The latest flurry of forum messages concerning published data making it possible to nullify the intent of the challenge isn't the only problem facing this challenge. I was going to handle this in the background, having already sent a private message to Matthew Hepburn of DARPA, but now there seems there is little point in holding back. At the request of a senior biological scientist on several of the major genome sequencing projects I have been independently reviewing for more than 10 years the statistical methodology used to analyze microarray data, especially the Affymetrix oligonucleotide chips. The conclusion i came to is that much of the published literature should be discarded in favor of methods that haven't been used by the bioinformatics community. This would hopefully increase the power and replicability of the scientific results. I hope to apply this especially in the understanding of cancer gene expression and metastasis. Almost the entire technical literature on normalization should be discarded, which has already been the subject of forum messages in another thread. But what is worse is that my preliminary analysis of the challenge data appears to show that the subjects already seem to be infected with who knows what even before the actual challenge, as their immune systems already appear to be activated. Apparent chronic viral infections have been the subject of a certain amount of controversy over the years, e.g. chronic Epstein-Barr Virus infection as a proposed etiology for Chronic Fatigue Syndrome (now largely disproved). There is now an ongoing controversy over the physiological significance of MGUS - monoclonal gammopathy of undetermined significance - which may be a marker for the disease multiple myeloma, or it may be nothing: http://www.myelomabeacon.com/resources/mtgs/ash2013/abs/3116/ There is still a lot we don't know about virology and immunology, and the further possibility of unanticipated contamination and cross infection is always present. A famous example was the African Green Monkey Virus and the polio vaccine. Another example of a slightly different sort was the discovery that many cancer cell lines were cross-contaminated with the original HeLa cells. I know that a lot of work has gone into the Duke research program and the Challenge, but unfortunately sometimes one has to be ruthless in striving for the truth. It's the only way we can be really sure.

Created by Alan Robinson robin073
"The conclusion i came to is that much of the published literature should be discarded in favor of methods that haven't been used by the bioinformatics community." can't agree more with you, alan. but what can you do? and that's why there is dream. and that's why i still stay in dream even they hate me. no matter what, the competition phase is usually fair and transparent. because only blind test is where i would do well, instead of reporting p values that are lower than e-100, or an auc of 0.95. clearly the authors of these papers haven't realized that there are only 10**80 atoms in the universe, and the chances that some atoms from universe hit the computer and caused a wrong calculation is way, way higher than that p value. but without reporting such values you cannot even publish well! the biggest problem of our field is now that we are consistently operating on a scale orders beyond the size of the universe!!

Not the only problem page is loading…