Select Page

Carnegie Mellon's New Algorithm Examines the Cancer Genome by Mickael Marsali

Imagine being handed puzzle to solve that’s made up of literally millions of tiny pieces. Maybe you would sit down and start with the corners, work your way across the edges, and then start piecing together the middle based on like-colors. Now imagine that as you’re trying to piece these things together, certain pieces spontaneously duplicate and others edges keep changing to fit into different parts of the puzzle as you work. Even your most well versed puzzle enthusiast is going to think the task nearly impossible.

What makes cancer so difficult to treat or cure is similar to what makes this puzzle so hard to solve: it’s inherently complex and rapid-changing. What we call cancer is really the uncontrolled division of abnormal cells in the body. When these divisions start to take place, large chunks of DNA made up of millions of chromosomes are duplicated and/or rearranged. It’s hard enough for scientists to find a cure to an illness without it constantly changing it’s expression.

Researchers from Carnegie Mellon University are looking at new ways to solve this puzzle using advanced algorithms to better understand a cancer genome (i.e. the complete set of genetic material present in a cell or organism.)

Their latest algorithm is named “Weaver” after a character in the video game Defense of the Ancients, which is also descriptive of its function: it weaves together disparate pieces of genomic information. It does this by analyzing two major classes of mutations in tumor DNA: (1) copy number variations and aneuploidy (i.e. how chromosomes get duplicated) and (2) structural rearrangements (i.e. changes to the DNA’s order such as pieces being inserted, deleted, duplicated, or moved around.)

Jian Ma, an Associate Professor of Computational Biology, and her team at the Computational Comparative Genomics Lab are heading up the project. As she explains, “The cancer genome is reshuffled and scrambled compared to the normal genome. Weaver’s goal is to interlace genomic pieces and keep things in the right order.”

Using a model called the Markov Random Field, researchers are are able to visualize interrelationships in complex data.

This is the first of it’s kind to be able to simultaneously analyze genomic sequencing data for both types of variations, which provides a more comprehensive view of the cancer genome and how these variations interact.

As Ma explains, “The goal is to look at the sequencing data from the cancer genome and recognize these complex alterations. None of the current structural variant detection methods are specifically designed for genomes with aneuploidy, a hallmark of cancer. Our algorithm can more precisely quantify complex rearrangement structure variants in the context of aneuploidy.”

This more holistic view of the cancer genome in action will be crucial to scientists trying to understand why cancer progresses in the way that it does.

“In cancer, we can see that certain regions are frequently amplified,” continued Ma. “Typically, we don’t know why that amplification is happening. By applying this method, we should at least be able to get a sense that the amplification is due to a specific type of structural variation.”

Moreover, analyzing these genomic interactions as they take place in time could help scientists better understand the full lifecycle of a tumor and provide a more detailed understanding of genomic cause and effect.

“Which variation happened first?” asks Ma. “Did the structural variants or deletion happen before or after the chromosome duplication? This approach could give us a better picture of how these copy number alterations and structural variants are connected.”

So far, Ma and her team have successfully test-driven Weaver in a variety of cancer cell lines (such as HeLa, MCF-7) and samples from the National Institutes of Health’s Cancer Genome Atlas program and published their research in the journal Cell Systems. The next step will be to study specific tumors to determine patterns between different types of cancer could have impacts on their gene expression or phenotype.

Mickael Marsali is a Senior Consultant and founding member of Arterial Capital Management in London. To learn more about his life and career, please visit his professional website.