Saturday, October 29, 2005
Scientists Notice a Pattern (i.e., Design) to Catch Fraud Science
Michael Borowitz, at the Johns Hopkins University School of Medicine in Baltimore, Maryland, says: "The shapes of the major clusters are often similar but in any system there is noise, and those noisy dots are in the same place too. That's hard to explain by biology. It is very difficult for me to believe that these were independent experiments." Borowitz is an expert in interpreting flow cytometry graphs, which he regularly uses to identity abnormal populations of cells in the blood and bone marrow of leukaemia patients.
Three other experts contacted, including Paul Robinson, a professor of immunopharmacology and biomedical engineering and Director of the Flow Cytometry Labs at Purdue University in West Lafayette, say that the graphs appear concerningly alike.
Robinson emphasises that it is impossible to prove that the data underlying the supposedly different graphs is the same without seeing the raw data from which they were generated. But he and four members of his lab have conducted a simple experiment to gauge the likelihood that data taken from different mice would produce such similar graphs.
They ran a computer model that produced graphs from two sets of 93,000 cells and reduced the resolution to below that of Van Parijs’ published figures. Their conclusion, detailed in a brief report seen by New Scientist, is that even graphs produced from samples taken from the same animal are unlikely to look the same, and that outlying data points are unlikely to appear in the same positions by random chance.