Prediction of disease-related genes based on weighted tissue-specific networks by using DNA methylation
In class, we studied the effect of applying PageRank algorithms to rank the influence of a website. However, there are applications of PageRank outside of website ratings. Last month, a paper was published that predicted disease-related genes based on weighted tissue-specific networks by using DNA methylation. The method to do so begins by creating a network graph comprised of protein-protein interactions, called PINs. These protein-protein interactions allow the researchers to identify disease-related genes. One drawback, however, is that false-positives exist within the graph of protein-protein relationships.
To tackle this problem, researchers implemented a version of the PageRank algorithm to isolate the false positives from true positives. First, they incorporated DNA methylation into their PIN, since they found that DNA methylation has been linked to certain cancers and could pinpoint disease-related genes more accurately. Next, they decided to focus more on tissue-specific networks (TSN) instead of PINs as they felt that TSNs related disease-riddled tissues better. These TSNs were then given weights based on DNA methylation and other factors.
Then, to determine whether a gene was a disease carrier, the researchers ran made a thousand instances of the tissue network graph from the gene and ran the PageRank algorithm. Analogously, the graph was set up so that the nodes represented the different tissues and the edges represented tissue networks between each node. This setup inherently gives rise to the idea of PageRank for each node, as nodes with more ingoing edges would be deemed to be more “important” meaning that it would have a higher probability of being a disease carrier. Scientifically, nodes with a PageRank value higher than the average would be considered “distinctiveness” which they classified as a potential candidate for disease-related genes.
To test the correctness of their approach, the researchers ran their experiment on colon cancer and leukemia. Through their experiment procedure, they were able to identify disease related genes more accurately.
http://www.biomedcentral.com/1755-8794/7/S2/S4