The following article is from the January 11, 2006 Boston Globe, and was written by Gareth Cook. The article addresses a problem in scientific research where laboratory results are manipulated by use of inexpensive digital technology. With pharmaceuticals and bio-med being such potentially lucrative businesses, have there been any significant changes to peer review procedures?
Technology seen abetting manipulation of research
By Gareth Cook, Globe Staff | January 11, 2006
An explosion of new digital image technology has left many of the world's top biology journals vulnerable to fraud, scientists say.
The same advances that have given consumers inexpensive digital cameras -- and software to easily copy, crop, or alter an image with a few clicks -- have also proven a temptation for unscrupulous researchers. Federal science fraud investigations involving questionable images have grown from 2.5 percent of the cases in 1989-90 to 40.4 percent in 2003-04, according to the federal Office of Research Integrity, which investigates scientific misconduct. And in just the past few months, there have been two high-profile cases -- those of discredited South Korean scientist Hwang Woo Suk and the fired MIT biologist Luk Van Parijs -- that involve duplicated images.
The manipulation of images of cells can give a scientist a way to convince colleagues that experiments were successful when they were failures. By merely changing images on a laptop, scientists can earn acclaim, win lucrative research grants, and advance their academic career.
For decades, many scientists and journal editors have assumed that cases of scientific fraud are extremely rare because few come to light each year. The scientific community, they believed, had adequate checks against fabrication, such as the practice of other scientists repeating experiments after they're published.
But an innovative program at one leading biology journal challenges these assumptions. In September 2002, the Journal of Cell Biology began examining all the images in papers it had tentatively accepted but not yet published, using the software program Photoshop. The journal has had to reject 1 percent of these papers because authors manipulated images in a seriously misleading way, adding or subtracting elements that changed an experiment's results, according to Mike Rossner, the journal's managing editor. Photoshop makes it easy to manipulate an image, but it also allows someone to adjust an image and look for signs of manipulation.
''We can catch a lot of stuff before it comes out," said Rossner. ''For me, I consider this a matter of personal responsibility as an editor."
Rossner has been warning editors at other journals of the scope of the problem, and what can be done about it, and some are listening. The journal Science, which published both papers in which Hwang falsely claimed to have cloned human embryonic stem cells, announced yesterday that it would start using new safeguards to detect altered images this month. The journal will use the techniques developed by Rossner, according to Science editor in chief Donald Kennedy, but he said he didn't think these procedures would have caught the fraud in the Hwang case.
None of the other biology journals contacted by the Globe, including Nature, Cell, PLoS Biology, and Proceedings of the National Academy of Sciences, has such a system for checking images. But all said they were now reviewing their procedures for handling images, and may change their policies.
Images play an important role in documenting the results of biology experiments, and some of them can be especially prone to fraud, scientists say. For example, biologists commonly include in their papers images of a gel used in tests of characteristics of cells, with black bands indicating the presence of DNA, or particular proteins. These bands are easy to fake to change the results of an experiment, and this is one of the most common types of misconduct Rossner finds.
At Science, which publishes a full range of scientific papers, the new screening will focus on papers in biology, which tends to use a type of image vulnerable to manipulation, Kennedy said. He and other journal editors said that they were shocked to hear that 1 percent of the papers accepted in the Journal of Cell Biology had to be rejected because of apparently fraudulent images.
Rossner said that his involvement in detecting fraud came entirely by accident. Several years ago, the journal started requiring its authors to submit their papers, including all of the images, electronically. One author had submitted an image in the wrong format, so Rossner was working to fix it. As he did this, he noticed a box around part of an image, where the background didn't quite match -- a sign that someone had altered the image.
''I said, 'Oh, boy,' " Rossner remembered.
Eventually, he set up a list of checks that a production person at the journal does for every image to be published. For example, he said, they will take the image and enlarge it, then increase the contrast of the image to search for signs that the background is not consistent. They also look at magnified images to see whether any are duplicates. On average, the checking takes about 30 minutes per manuscript.
Since the system has been in place, about 1,300 papers have been reviewed. In about 25 percent of the cases, there is at least one image that has been changed so much that the editors think it is not an accurate representation of the original, and the authors are asked to resubmit the photo. For example, it is common for authors to increase the contrast on the image so that some of the fainter lines disappear, making the image look cleaner.
In these cases, the manipulations do not affect the conclusions of the paper, but violate the journal's strict policy on image manipulation.
In 13 cases, Rossner said, the screening has found what the journal considers to be fraudulent manipulations and has rejected the paper. These include duplicating cells, and adding bands on gels. He said that the journal rejects papers only if he and three other editors agree that the misdeed is that serious. Rossner has been consulting with Hany Farid, a computer scientist at Dartmouth College, who is developing computer software that will automatically check digital images for signs of fraud, including portions of images that are duplicated. He plans to make the software available for free.
Hwang published two papers claiming to have created cloned embryonic stem cells, and both had duplicated images -- one used the same image more than once, and another used images that had appeared in other journals, which would have been much more difficult to catch. Science said yesterday it would retract both papers, after the release of an investigation in South Korea showing that Hwang had invented his findings.
In a separate case, the Massachusetts Institute of Technology fired Van Parijs for research fraud last year. MIT did not identify the specific problem, but several papers he wrote before coming to MIT contain what appear to be duplicated images.
Part of the problem, scientists said, is the speed with which the technology has changed, meaning that journals, and the scientific community, have not had much time to think through the implications. Now, a single scientist can acquire data from an experiment, analyze it, and generate a final image for publication by computer, without anyone else being a part of the process, eliminating a check on fraud. There is also something of a generation gap -- the people who head labs and train the next generation in what is acceptable are often not as familiar with the full power -- and potential for abuse -- that the new tools provide, according to Emilie Marcus, editor of Cell.
Rossner said that while some editors have been enthusiastic to hear about what can be done to detect fraud, others are complacent. Twice, he said, he has seen papers rejected by his journal for fraudulent images appear later in another journal, with the same, problematic figure. Once, he said, he wrote to the editor of the journal to tell him about a band whose intensity had been manipulated. Rossner got a note back; the editor didn't see any problem.
Gareth Cook can be reached at firstname.lastname@example.org