Groundbreaking text mining project highlights 'gender gap' in scientific research

March 03, 2016

A project at The University of Manchester to analyse 15,000 mouse studies - the largest of its kind ever undertaken - has revealed that about half of these studies failed to report the sex and age of the mice involved, despite these being recognised as key variables that can affect the outcome of scientific studies. The project utilised text mining software developed at the University, which can analyse large collections of documents to unearth information which would otherwise have been virtually impossible to discover. The software relies on a number of rules, which automatically scan the method section of papers to identify mentions of gender and age.

The results of the project, published this week in eLife, highlight the issue of reproducibility of scientific research - around £20 billion is spent every year on research which is not reproducible, and over 80% of potential therapeutics fail in humans after being tested in mice. Previously published studies have suggested that research done on female animals may not be applicable for men, and in many of the studies analysed in this project, the animals used were overwhelmingly female. This may be due to female mice being less aggressive, which makes them easier to use in the studies. This is important, because the sexes can have markedly different responses to the same investigations - for example, in infection research. This may significantly reduce the reliability of studies, and lead to drugs that won't work for half of the population.

The reproducibility of studies often focuses on the interpretation of statistics, but this project has highlighted that the methods used may not be reported rigorously enough to assess whether they were done correctly. By looking at the methods used, it is possible to infer whether or not the statistics produced are sound, and reproducible in the future. Without knowing these methods, this cannot be inferred at all, which hampers cross-disciplinary research and longevity of data.

The project has produced a vital tool to measure the reproducibility of scientific studies, but there is a long way to go - failure to consider gender in research is still very much the norm, and according to one analysis of scientific studies published in 2009, only 45% of animal studies involving depression or anxiety and only 38 percent involving strokes used females, even though these conditions are more common in women.

"The opportunity to use text mining to cover such a broad portfolio of research was brilliant, and vital to see the bigger picture," said Sheena Cruickshank, Senior Lecturer in Immunology at The University of Manchester. "We are an interdisciplinary team, and it was this which enabled us to spot this issue and then explore it. The paper builds on several pieces of work we have done together, and highlights the importance of the scientific community to come together and define what is important in the current reproducibility crisis."

"This study has demonstrated how state-of-the-art computer science technology is instrumental for a large-scale and systematic analysis of literature," said Dr Goran Nenadic from The University of Manchester's School of Computer Science. "It avoids small sample bias, and allows us to explore the research landscape on a large scale to identify key issues in reporting details of scientific methodologies, which are necessary for reproducibility, transparency and fidelity of research."
-end-


eLife

Related Scientific Research Articles from Brightsurf:

Who's Tweeting about scientific research? And why?
Although Twitter is best known for its role in political and cultural discourse, it has also become an increasingly vital tool for scientific communication.

Weaving Indigenous knowledge with scientific research: a balanced approach
Insights from bicultural research can enhance practical applications from a palaeotsunami database to land-use decisions, according to a new review in Earth Surface Dynamics

Level of media coverage for scientific research linked to number of citations
An analysis of over 800 academic research papers on physical health and exercise suggests that the level of popular media coverage for a given paper is strongly linked to the attention it receives within the scientific community.

Spotting cutting-edge topics in scientific research using keyword analysis
Researchers from the University of Tsukuba conducted a quantitative keyword analysis of 30 million articles in the life sciences over a nearly fifty-year period (1970-2017) and found that 75% of total emerging keywords, at 1-year prior to becoming identified as emerging, co-appeared with other emerging keywords in the same article.

Calibration method improves scientific research performed with smartphone cameras
Although smartphones and other consumer cameras are increasingly used for scientific applications, it's difficult to compare and combine data from different devices.

AccessLab: New workshops to broaden access to scientific research
A team from the transdisciplinary laboratory FoAM Kernow and the British Science Association detail how to run an innovative approach to understanding evidence called AccessLab in a paper published on May 28 in the open-access journal PLOS Biology.

University of Idaho study finds scientific reproducibility does not equate to scientific truth
Reproducible scientific results are not always true and true scientific results are not always reproducible, according to a mathematical model produced by University of Idaho researchers.

Scientific research will help to understand the origin of life in the universe
Scientists from Samara University and several universities in the USA have proposed and experimentally confirmed new fundamental chemical mechanisms for the synthesis of polycyclic aromatic hydrocarbons (PAHs).

New research helps to inform the design of scientific advisory committees
At a time of 'fake news' and a growing mistrust of scientific experts, researchers at York University's Global Strategy Lab have produced new research to help inform the design of scientific advisory committees and help maximize the application of high-quality scientific research towards future policy and program decisions.

Jumping to scientific conclusions challenges biomedical research
Improving experimental design and statistical analyses alone will not solve the reproducibility crisis in science, argues Ray Dingledine in a societal impact article published in eNeuro.

Read More: Scientific Research News and Scientific Research Current Events
Brightsurf.com is a participant in the Amazon Services LLC Associates Program, an affiliate advertising program designed to provide a means for sites to earn advertising fees by advertising and linking to Amazon.com.