Reproducibility crisis

From Bioblast
Jump to: navigation, search
Bioblasts - Richard Altmann and MiPArt by Odra Noel
MitoPedia         Terms and abbreviations         Concepts and methods         MitoPedia: SUIT         MiP and biochemistry         Preprints and history


Reproducibility crisis


An experiment or study is reproducible or replicable when subsequent experiments confirm the results. This is re-search. However, we can define different types of reproducibility depending on the conditions that we use to replicate the previous work or in the information available. Our aim is to focus mostly on two different kinds1:

1. Direct: Is when we obtaining the same results using the same experimental conditions, materials, and methods as described in the original experiment. This would be the ideal reproducibility of an experiment however, it requires a very accurate description of how the original experiment was performed. Some journals are trying to resolve the reproducibility crisis improving the rigor and the excellence on the reported methods and results (e.g. STAR Methods in Cell Press).

2. Systematical: Refers to obtaining the same results, but under different conditions; for example, using another cell line or mouse strain, or inhibiting a gene pharmacologically instead of genetically. This open the door to subsequent studies to find the conditions under which an initial finding holds.

But, is there a reproducibility crisis? According to a survey conducted by Nature2 of 1,576 researchers, "52% agree that there is a significant 'crisis' of reproducibility, less than 31% think that failure to reproduce published results means that the result is probably wrong, and most say that they still trust the published literature" (Baker 2016). Chemistry and biology are the subjects with the highest share of failed attempts of reproduction of results.

Asking the researchers for the causes of this inability to reproduce published results, the top three answers are:

  • Selective reporting
  • Publication pressure
  • Low statistical power and poor analysis

The top three mentioned countermeasures are:

  • Better understanding of statistics
  • Better mentoring and supervision
  • More robust design


1. Stanley E Lazic (2016) Experimental design for laboratory biologists. Maximizing information and improving reproducibility. Cambridge University Press.
2. Baker M (2016) 1,500 scientists lift the lid on reproducibility. Survey sheds light on the ‘crisis’ rocking research. Nature 533:452–4.

Further links

Solve the reproducibility crisis

While it is probably impossible to fully prevent human self-deception and inadequate command of statistical methods, what we can do is minimize sources of error connected to the instrumental equipment and its handling:
  • Select instrumental equipment for which appropriate specifications are available.
  • Have yourself trained on your equipment and make sure you know what you (both, you and the device you operate) are doing in each step of your experiment.
  • Avoid black-box performance of software.
  • Same for data analysis: get trained on analysis software. In the best case, use software that comes with your instrument in order to minimize errors during data transfer and translation.
  • An Open Access policy fosters the establishment of an error culture and a culture of transparence in science. In this way, Open Access - as manifested in the Bioblast website (see Gentle Science - contributes to solving the reproducibility crisis.
  • Methods: Identify the methods, apparatus (manufacturer's name and address in parentheses), and procedures in sufficient detail to allow other workers to reproduce the results. Give references to established methods. - Quoted from International Committee of Medical Journal Editors.

Further links

» Validating key experimental results via independent replication
» Reproducibility Initiative

MitoPedia concepts: MitoFit Quality Control System