process of ranking related genes and signaling pathways based on importance and verifying P-NET by weight, the team also determined that the MDM4 gene may be related to prostate cancer progression and drug resistance

Researchers say that through modification, P-NET can also help oncologists predict the disease progression and treatment response of other cancers