Imprimir Republish

Good practices

A network for reproducing scientific experiments

Pedro HamdanA group of researchers from the Leopoldo de Meis Institute of Medical Biochemistry at the Federal University of Rio de Janeiro (UFRJ) are organizing a network of laboratories to repeat up to 100 biomedical experiments published in Brazilian scientific articles in order to verify the reproducibility of the results. Named the Brazilian Reproducibility Initiative, the project is funded by the Serrapilheira Institute. The first step will be to select five widely used methods for repeating scientific experiments—potential candidates include the Western blot technique for protein detection, the ELISA antibody test, and RT-PCR for RNA quantification, as well as models based on cell cultures and rodents. “We want to choose established methods that are easy to implement. The next step will be to invite research centers from around the country that are able and interested to participate in the network,” says physician Olavo Amaral, coordinator of the project. He hopes to accredit participating laboratories by the end of the year.

The experiments to be replicated will be chosen at random from published Brazilian scientific articles. “The aim is to select at least 20 experiments related to each method and to determine the proportion of that sample for which the results are reproducible,” explains Amaral. The Serrapilheira Institute will invest R$145,000 in the first year of the project, during which the research replication protocols will be established. After this stage is complete, the UFRJ group hopes to obtain a new funding round to conduct the experiments—the total estimated cost of the project is roughly R$1 million. “The cost is low compared to the total invested by research funding agencies in Brazil. It is important that we improve the quality control mechanisms for research results.” Each selected experiment will be repeated by at least three laboratories.

Amaral notes that peer review alone is not always enough to detect flaws in the methodology or execution of experiments. “But no one knows the extent of the problem,” he says. “Studies conducted in other countries for specific areas of research have found that less than half of the results published in articles are reproducible.”