Algorithm Ranks Genes’ Likelihood to Contribute to Cancer

The search for cancer genes is increasingly a matter of molecular “To Tell the Truth,” as scientists seek to distinguish genes actually involved in the disease from those that are imposters. Powerful gene-scanning technology often reveals hundreds or thousands of genetic irregularities in tumor tissue, making it critical for investigators to winnow out the true culprits.

In a new study in the April 8 issue of the journal Cancer Cell, researchers at Dana-Farber Cancer Institute and Memorial Sloan-Kettering Cancer Center describe a new algorithm for ranking abnormal genes according to their likelihood of contributing to a cancer. And they show that a gene identified by the algorithm as a likely restrainer of tumor growth does indeed play that role in a common type of brain cancer, and is not a mere “bystander” to another restrainer gene.

“As the Human Cancer Genome Project begins to map the genetic alterations in different kinds of cancer, we need to be able to discriminate between alterations that truly are relevant to the disease and those that are not,” says the paper’s senior author, Lynda Chin, MD, of Dana-Farber. “Using a new algorithm developed in collaboration with Cameron Brennan, MD, of Memorial Sloan-Kettering, we were able to identify genes with too many or too few copies in cancer cells.”

The algorithm can be used to analyze similar genomic data generated by The Cancer Genome Atlas pilot project, a federally-led effort to explore genomic changes involved in human cancer. It is also being submitted to BioConductor, a collection of open-source computational tools for free download by researchers.

The Cancer Cell paper describes how the algorithm was applied to a study of glioblastoma, the most common form of brain cancer in adults and one of the most difficult malignancies to treat successfully. Chin and her colleagues performed high-resolution genomic scans of glioblastoma tumor samples and cell lines, and the results showed dozens of gene copy alterations, some of which had already been linked to the disease and some of which had not. To determine which of these “suspect” alterations were most likely to contribute to cancer, the researchers ran the results through their new algorithm.

One of the highest-scoring abnormalities - meaning it had a high potential to cause rampant cell growth - involved a gene known as p18INK4C. Researchers knew that one of p18INK4C’s cousins, a tumor-restraining or – suppressing – gene called p16INK4A, is missing in a majority of glioblastoma cases. p18INK4C itself, however, wasn’t previously known to be missing in the disease. Although the two genes are thought to have similar functions, investigators suspected the co-disappearance was more than a coincidence, and that the loss of p18INK4C plays a role in glioblastoma.

Based on the algorithmic analysis, lead author Ruprecht Wiedemeyer, PhD, a postdoctoral fellow in Chin’s lab, went in search of a connection between p16INK4A and p18INK4C that can explain their joint disappearance. It turns out that the loss of p16INK4A triggers a shutdown of a “pathway” (a series of interconnected genes) called RB. That, in turn, causes cell proliferation and a giant step toward cancer. At that point, p18INK4C steps in as a backup system, pulling the reins on the hectic cell growth permitted by the loss of p16INK4A. If p18INK4C is lost, it’s as though the emergency brake on growth is gone.