
Some studies show the expression of proteins such as caspase-3 (pictured), in cell lines that should not be able to produce them. Credit: Dr Mark J. Winter/Science Photo Library
A scientific sleuth has identified more than 220 research papers that describe impossible patterns of protein expression in genetically altered cell lines. In these studies, particular proteins show up in imaging experiments, even though the cells do not have the gene necessary to make them.
Journals are now taking action to clean up the literature by retracting or correcting the work, and researchers warn that there are probably many more affected studies.
“This problem will not go away,” says Thierry Soussi, a cancer geneticist at Uppsala University in Sweden, who notes that there are several other examples of the issue in now-retracted papers that were not included in the sleuth’s analysis. “This is only the tip of a gigantic iceberg.”
Modified cells
Cell lines are groups of cells derived from a single origin cell and cultured for use in experiments. Some cancer cell lines have been genetically modified with mutations and deletions — often to remove genes that usually control cell division. This means that the cultures continually divide and grow, giving researchers the chance to study tumour progression and test potential treatments.
The MCF-7 cell line is one of these. It lacks the gene for caspase-3 protein, which is involved in cell death. In 2009, researchers warned1 that some studies that report working with MCF-7 have mysteriously observed caspase-3 expression in western blots, an imaging technique that identifies proteins based on their molecular weight.

Sleuth identifies dozens of studies that used the wrong antibody
In the most recent analysis, published as a preprint on 21 October to the data repository Zenodo2, Sholto David, a sleuth based in Arbroath, UK, finds that this problem persists and extends to other cell lines.
David compiled a list of 20 commonly used cell lines with genetic mutations or deletions that leave cells unable to express specific proteins. He then scoured the literature using Google text and image searches to identify examples of western-blot images that unexpectedly identify these proteins. He also searched a collection of 140,000 articles that he had previously downloaded and screened for image-integrity issues and other problems, most of which were biomedicine papers.
He found 230 reports of phantom proteins across 221 papers published since 2004. The most common errors were observations of caspase-3 expression in the MCF-7 cell line, affecting 49 papers; expression of the cell-death protein BAX in the Jurkat cell line in 47 papers; and expression of the signalling protein STAT3 in the PC-3 cell line in 35 papers.
Mistakes or misconduct?
David suggests several explanations for these impossible protein-expression patterns. They might be the result of honest errors: researchers could have misidentified the cell lines or contaminated them with others. Or the antibodies used to identify proteins in the western-blot experiments could have been non-specific and bound to another protein with the same molecular weight. In fact, David says that he stumbled across the anomalous proteins while working on a project that flagged hundreds of studies that used the wrong antibodies in experiments.
Another possibility is that some of the papers are the product of papermills, companies that fabricate and sell scientific manuscripts. David says that some papers in the cache contain the hallmarks of being produced by such outfits, including duplicated images or text templates that are similar to other papers. The proportion of the papers that fall into this category is not clear, he says.
Nature has contacted several authors of papers containing anomalies to ask for comment, but none responded.

Science sleuths raise concerns about scores of bioengineering papers