Interested in reading more?

Photo: Youngblood PLAY IT AGAIN: Replication is the key to reducing inadvertent errors or fraudulent results, maintains USC's W. French Anderson. Last fall, Francis Collins, a prolific and widely respected scientist, retracted all or parts of five papers he had coauthored in the preceding two years. Collins, the director of the National Center for Human Genome Research at the National Institutes of Health-which in January became the National Human Genome Research Institute-apparently was the victim of a dishonest graduate student in his lab from the University of Michigan, where he had been before coming to NIH. According to published interviews with Collins last fall, the student fabricated and misrepresented data (Notebook, The Scientist, Nov. 11, 1996, page 30). Alerted to possible problems by a journal editor who was considering a paper submitted by Collins and the student, Collins checked some results and redid some work, then confronted the student, who allegedly confessed. VULNERABLE P.I.'s: Francis Collins, who retracted all or parts of five papers last fall, is far from the only high-profile researcher stung by allegations of misconduct in his lab. In a "Dear colleague" letter to 100 researchers, Collins acknowledged that the issue raises the question of whether "I as the research mentor was paying sufficient attention to this individual." Of course, Collins is not the only high-profile researcher stung by misconduct in his lab. As senior researchers get farther from the bench and increase the number of graduate students and postdoctoral fellows they supervise, are they vulnerable to the infrequent dishonest scientist in their lab? Are these principal investigators (P.I.'s) putting too much pressure on junior scientists, whose fates they control, to get certain results? Are they failing to properly scrutinize junior scientists in their labs or to see that results are sufficiently checked? The Scientist interviewed prominent P.I.'s, postdocs, and others to get their views on these questions. All agreed that the sort of committed cheater who apparently victimized Collins would not necessarily be caught by stricter supervision. "It is very difficult to completely avoid a skillful and dedicated fraud," says the Massachusetts Institute of Technology's David Baltimore, when asked to comment on what befell Collins. Baltimore, who presides over a laboratory of 31 people, has a reputation for being on top of what is going on in his large laboratory, but found himself embroiled in a 10-year misconduct proceeding as a result of his collaboration with Thereza Imanishi-Kari of Tufts University. Imanishi-Kari was exonerated by a Department of Health and Human Services appeals board last year (B. Goodman, The Scientist, Aug. 19, 1996, page 1). Citing advice from lawyers and the continuing investigation of the student, Collins would not agree to an interview. In a brief E-mail message, he asserted that "the problem with the student in my lab was not a consequence of a busy mentor. It was the clever and intentional deception of numerous individuals by a student determined to fabricate data." Paul Liu, Collins's former postdoc and now an independent investigator at the Genome Institute, notes that Collins was in the lab every week. About the student, he says, "This guy was very clever and gave a complete story, thinking through all the loopholes. The methods section [in his papers] is very detailed. I didn't doubt that he did the experiments." Supervision, or lack of it, seems to be an issue mainly in large laboratories, although it is not clear that the problem of cheating is worse in large labs. Certainly there are noteworthy examples of fraudulent science occurring in small labs. Baltimore suggests that interactions in a large laboratory among grad students, postdocs, the lab head, and others are "part of the safeguard" against fraud as well as sloppy science. Indeed, many of those who talked to The Scientist agreed that the atmosphere in the laboratory is far more important to encouraging good science and stifling dishonesty than the size of the laboratory. A lab-large or small-with good lines of communication, little intralab competition, and an accessible, engaged mentor is not, in this view, a likely venue for successful fraud. Photo: EW & WL Branson OUTSIDE INFLUENCE: Funding crunches can lead to "soft results that get published," contends Cecil Fox of Molecular Histology Labs. A corrosive atmosphere may be more likely in a large lab than a small one, however. Cecil Fox, a former NIH scientist who founded and runs Gaithersburg, Md.-based Molecular Histology Labs Inc., a contract research organization, sees problems in "large labs where you have a great kahuna who doesn't work in the lab, doesn't know how to work at the bench, but goes in and says, 'We need some results to get funded.' And sure enough, someone in the lab produces soft results that get published." Fox and others say soft or sloppy science is far more common than fraudulent science-and more damaging. For example, a researcher feeling pressured to get a result to enhance job prospects or a grant application might report one experiment with conclusive results and ignore others with ambiguous or negative results. FAIL-SAFES? University of Illinois’s C.K. Gunsalus asks if there are enough "checks and balances" to prevent fraud in the lab As the administrator who directs the institutional response to misconduct allegations at the University of Illinois, associate provost C.K. Gunsalus has ample experience dealing with such cases. She maintains that the way a lab is run is a key element in whether a lab member could commit research fraud. Are there, she asks, enough "checks and balances so that a smart, committed liar has to work pretty hard to get away with something? Do people regularly challenge each other often enough as part of their own internal rigor to come across such a liar?" Checks and balances start with a P.I. who, even if not doing bench work, is a presence in the lab, observing, answering questions, and providing advice. Virologist Gail Wertz of the University of Alabama, Birmingham, supervises a laboratory of about 10 people. She says, "I walk around the lab and try to make contact with everyone every day." She is especially careful to talk to her graduate students, inquire about their controls, and see their data. Photo: Robert Reichart KEEPING IN TOUCH: Rockefeller’s James Darnell, center, visits with his graduate students and postdocs "every day or every week." More senior researchers, with heavy travel and administrative commitments, may not be as readily available. James Darnell, the Vincent Astor Professor at Rockefeller University, supervises about 10 people in his lab. "I haven't been shaking test tubes for 20 years," he acknowledges, but "I supervise [graduate students' and postdocs'] experiments and I don't think they could pull the wool over my eyes. I see them every day or every week." James Wilson at the University of Pennsylvania Medical Center has more than 20 people, mostly postdocs, in his lab. He also has heavy administrative responsibilities as director of Penn's Institute for Human Gene Therapy. "You can't look over every experiment," he contends. "You have to challenge people in your lab, and they challenge you. It's important to get a sense of integrity of the people in your lab." In many large labs, the immediate day-to-day supervision is carried out by senior postdocs, senior technicians, or even other faculty. Rita Colwell, president of the University of Maryland Biotechnology Institute, continues to run a lab that varies between 10 and 15 people. Two research assistant professors, who have worked for her for a long time, are responsible for daily operations, she says. In the Darnell lab at Rockefeller, the immediate supervision is by postdocs, according to Daniel Kessler, an assistant professor of cell and developmental biology at the University of Pennsylvania. Kessler was a graduate student in Darnell's lab, then a postdoc in the slightly smaller lab of developmental biologist Douglas Melton at Harvard University. Kessler says the key to the success of Darnell's lab was his ability to attract high-quality postdocs. "Most of my hands-on technical training came from one postdoc," recalls Kessler. "That was an extremely good situation for me." Kessler was lucky. Many postdocs, especially at the end of their training, focus on job hunting. Full of anxiety, they may not be eager to tutor less experienced scientists. Most laboratories, large and small, hold regular lab meetings at which members present new data, discuss technical difficulties, and plan experiments. The lab meeting, according to Paul Wilson, a senior postdoc in developmental biologist Ali Hemmati-Brivanlou's lab at Rockefeller University, "is one of the best constraints on fraud and bad science. In lab meetings, senior postdocs have a responsibility to ask penetrating questions." Baltimore says, "I have a meeting with the lab every week and go over data. I make an effort to have people conscious of what each person is doing-not to stop fraud, but because it keeps people from stepping on toes and keeps them aware of opportunities." HANDS-ON RESEARCH: Actively working in the lab gives Alabama virologist Gail Wertz "a feeling for what kind of results" to expect. Whether at lab meetings or individually, many lab heads say it is critically important to see and discuss raw data with junior scientists. Alabama's Wertz, for one, maintains that tables and charts of compiled data are not enough: "Every person in every lab, no matter what its size, has got to pay attention to primary data." CULTURAL AWARENESS: Sometimes students from different cultures "are less willing to show" problematic data, notes Scripps Research Institute’s Roger Beachy. Typically, the more experience a junior scientist has, the less scrutiny he or she will receive from a supervisor. Roger Beachy, head of the division of plant biology of the Scripps Research Institute in La Jolla, Calif., says he tries to see raw data from his graduate students every week or so, but may see raw data only every few months, or not at all, from postdocs. "I ask to see raw data more on projects I really care about," he explains. Beachy supervises two labs totaling about 30 members. He raises an issue that several lab heads and postdocs would discuss only gingerly: whether students and postdocs from different cultures need stricter supervision, either because they have been taught to handle data differently or because their permission to remain in the United States hinges on success in the lab. In Beachy's case, some of his foreign students and postdocs want him to see only their good data, he says, and "are less willing to show me inconclusive data or data they think might be mistaken." He acknowledges that seeing just clean data or just summaries may make it easier for someone to deceive him. When he senses that someone is pushing positive data and ignoring negative or inconclusive data, he questions the person more thoroughly, he reports. Speaking of arms control with the Soviet Union, former President Ronald Reagan said, "trust, but verify." In science the stated ethic is "trust, but replicate." True replication, however, is an ideal that is not often met in practice. Researchers eager to publish quickly do not want to repeat a long or complicated experiment. And there is usually no funding available just to repeat the result from a competing lab. Nevertheless, some lab heads say they insist on repeating work in-house before publishing, especially if the result is potentially important. "Any time you have an exciting finding, you have to be very cautious," states W. French Anderson, director of the Gene Therapy Laboratories at the University of Southern California School of Medicine. "You have to have replication. If two people in the lab have independently gotten the same results, the chances of inadvertent error or fraudulent results are greatly reduced." Virologist Wertz says she never lets an experiment go out that has been done only once. Instead, she wants to see experiments repeated several times. "I'll know from looking at the gels that . . . it's representative data," she explains. Usually, she has the primary author repeat the work, though sometimes, if it's controversial, she or someone in the lab might do it. More common than simply repeating experiments is for someone else in the lab to extend interesting results by a colleague. "We do things which build on previous results," says Baltimore. "That's a positive thing to do and checks the original result. If you literally said to someone that you wanted to repeat the experiment . . . it indicates a lack of trust and breaks down the human relations [of the lab]. You might as well kiss your lab good-bye if people think you don't trust them." Photo: Karen Young Kreeger LIMITED SUPERVISION: With a large lab and heavy administrative responsibilities, Penn's James Wilson "can’t look over every experiment." With administrative and fund-raising responsibilities multiplying, lab heads rarely continue hands-on research. Wertz, who recently completed a term as president of the American Society for Virology, keeps her lab to about 10 people to make it easier for her to remain active at the bench. She says that continuing her own research gives her "a good feel for what can be done with a project. I know whether it's going to work or not, and have a feeling for what kind of results" to expect. Some P.I.'s keep a small project going. Anderson maintains that working in the lab enables him to "find out who doesn't clean up the centrifuge rotors after they use it, and also hear lab scuttlebutt. If you are working in the lab at 11 p.m., you learn things you don't learn in the office." His students, he says, occasionally bring visitors to the lab to "see me working in the hood." Harvard's Melton, a developmental biologist, reserved one lab chore for himself, according to two of his former postdocs. He would do the microinjections of embryos that were a staple technique of his lab. Many postdocs learned to do this themselves, says Kessler, one of Melton's former postdocs. But the advantage of the arrangement, he notes, is that "it guaranteed time when you were sitting at the bench with him and you could talk to him about things in the lab." Some lab heads say being active at the bench is neither possible nor desirable in a large laboratory. "Where people need me most is not doing my own experiments," asserts Penn's Wilson, "but being available to advise on their experiments." While lab heads' participation in the intermediate steps of putting out a paper-designing the experiments, performing them, deciding issues of data analysis and presentation-varies, one job requires the utmost attention. In this era of tight competition for funding, lab heads are finding that they have greater responsibilities than ever before in teaching proper scientific conduct to the researchers they supervise. Billy Goodman, a freelance science writer based in Montclair, N.J., can be reached online at goodmanb@idt.net.

Top P.I.'s Say That Their Presence In Labs Acts As Safeguard Against Fraud, Sloppiness

May 11, 1997

Interested in reading more?

Become a Member of