to BioTechniques free email alert service to receive content updates.
Google Goes Cancer

05/18/2012
Diana Gitig, PhD

Taking Google’s PageRank algorithm as their model, researchers in Germany aim to rank gene expression networks to predict clinical outcomes of cancer patients.

Bookmark and Share

Two patients with the same type of cancer can respond differently to the same course of treatment, but we have not yet been able to determine why, or predict who will respond well to a particular drug. Although many studies have tried to correlate gene expression to clinical outcomes, there is little agreement between those studies, and they are not particularly accurate.

Taking Google’s PageRank algorithm as their model, researchers in Germany aim to rank gene expression networks to predict clinical outcomes of cancer patients. Source: PLoS Computational Biology





As a result, Christof Winter, a postdoc at the Dresden University of Technology, was asked by colleagues treating pancreatic cancer patients to develop a computer system identifying clear biomarkers. “We had often used networks to make sense of gene expression data sets, but we wanted a more automatic, objective, reproducible prognostic indicator. We first experimented with our own algorithm, but then we realized that what we needed was already there, in PageRank,” says Winter.

Much like Google’s PageRank algorithm chooses the most relevant documents based on the number of hyperlinks it contains, their system—NetRank—evaluates the biological interaction between genes’ products to assess their ability to predict patient survival. Their approach was published May 17 in PLoS Computational Biology (1).

The researchers looked at gene expression profiles from 30 deceased patients who had surgery for pancreatic cancer from 1996–2007. Each patient was assigned to a good prognosis group if they outlasted the median survival time of 17.5 months, and a poor prognosis group if they did not. The expression of 8000 genes was then evaluated to see if any varied significantly between the two groups.

Because genes can be linked to each other in a number of ways, Winter and colleagues tested three different networks for NetRank: transcription factor–target gene relationships, protein-protein interactions, and gene co-expression. Transcription factor–target gene relationships gave the highest predictive accuracy, and using that network, they identified seven genes that predicted survival in their sample groups.

To validate their findings, they analyzed the expression of these genes in 412 additional patients who had surgery for pancreatic cancer from 1991–2008. Because immunohistochemistry can be easily performed in a clinical setting, they used that technique to assay expression of their seven genes in each patient’s tumor.

For patients who had received adjuvant therapy, six of the seven genes were sufficient to achieve the best predictive accuracy; for patients who had not received such therapy, only five genes could do it. NetRank improved the predictive value over extant clinical parameters including tumor size, regional lymph nodes, distant metastasis, histological grade, and residual tumor by an additional 9% in patients who were treated with adjuvant and 6% in those who were not.

Proteins and genes do not act in a vacuum but rather form complex networks of interactions. NetRank assesses the relevance of a gene as a prognostic indicator not only by correlating its expression to survival, but by factoring in the expression of its interaction partners as well, making it more accurate than methods used to date.

References

  1. Winter, C., G. Kristiansen, S. Kersting, J. Roy, D. Aust, T. Knösel, P. Rümmele, B. Jahnke, V. Hentrich, F. Rückert, M. Niedergethmann, W. Weichert, M. Bahra, H. J. Schlitt, U. Settmacher, H. Friess, M. Büchler, H.-D. Saeger, M. Schroeder, C. Pilarsky, and R. Grützmann. 2012. Google goes cancer: Improving outcome prediction for cancer patients by Network-Based ranking of marker genes. PLoS Comput Biol 8(5):e1002511+.



In order to deliver a personalised, responsive service and to improve the site, we remember and store information about how you use it. This is done using simple text files called cookies which sit on your computer.

By continuing to use this site and access its features, you are consenting to our use of cookies. To find out more about the way Bio Techniques uses cookies please go to our Cookie Policy page.