Guia Covid-19
Imprimir Republish

Good practices

Irreproducible experiments

009_BoasPraticas_208Daniel BuenoThe journal Nature announced measures to prevent the publication of scientific articles that have to be canceled after publication when further experiments cannot replicate the same results. Although the errors originate in laboratories, noted the Nature editorial, scientific publications exacerbate the problem when they fail in their assessment of the robustness of the data or the reliability of the methods.

The most interesting measures are increased editorial space for a description of the methodologies used and closer examination of statistics. The authors will also be encouraged to provide more transparency, including the publication of raw data online. Another tool is a check during the submission of life sciences papers specifying the necessary information about sample sizes and statistical tests, among other aspects. As reported by the magazine The Scientist on its website, the editors of Nature also want researchers to reveal the sources and profiles of biological substances, such as cell lines and antibodies.

The problem with irreproducible data is growing and there is already an Internet platform, the Reproducibility Initiative (www.scienceexchange.com/reproducibility), which offers researchers the opportunity to test experiments in independent laboratories before publishing their findings. The site’s creator, Elizabeth Iorns, is executive director of the Science Exchange, an online portal that links researchers to experimental service providers.

The idea to create the service came to her when she learned that pharmaceutical companies were hiring researchers to validate published results. The portal focuses on preclinical biological studies. “There is a lot of negative information about the research reproducibility problem. I believe this is the first time someone is trying to do something positive in this regard,” said Iorns to Nature‘s website.

The Reproducibility Initiative is based in Palo Alto, California. Authors interested in having their results tested before publication must pay a specific amount for validation. But the hope is that, in time, funding agencies will also support the initiative.

The Nature Publishing Group, London, and Rockefeller University Press, New York, have already expressed support for the initiative.

Republish