Skin Cancer: Algorithms to Predict Immunotherapy Response

Melanoma is the deadliest of all skin cancer types. Although immunotherapy can be used in the fight against melanoma, no efficient clinical test can determine ahead of time whether a patient will benefit from this treatment that works by stimulating their immune system.

John Stagg, a researcher at the CHUM Research Centre, Ian Watson from the Goodman Cancer Research Centre and Hamed Najafabadi from the McGill Genome Centre intend to change this state of affairs by using predictive algorithms.

Convinced of the importance of their research project, Genome Quebec, Oncopole and IVADO awarded them a 2 year $300,000 grant as part of the Omics Data Against Cancer competition.

The three researchers’ project involves developing algorithms to predict immunotherapy response in patients with metastatic melanoma.

Clinicians are currently able to sequence their cancer patients’ genome to determine the best therapeutic options. However, this genomic information is rarely used to orient patients towards immunotherapy.

Why? Simply because scientists still don’t know which DNA sequences to use to predict the response to this type of treatment.

To address this, John Stagg and his colleagues will rely on a branch of artificial intelligence known as transfer learning to try to predict the gene expression signature characteristic of a good response to immunotherapy.

First, the algorithms developed by the team will be trained on public data from The Cancer Genome Atlas Program before being tested on a cohort of 300 Montreal patients with metastatic melanoma.

If this approach proves successful in the coming years, it will probably be possible to transpose it to other cancers that can be treated by immunotherapy, especially lung cancer.

> Official communication of November 5