The study was led by Jingwu Zang, senior vice president and head of the R&D center's department of neuroimmunology in Shanghai. The paper's top author was GSK researcher Xuebin Liu.
Zang, who had held the job since 2007, has been fired; Liu resigned in light of questions about the data.
Their paper claimed that research had "found data suggesting that the signaling molecule interleukin-7 caused a subset of T-cells .... taken from people with multiple sclerosis to multiply," according to the Nature Medicine blog Spoonful of Medicine. The blog pointed out that the finding corresponded with research that some people might be at higher risk of MS as a result of genetic differences in that molecule's cell receptor.
The paper had remained unquestioned until "several weeks ago," when a tipster contacted GSK and suggested that there were problems with the study, GSK spokeswoman Melinda Stubbee told CNBC.
"It was an email," Stubbee said. "The information was given to us anonymously."
GSK's subsequent probe discovered that the data that purportedly was based on blood samples from people with MS had either come from "normal, healthy donor samples," or from sources that "can't be documented at all," Stubbee said.
"The integrity of our research is critical to our work, and when these allegations came to light we immediately contacted the journal to tell them we were taking the charges seriously and would be investigating thoroughly," GSK said in its statement.
"Regretfully, our investigation has established that certain data in the paper was indeed misrepresented," GSK said.The U.K.-based company noted that "the published study was from pre-clinical, early-stage research and did not directly involve patients."
(Read More: China Downturn Might Be Most Drawn Out)
Zang told Spoonful of Medicine that he was "shocked" and "disappointed" by his termination. "I was never involved in any data fabrication," he told the blog. "I'm not saying I'm free of any responsibility. ... I'm a senior author. For that I should accept certain responsibility ... but not as currently accused."
Liu told the blog that the disputed data were incorrectly described in the study because there was "a rush" to complete it, and he had forgot to update the document when it was submitted. He also said that Zang did not realize the errors or participate in analyzing the data.
—By CNBC's Dan Mangan. Follow him on Twitter