Researchers Use Google's Search Algorithms To Fight Cancer 52
MatthewVD writes "German scientists have modified Google's PageRank algorithm to scan tumors and learn more about how cancers progress. PageRank orders results based on how other web pages are connected to them via hyperlinks; the modified algorithm, NetRank, scans how genes and proteins in a cell are similarly connected through a network of interactions with their neighbors. This approach could also yield new therapies to help combat tumors."
Interesting idea, flawed paper (Score:4, Interesting)
The idea is interesting but the paper seems flawed. They integrate data and then look at how genes are positioned in the network with page-rank. The problem is that genes positioning in the network is highly dependent on how studied they are. Therefore, very well studied will get a high "NetRank." Genes known to be predictive of cancer progression are very well studied (lots of fudning in that area). This means the algorithm is basically finding and returning a list of what we already know, and it turns out that what we know is reasonably predictive when you combine 400+ markers.
I'm surprised this made it by peer review without additional experiments to assess the role that this bias plays.
Re:Oh Geez (Score:2, Interesting)
It has nothing to do with anatomical connection between body parts. That's a naive assumption. People have known that cancer spreads from the original tumor to a metastatic site via blood and lymph vessels for a long time. The article is about identifying correlations between changes in expression of genes and proteins, and how these are linked to cancer progression and metastasis. Believe it or not, but every cell in your body is a machine with millions, if not billions, of working parts. One small change in one gene can have dramatic consequences on a number of molecular signaling pathways. Biological, and specifically cancer biological, research has long been focused on deciphering these molecular pathways and identifying the connections between them. So there's already a wealth of data, the key is making sense of it, generating hypotheses, and then testing the hypotheses. So, any additional tools in the arsenal are more than welcome.
Granted, I have no clue if the new application of Google's algorithm is useful or just something flashy to get a publication. But to dismiss the idea behind it as trivial is misguided.