An Institute of Medicine (IOM) study on “omics” tests at Duke University has found that such tests are susceptible to errors. The initial investigation began in 2009, when biostatisticians at MD Anderson Cancer Center voiced concern over the published claims of two Duke scientists that molecular patterns in tumor cells could indicate patients’ reactions to different chemotherapy treatments. The IOM deemed “overfitting” a key cause of the errors, as researchers relied on a small number of samples to establish patterns in hundreds of biomolecules. Duke also had relationships with the test makers, and a prominent professor at the university spearheaded the research. The IOM suggested a series of steps to prevent such issues. These steps include administering the test at another institute on blinded samples; sharing information in papers among researchers; and making sure to “lock down” the computational model to prevent alteration. Institutions should also confer with the FDA early in the process. Duke now faces lawsuits from the research and is likely to also retract 27 papers.

Source: Science Insider