Scientists have found a way to predict the prognoses of patients with pancreatic cancer based on analysis of gene expression patterns within each patient’s tumor by using a variation of the algorithm Google’s search engines use to scour the Internet for relevant websites, according to a study published today in PLoS Computational Biology.
High-volume genetic techniques have allowed scientists to identify thousands of genes that may be differentially expressed in tumor cells. Many have hoped that this capability would enable researchers to better predict a patient’s prognosis or better match treatments to a patient’s individual tumor. But to make these expression signatures a useful clinical tool, strategies are needed that identify those genes that predict a tumor’s aggressiveness and susceptibility to therapy while weeding out the many other genes that are differentially expressed in tumors but are not clinically relevant.
A team of researchers from Germany and Switzerland tackled this challenge by using a strategy based on Google’s PageRank search strategy to search through gene expression profiles and identify the most relevant. To identify the web pages most relevant to a particular search term, PageRank finds web pages on related topics and then ranks them based on which of these web pages are most heavily linked to by other websites. The NetRank algorithm used by the researchers similarly uses information on biological interactions between genes’ products to decide which of the genes are likely to be most relevant to a patient’s clinical outcomes. The scientists used NetRank to analyze tumor samples from 30 patients with pancreatic cancer and found they were able to predict whether patients had good or bad prognoses with 72% accuracy. They then validated their findings with a separate set of tumor samples from 412 patients with pancreatic cancer.
With further validation, such a strategy may be used to determine which patients should receive chemotherapy and which patients may survive longer without such treatment, according to the authors.
“Reliable prediction of survival and response to therapy based on molecular markers bears a great potential to improve and personalize patient therapies in the future,” the authors state.