Nature ’s roundup of the papers and issues gaining traction on social media.

Lately, researchers have been taking to social media to hash over some big questions: just how much scientific information is out there, and how can anyone start to make sense of it?

Much conversation centred around a recent PLoS ONE article on accessing that knowledge. Using data gleaned from Google Scholar and Microsoft Academic Search, researchers at Pennsylvania State University in University Park estimated that more than 114 million English-language scholarly works — journal papers, dissertations, master's theses, books and technical reports — exist as resources online. But much of the bounty can be had only for a price. Pollution scientist Andrew Singer of the NERC Centre for Ecology and Hydrology in Wallingford, UK, noted on Twitter that “only” 24% of the works were open-access. “That's quite a low number,” he says. “It's disappointing to see how resistant we are to change.” Coincidentally, the paper came out just days before leading Chinese science agencies called for their researchers to make all articles freely available within one year of publication (see go.nature.com/hvs2yp).

Based on data from Altmetric.com. Altmetric is supported by Macmillan Science and Education, which owns Nature Publishing Group.

Khabsa, M. & Giles, C. L. PLoS ONE 9, e93949 (2014)

Contemplating those 114 million scholarly works may be daunting, but an article by two clinical researchers makes the case that scientists should also consider work that is not included in that collection. They write that bias in choosing which hypotheses to test and in selecting results for publication has undermined the movement towards evidence-based medicine. Industry-sponsored studies, they argue, are especially prone to bias: positive results are trumpeted, but negative or unexpected results are too often swept into the dustbin. The paper hit a nerve. Médicos Sin Marca, a medical group in Chile, tweeted a quote from the paper: “Indiscriminate acceptance of industry-generated 'evidence' is akin to letting politicians count their own votes.”

Every-Palmer, S. & Howick, J. J. Eval. Clin. Pract. http://dx.doi.org/10.1111/jep.12147 (2014)

A deluge of data has also swamped the medical field, prompting one opinion article to call for a new subspeciality in clinical informatics. Physicians in the field would receive training for translating electronic health records, genetic information, epidemiology and other data sources into more effective patient care. The authors note that, for this to become a reality, questions about accreditation and funding for fellowships still must be resolved. Some observers see the movement as long overdue. Adam Dunn, a health informatician at the Australian Institute for Health Innovation in Sydney, tweeted “only about a decade too late? Next thing you know, paper records will be replaced by computers!”

Detmer, D. E. & Shortliffe, E. H. J. Am. Med. Assoc. http://dx.doi.org/10.1001/jama.2014.3514 (2014)

A timely piece highlights the pitfalls of in silico research that draws on published data. Specifically, genetic data from the 1000 Genomes Project, a global effort to sequence the DNA of more than 1,000 humans, contains some unpleasant surprises, warns William Langdon, a computer scientist at University College London. After comparing 50 billion of the published sequences to known genomes, Langdon concluded that perhaps about 7% of the samples were contaminated with bacterial DNA. Geneticist Jason Moore of Dartmouth College in Hanover, New Hampshire, tweeted that the paper was “getting a lot of attention”. It's easy to see why. “As scientists use publicly available datasets rather than collecting their own samples, there is a risk of people using data and taking it as gospel,” Langdon told BioMed Central, publisher of the journal.

Langdon, W. B. BioData Mining 7, 3 (2014)

Social-media users worldwide also took note of a provocative essay that questioned the industrialized world's approach to global health. Editor-in-chief of The Lancet, Richard Horton, writes that health-outreach initiatives unfairly benefit relatively prosperous countries at the expense of poorer ones in Africa and the tropics. He points, for example, to an analysis that found that 80% of investments made by the Bill & Melinda Gates Foundation stayed in the United States. He also takes his own journal to task by citing a recent Lancet paper that examined a particular health issue across 56 countries. The authors were from Geneva, London, Gothenburg and Cape Town — a sample that didn't exactly represent the developing world. Jocalyn Clark, a medical editor working in Bangladesh, took to Twitter to voice her agreement: “We must do more to examine the power + politics of #globalhealth.”

Horton, R. Lancet 383, 1705 (2014)

Social media loves a good weight-loss story — especially one that shakes up conventional thinking. A Viewpoint in the Journal of the American Medical Association made a surprising suggestion: some people start overeating after they start gaining weight, not the other way around. It all comes down to food choice, the authors say: sugary or starchy foods with a high glycaemic index can cause underlying metabolic abnormalities to encourage fat accumulation, even when a person cuts back on calories. Such foods stimulate overproduction of insulin and increased absorption of glucose by fat cells, a metabolic state that promotes overeating. The researchers conclude that people who adopt a low-carbohydrate or low-glycaemic diet may have an easier time losing weight even if they don't cut back on calories. The paper generated comments from dieters, dietitians and more than a few physicians. Oncologist Michael Fisch at the M. D. Anderson Cancer Center in Houston, Texas, tweeted that the common phenomenon of fat gain despite cutting calories “doesn't add up with [the] calorie-centric view of #obesity”.

Ludwig D. S. & Friedman, M. I. J. Am. Med. Assoc. http://dx.doi.org/10.1001/jama.2014.4133 (2014)