Although the human genome has been mapped, many questions remain about how genes are regulated, how they interact with one another, and what function some genes serve. A new tool developed by researchers at the University of Illinois at Urbana-Champaign and the Massachusetts Institute of Technology distills the huge amount of genomic data into gene networks that can point to the function of genes, highlighting relationships between genes and offering insights into disease, treatment and gene analogs across species.
The researchers described the tool, called Mashup, in the journal Cell Systems.
“A single gene cannot do the job. You need to have a lot of genes in pathways that work together through different types of interactions, and gene functions are determined by these interactions,” said Illinois computer science assistant professor Jian Peng, who led the study. “You can think of genes as being connected in a social network, where each gene is a person. There are different relationships between any two persons. They could be friends, they could be family, they could share similar interests, and the different interactions between them define who they are in the society.”
Mashup uses machine learning tools to computationally integrate multiple gene interaction networks. Each gene is represented by compact patterns, or topologies, so that researchers can more easily analyze its function. By compressing the landscape of gene interactions, the main fingerprints of a gene pathway are revealed as the background noise of genomic data is stripped away, Peng said.
The researchers compared Mashup with other gene analysis methods and found it to be faster and more accurate at predicting a gene’s function and at identifying genes of similar functions in other species. This can yield insights into human disease in cases where pathways in other organisms are better documented than in humans. For example, Peng and other collaborators are working to compare certain genes in yeast with the human genes that contribute to neurodegenerative diseases such as Parkinson’s or Alzheimer’s.
“Where genes were conserved during evolution, we easily can map yeast genes back to human genes by comparing DNA sequences to find functionally similar gene pathways,” Peng said. “However, a lot of yeast genes don’t look much like human genes in sequence. Mashup is a way to tell whether they come from similar functional pathways by comparing the network topology. You can think of it like, if my friends are similar to your friends, we probably have common interests. In this way, Mashup can identify key pathways in neurodegenerative diseases by mapping discoveries from yeast experiments onto the human genome.”
Mashup also could be a tool for studying and treating cancer, both by looking at how genetic interactions change in cancer cells and in predicting drug efficacy. The researchers tested Mashup’s ability to predict whether certain cancer treatments would be effective for a particular tumor by predicting the efficacy of different drugs over hundreds of cancer cell lines and comparing the results to their known profiles from the Cancer Genome Project. They found the tool accurately predicted a large number of drugs’ efficacy against specific cancer types.
Next, the researchers will test Mashup with clinical samples through a collaboration with Mayo Clinic, Peng said.
“After doctors get genomic data from a patient’s tumor sample, we use our algorithms to predict which genes are essential to the tumor. We will be able to recommend treatments by finding drugs that target essential genes for that tumor, or determining whether a drug combination would be more effective,” he said.
The National Institutes of Health supported this work. The paper “Compact Integration of Multi-Network Topology for Functional Analysis of Genes” is available online.
Contact Jian Peng, Department of Computer Science, University of Illinois at Urbana-Champaign, 217/300-2825; email: email@example.com.
Writer: Liz Ahlberg Touchstone, Biomedical Sciences Editor, 217/244-1073, firstname.lastname@example.org.
Photo by L. Brian Stauffer