Researchers are using artificial intelligence for lexicon of cancer genes

Researchers at the Danish Cancer Society have, together with international colleagues, developed an information tool, which can identify genes that drive the development of cancer. 'Moonlight' is the name of the tool that someday hopefully can be used for helping find the right treatment for each cancer patient

Let us present: BCL2. It is the name of a gene that plays a role in the development of cancer.

The special thing about BCL2 is that it appears to have two very different roles, depending on the type of cancer it is found in. While in one cancer type, BCL2 is a tumor suppressor, in another cancer type it is a tumor promoter.

This is new knowledge that has emerged with the help of 'Moonlight', a new information tool that, by using lots of data and computing power, is able to identify cancer driver genes and has just been published in the acclaimed scientific journal Nature Communications.


Elena Papaleo

Not least, Moonlight can identify dual role cancer genes, that is, cancer genes that promote cancer progression in one cancer while counteracting cancer development in another cancer. Of the 151 dual role cancer genes, 31 were already known, while 120 were newly discovered. The predictions from Moonlight now have to be tested on real cancer cells in the lab.

-Dual role cancer genes are still fairly new territory for cancer researchers, but there are great prospects. A drug that affects a dual role cancer gene, should only be given to patients where the cancer gene is a driver in the development of cancer. In other patients we could risk exacerbating the disease, says group leader at the Danish Cancer Society's Research Center and expert in bioinformatics Elena Papaleo, who developed the Moonlight idea together with Antonio Colaprico, researcher at the University of Miami.

They gathered an international group consisting of 17 researchers from research institutes in 8 countries. To them, the presentation of Moonlight is the culmination of three years of work, and the work continues:

-We lack new and visionary tools to understand the complexity behind the genomic profiling of cancer. The goal of Moonlight is to create an information tool that can show, for each cancer type, which genes drive the development of cancer, and in the long term use that knowledge to provide the best possible cancer treatment for each patient, says Elena Papaleo.

Data and algorithms

Similar information tools have been made in the past that focus on gene mutations, she says. The idea of focusing on cancer driver genes and dual role cancer genes is new. This can be done by combining a lot of different information from different sources. For example, the level of the gene in the cancer cell, how it changes the molecules in the cell, and other information.

 The Moonlight algorithms mainly build on data from the American Cancer Genome Atlas (TCGA). The conclusions have then been validated by comparing with data from the Cancer Cell Line Encyclopedia, which also contains data on cancer genes.

For example, based on the information in The Cancer Genome Atlas, Moonlight identifies BCL2 as a dual role cancer gene, and the researchers validate the result by comparing with the other set of data.

 -We found a good agreement between the two data sets, which confirms the value of Moonlight's algorithm, says Elena Papaleo.

The BCL2 gene is tested in the Danish Cancer Society's laboratories

Next step will be testing the results with real cancer cells in the laboratory. For example, the researchers in the Danish Cancer Society Research Center will test cancer cells with the BCL2 gene.

 -There is a lot of work ahead of us in validating the computer's predictions in real cancer cell lines. Also, we hope that scientists around the world will use Moonlight to validate genes of their specific interest.

 - When it comes to BCL2, there are actually drugs that target this gene. Therefore, it is important to know that we are helping one type of patient by using the medication, but in another patient, we will actually promote cancer progression, says Elena Papaleo.

 Moonlight is now freely available online for use by scientists and doctors.

The research is published here:

Colaprico et al. Interpreting pathways to discover cancer driver genes with Moonlight. Nature Communications 11, 69 (2020)

More about the research

The researchers have analyzed data for over 8,000 tumor samples from 18 cancer types, discovering 3,310 oncogenic mediators, which they define as genes controlling biological processes, 151 having dual roles.