Scientific research has changed: How Science Goes Right

Science appears in public media headlines mostly when new breakthroughs have been achieved or when experts share their knowledge on political issues. Last week’s print issue of The Economist, entitled “How Science Goes Wrong”, challenges this tradition by addressing several flaws in the scientific infrastructure. Problems that have been discussed within the scientific community are now being brought to public attention. Rightfully so, since scientific research and the latest findings have a large impact on all of us, with significant amounts of public resources are spent on achieving scientific results.

The Economist article identifies the following shortcomings within scientific research:

• Statistical Misinterpretation: Experimental results are often published with insufficient statistical backing.

• Data Validation: Many scientific research findings cannot be replicated due to poor data validation even within peer-reviewed publications.

• Publication Bias: The currently established data exchange infrastructure via peer-reviewed print journals drives scientists towards trendy scientific research topics with risky hypotheses and allows the sharing of only a fraction of their data and methodology.

‘Good science’ is … documenting scientific research data digitally in an electronic lab notebook.

The trouble with scientific research – What are the reasons behind?

In our opinion, the reasons leading to the problems with scientific research stated above are manifold, complex and politically as well as historically driven. In the following, we discuss a couple of them:

Publish or perish: In modern science, the one and only accepted accreditation for “good” scientific research seems to be the sheer number of publications per year in journals with the highest possible impact factor. Thus, rather than being inspired and driven by curiosity in the search for knowledge, scientists obey the demands and predetermined topics of publishers when doing scientific research.

Risky Hypotheses: In order to publish novel high-impact stories, scientists need to come up with hypotheses, which put their experiments in a favorable context. Hypotheses are necessary for the planning and interpretation of results. If a hypothesis is formulated too lurid, it may jeopardize the entire scientific research process, especially when the data to back the hypothesis is flawed in statistics.

Statistical misinterpretation: Mostly, statistical flaws arise from running too few repetitions of an experiment, which in turn, allows for a selective data interpretation that might not be reproducible by others.

Availability of Details: Current scientific publication formats mostly publish a ‘story’, focusing on the hypotheses rather than on a technical description of the scientific research processes. Very often, the raw data which led to the conclusions described are not available in scientific journal papers.

Peer Reviewing: The safeguard of scientific quality and integrity is a resource-intensive process, which is increasingly incapable of handling the amounts of scientific data that are produced within scientific research. In a recent issue of Science, the editor sent an obviously flawed scientific manuscript to a large number of journals to find that the peer-review process can indeed be very leaky.

Reproducibility: The interplay of the factors mentioned above lead to the fact that a vast number of scientific data cannot be reproduced. This fact turns into an economic problem when industries and politics cannot rely on scientific data to a satisfying extent.

Solutions to improve scientific research

With a long history of the current scientific communication infrastructure, solutions to the problems will not happen overnight. What are the most promising approaches to better scientific research?

New strategies for communicating scientific data: Emerging platforms for exchange of scientific research data nowadays offer a useful complement to classical journal publications. Publishers are already experimenting with extended formats which allow the inclusion of raw data (e.g. Elsevier’s Article of the Future or Macmillan´s Scientific Data service). A promising utilization of these options will be the publication of smaller fractions of experimental evidence (see also our blog post on nanopublications), which can help to remove hypothesis bias and instead focus more on the core of science: experimental data.

Methodological Focus: Technical details of the experimental setups should be a focus of scientific communication instead of being only a side note of information as in the established formats. Platforms like nature protocols are a good starting point, however, we are convinced that there is still room for improvements in scientific research. A more detailed, and also more visual way of communicating technical details will help scientists achieve the reproducibility that is required.

Data integration: Making scientific raw data available is the biggest step towards a trusted validation of scientific findings. With published findings being directly linked to the data sources from which they are derived it will be easier to validate and re-use research findings. For the effective re-use of large amounts of scientific research data new infrastructures for their handling will be necessary.

Our approach is to tackle the challenges where they arise: Documenting scientific research data digitally in an electronic lab notebook – directly where they are generated – is the first step towards improving research communication. We are excited to be part of an innovative ecosystem, which aims at nothing less than optimizing the way scientific procedures and results are communicated – a revolution that is long overdue.